Production Networks meet Industry 4.0


Fachbuch, 2020

180 Seiten


Leseprobe


Contents

1 INTRODUCTION
1.1 MANUFACTURING AND ORGANIZATIONAL TRENDS
Reconfigurable Manufacturing System
Virtual Enterprise
1.2 NEW INDUSTRIAL PLATFORM: INDUSTRY
1.3 POTENTIAL OF SMALL AND MEDIUM-SIZED ENTERPRISES
1.4 WHAT THE FUTURE HOLDS

2 PRODUCTION NETWORKS
2.1 PRODUCTION NETWORK MODELS
Competence-cell-based network
Competence cell
Networking of competence cells
Complexity-based model
Core competence cell model
2.2 PRODUCTION NETWORK LIFECYCLE
2.3 CHALLENGES FOR PRODUCTION NETWTORKS

3 PARTNER SELECTION PROBLEM
3.1 EVALUATION AND COMPARISON OF ENTERPRISES
Multi-criteria decision-making approach
PROMETHEE method
Example of usage of the PROMETHEE method
Ranking of enterprises by using the PROMETHEE method
3.2 OPTIMIZATION OF THE VIRTUAL ENTERPRISE CREATION
Optimization of the VE creation by using Ant Colony Optimization
3.3 MATHEMATICAL MODEL OF THE PARTNER SELECTION PROBLEM
Labeling Partner Selection Problem instances

4 SOLVING PARTNER SELECTION PROBLEM BY USING THE HUMANT ALGORITHM
4.1 HUMANT ALGORITHM
Metaheuristic Optimization
Multi-Objective Ant Colony Optimization
Concept and methodology of the HUMANT algorithm
HUMANT algorithm and single-objective optimization
HUMANT algorithm and multi-objective approach to single-objective optimization
4.2 HUMANT ALGORITHM AND PARTNER SELECTION PROBLEM
HUMANT algorithm and simple PSP instance
HUMANT algorithm and PSP instance wu-CS-7-19
HUMANT algorithm and PSP instance ve-CSQ-4-10
HUMANT algorithm and PSP instance mla-SQ-35-12

5 APPLICABLE REAL-WORLD SOLUTION FOR SUSTAINABLE PRODUCTION NETWORKS
5.1 SUSTAINABLE PRODUCTION NETWORKS
Production network management
Smart Collaborative Platform VENTIS
Case Study: Integration of sustainability into Partner Selection Problem
5.2 PHENOMENOLOGICAL APPROACH TO THE VIRTUAL ENTERPRISE CREATION
Procedure for creation of Push-type VE
Procedure for creation of Pull-type VE
Case Study: VE creation in the Production network of Split-Dalmatia County
Case 1: ‘The best Virtual Enterprise is a good one’
Case 2: ‘The best Virtual Enterprise is a bad one’

6 ANTHROPOLOGICAL ANALYSIS OF PRODUCTION NETWORKS
6.1 CHALLENGES OF THE TRUSTFUL COLLABORATION
6.2 ANTHROPOLOGICAL DIMENSION OF THE COLLABORATION PROBLEMS

7 CONCLUSION: PRODUCTION NETWORKS IN THE DIGITAL ERA
7.1 PRODUCTION NETWORKS AND TRACEABLE SMART PRODUCTS
Case Study: Demonstration of the vertical integration of production
7.2 PRODUCTION NETWORK OF SMART ENTERPRISES
Smart Enterprises and Lean Automation
Guidelines for Smart Enterprise development
7.3 FUTURE RESEARCH

APPENDIX: SUPPLEMENTAL MATERIAL FOR THE PARTNER SELECTION PROBLEM INSTANCES

BIBLIOGRAPHY

ABOUT THE AUTHOR

Preface

Ten years ago, I’ve started with the research on Production Networks as a part of my doctoral study. I’ve completed my Ph.D. dissertation “Phenomenological approach to the design of production networks” in October 2014. During post-doctoral period from 2014 till 2019, I’ve continued my research on Production Networks and published few scientific papers in distinguished scientific journals. So, this book represents collected works of my doctoral and post-doctoral research. I’ve never wanted to translate my dissertation on English, since I don’t believe the dissertations are interesting scientific text. I believe the book of collected works is much more appropriate format. Those who are interested in this topic, especially in the context of Industry 4.0, will find this text interesting.

One of the main contributions of my research was the design of the HUMANT (HUManoid ANT) algorithm for a priori approach to the multi-objective optimization. I’ve managed to combine Ant Colony Optimization and multi-criteria decision-making PROMETHEE method and used it to solve one of the most important optimization problems in production networks – the Partner Selection Problem. So far, HUMANT algorithm is the only fully operational algorithm which combines Ant Colony Optimization and PROMETHEE method. I’ve also contributed to a priori approach to the multi-objective optimization by designing its special case: the phenomenological approach. In the real-world application, I’ve contributed by designing the concept of Smart Collaborative Platform VENTIS and with anthropological analysis of the problems of trustful collaboration among enterprises.

At the beginning of this book, I wish to express my gratitude to God, who expressed His Love for me through my family – my greatest support and reason to get out of bed in the morning. Many thanks to my parents Prof. Nenad Mladineo, Ph.D. and Prof. Mirjana Mladineo, to my six brothers and my three sisters, and especially to my wife: Dr. Ana Mladineo, M.D. to my children: Sara, Samuel, Maksimilijan, Rafael, Ignacio, Leopold & Noel and to kids who are in God’s plan for us.

Dedicated to my scientific and spiritual inspirations:
Dr. Edith Stein, Dr.h.c. Carmen Hernández, padre Giacomo Raineri, SL.L.

For my eyes have seen Your Salvation, which You prepared in sight of all the peoples, a Light for revelation to the Gentiles, and Glory for Your people Israel.

Acknowledgements

This book represents the result of 10 years of the author’s original doctoral and post-doctoral research about the Production Networks and application of the multi-objective optimization to Partner Selection Problem.

Parts of the chapters were arranged by using the text from author’s Ph.D. dissertation and the original manuscripts of the papers published in distinguished scientific journals during these years:

1) Mladineo, M., Celar, S., Celent, L. & Crnjac, M. (2018) Selecting manufacturing partners in push and pull-type smart collaborative networks. Advanced engineering informatics, 38, p. 291-305, doi:10.1016/j.aei.2018.08.001;
2) Mladineo, M., Veza, I. & Gjeldum, N. (2017) Solving partner selection problem in cyber-physical production networks using the HUMANT algorithm. International journal of production research, 55 (9), p. 2506-2521, doi:10.1080/00207543.2016.1234084;
3) Mladineo, M., Veza, I. & Gjeldum, N. (2015) Single-Objective and Multi-Objective Optimization using the HUMANT algorithm. Croatian Operational Research Review, 6, p. 459-473, doi:10.17535/crorr.2015.0035;
4) Mladineo, M., Veza, I. & Corkalo, A. (2011) Optimization of the selection of competence cells in regional production network. Tehnicki vjesnik - Technical Gazette, 18 (4), p. 581-588; and the original manuscripts of the papers presented on distinguished scientific conferences:
5) Mladineo, M., Veza, I., Gjeldum, N., Crnjac, M., Aljinovic, A. & Basic, A. (2019) Integration and testing of the RFID-enabled Smart Factory concept within the Learning Factory. Procedia Manufacturing, 31, p. 384-389, doi.org:10.1016/j.promfg.2019.03.060. Conference on Learning Factories - CLF 2019, Braunschweig, Germany;
6) Mladineo, M., Veza, I., Gjeldum, N. & Crnjac, M. (2018) Web Information System for Sustainability Optimization of Production Networks. ICIL 2018 - Conference Proceedings. International Conference on Industrial Logistics - ICIL 2018, Beer-Sheva, Israel;
7) Veza, I., Mladineo, M. & Gjeldum, N. (2015) Managing Innovative Production Network of Smart Factories. IFAC - PapersOnLine, 48 (3), p. 555-560, doi:10.1016/j.ifacol.2015.06.139. 15th IFAC Symposium on Information Control Problems in Manufacturing - INCOM 2015, Ottawa, Canada.

Therefore, it can be stated that content of the book passed rigorous scientific review by dozen of reviewers from the world-wide academic community. So, if you wish to cite this book in your scientific work, please consider citing some of the papers above.

Author wishes to express his gratitude to all the colleagues with whom he has been collaborating and working through all these years, but especially to his mentor Prof. Ivica Veža, Ph.D.

Furthermore, some parts of this research were supported by Croatian Science Foundation under the project 1353 Innovative Smart Enterprise (INSENT).

Abstract

In the era of globalization, on the one hand, and shortage of resources and climate changes, on the other hand, the manufacturing industry is looking for new sustainable production paradigms. The production of the 21st century should be based on efficient use of resources without threatening the environment and social stability. It means it could be reasonable for society to dislocate production from one industrial area to some other less industrial area with less manufacturing waste, or to search for free capacities in other factories, instead of purchasing new equipment to extend capacities in its own factories. This kind of production outsourcing or manufacturing outsourcing can be presented as the Production Network, which represents a special case of the Virtual Enterprise.

In the past, some enterprises literally outsourced their manufacturing, like for instance Apple Computer Inc. in 1996. Today, Apple Inc. kept similar practice, so most of its smartphones are manufactured and assembled in China, but “Designed by Apple in California” as it is proudly stated on their products. This trend is followed by other US enterprises, as well.

Europe has a different strategy. The leading European industrial countries want to keep the remaining manufacturing in Europe and they want to bring back some of the manufacturing from China and other Eastern countries. They believe this would be possible through development of new industrial platforms like Industry 4.0.

There is an interesting fact that Production Networks are seen as one of the important aspects of Industry 4.0. Namely, with Industry 4.0, the Production Networks have an opportunity to finally show their true potential. The introduction of Internet of Things and Information and Communication Technology into manufacturing environment enables the creation of Cyber-Physical System for agile and short-term production planning and control inside the Smart Factory. However, Cyber-Physical System can be extended outside of the factory and used to connect with other factories thus creating Cyber-Physical Production Network of smart factories. Since unique ID for each product (barcode or RFID tag) enables the traceability, the production network management could be done with ease. In short, this is the perspective of Production Networks within Industry 4.0, and their future looks bright, very bright.

The remaining research challenges are automated negotiations and automated optimization algorithms for solving Partner Selection Problem and similar problems in process of creation of Virtual Enterprise inside the Production Network. These issues are addressed in this research through design of the multi-objective optimization algorithm with a priori approach – the HUMANT HUMANT (HUManoid ANT) algorithm, unique approach to multi-objective optimization – the Phenomenological approach, and proposal of Smart Collaborative Platform – VENTIS (Virtual ENTerprise Information System), which is very close to the idea of Social Manufacturing.

Keywords

Production Network, Industry 4.0, Cyber-Physical System, Social Manufacturing, HUMANT algorithm, PROMETHEE method, Phenomenological Method

1 INTRODUCTION

All the world's a stage, and all the men and women merely players.

(W. Shakespeare, writer)

Summary

In this chapter, the short overview of the newest manufacturing and organizational paradigms is given. Some of them are caused by globalization, and some of them are driven by technological development. From manufacturing aspect, the reconfigurable production system, as a new paradigm for the new concept of personalized production, is presented. From the organizational aspect, virtual organization is shortly described emphasizing one of its forms – the production network.

The brief introduction of new industrial platform – Industry 4.0 – is given. And, since the production networks should be based on small and medium-sized enterprises, the potential of SMEs is shortly described with the comment on distribution of SMEs in the European Union and Japan.

At the end of the chapter, the perspective is presented on what the future holds from the aspect of changes in the market demand on global and regional level.

1.1 MANUFACTURING AND ORGANIZATIONAL TRENDS

The process of globalization, and recently global economic crisis, are forcing researchers to seek for new flexible business-organizational structures. It is clear that the classical vision of the enterprise and its activities no longer corresponds to economic realities. This fact is especially true when it comes to manufacturing enterprises.

Today's manufacturing enterprises need to have a high degree of specialization in different narrow fields of work, and, at the same time, a flexible manufacturing system that will listen to and adapt to the needs of customers (a very specific ones, and a wide range ones). New manufacturing paradigm called personalized production is taking place (Figure 1.1).

Abbildung in dieser Leseprobe nicht enthalten

Figure 1.1. Manufacturing paradigms (Koren 1999)

Reconfigurable Manufacturing System

According to Koren, globalization has created a new, unprecedented landscape for the manufacturing industry, one of fierce competition, short windows of market opportunity, frequent product introductions, and rapid changes in product demand. So, the globalization is challenging, and the main challenge is to succeed in a turbulent business environment where all competitors have similar opportunities, and where customer wants personalized product. There are two types of personalized products (Koren et al. 2010):

- Product’s regional fit – Besides culture and market, regionalization must take into account additional limitations: purchasing power, climate, and legal regulations. Market research that collects and analyzes information about the habits and needs of customers in the target country is a necessity for the product’s success.
- Product personalization – Products that are manufactured to fit the buyer’s exact needs are likely to become a new source of revenue in developed countries.

Personalized production creates a new vision of a modern enterprise which needs to unite the somewhat contradictory requirements: specialization vs. flexibility. Traditional flexible manufacturing systems are not able to fulfill those requirements and to be economical in the same time. There is a need of new production systems, like the one presented by Koren in 1999: Reconfigurable Manufacturing System - RMS. The reconfigurable manufacturing system is much more flexible than Flexible Manufacturing System. According to Koren there are three main principles of reconfigurable manufacturing system (Koren et al. 2013):

The reconfigurable manufacturing system provides adjustable production resources to respond to unpredictable market changes and intrinsic system events:

- RMS capacity can be rapidly scalable in small increments;
- RMS functionality can be rapidly adapted to new products;
- RMS built-in adjustment capabilities facilitate rapid response to unexpected equipment failures.

A reconfigurable manufacturing system is designed around a product family, with just enough customized flexibility to produce all members of that family. The reconfigurable manufacturing system core characteristics should be embedded in the system as a whole, as well as in its components (mechanical, communications and control).

Virtual Enterprise

For instance, the environment of many manufacturing enterprises is characterized by unpredictable market changes. Reconfigurable manufacturing system meets these requirements by rapidly adapting capacity and functionality to new situation. Implementing RMS characteristics and principles in the system design leads to achieving the ultimate goal: “living factory” (Koren et al. 2010). The “living factory” can rapidly adjust its production capacity while maintaining high levels of quality.

Abbildung in dieser Leseprobe nicht enthalten

Figure 1.2. Creation of virtual enterprise inside production network (Veza et al. 2013)

However, such a reconfigurable structure as “living factory” can be also achieved by networking small and medium-sized enterprises (SMEs) into Production Networks. It is only important that every SME of production network is capable and wiling to be part of special cooperation inside network called Virtual Enterprise - VE (Camarinha-Matos et al. 2001). For each new product a new Virtual Enterprise is created from different SMEs (Figure 1.2).

According to L. M. Camarinha-Matos (Camarinha-Matos et al. 2007) virtual enterprise is a temporary alliance of enterprises that come together to share skills or core competencies and resources in order to better respond to business opportunities, and whose cooperation is supported by computer networks. Two key elements in this definition are the networking and cooperation, as most important part (Schermerhorn et al. 2002). Clearly, there is a tendency to describe a virtual enterprise as a network of cooperating enterprises. A number of pre-existing enterprises or organizations with some common goals come together, forming an interoperable network that acts as a single (temporary) organization without forming a new legal entity nor establishing a physical headquarter. In other words, virtual enterprises materialize through the integration of skills and assets from different firms into a single business entity. The idea of virtual enterprise compared to other types of virtual organization is shown on the following figure (Figure 1.3).

Abbildung in dieser Leseprobe nicht enthalten

Figure 1.3. Types of virtual organization (Camarinha-Matos et al. 2001)

So, in production network each SME has its autonomy, because this network is non-hierarchical. Such a network contains elements of a holistic system, such as for example: ants in nature. Each ant is an autonomous, but all the ants communicate with each other and cooperate for the benefit of the entire anthill. This is the basic idea of production network. Which means, all enterprises in the network, in addition to already existing cooperation, are willing and able to develop new cooperation on new projects forming a new virtual enterprise.

This form of networking and collaboration allows easy reconfigurations (new virtual enterprises) thus enabling the production of complex products that require high degree of specialization, and, at the same time, these products must be personalized. So, flexible and agile response is needed to meet the needs of modern customers which are very specific with wide range of needs. It is not easy to unite these contradictory requirements: specialization and flexibility, but production networks have a full potential to be able to do so. This potential is characterized by the fact that networks consists of a small cells, i.e. small and medium-sized enterprises (SMEs).

1.2 NEW INDUSTRIAL PLATFORM: INDUSTRY 4.0

The first three industrial revolutions were result of mechanization, electricity and information technology. Now, with the introduction of the Internet of Things (IoT) and Information and Communication Technology (ICT) into the manufacturing environment, the fourth industrial revolution has started and it has been called – Industry 4.0 (Kagermann et al. 2013). This new type of industrial platform is based on Smart Factory (Figure 1.4).

The Smart Factory has a completely new approach to production: smart products are uniquely identifiable, may be located at all times and know their own history, current status and alternative routes to achieving their target state. Furthermore, the embedded manufacturing systems are vertically networked with business processes within enterprises and horizontally connected to the dispersed value networks that can be managed in real time (Kagermann et al. 2013). Smart Factories allow individual customer requirements to be met and mean that even one-off items can be manufactured profitably. In Industry 4.0, dynamic business and engineering processes enable last-minute changes to production. End-to-end transparency is provided over the manufacturing process, facilitating optimized decision-making Industry 4.0 requires implementation of following features into enterprise (Kagermann et al. 2013): horizontal integration through value networks, end-to-end digital integration of engineering across the entire value chain, and vertical integration together with networked manufacturing systems.

Abbildung in dieser Leseprobe nicht enthalten

Figure 1.4. The four stages of the Industrial Revolution (Kagermann et al. 2013)

To implement these features, an enterprise must be Smart Enterprise, i.e. it must incorporate its machinery, warehousing systems and production facilities in the shape of Cyber-Physical System (smart machines, storage systems and production facilities capable of autonomously exchanging information, etc). Modern information and communication technologies like Cyber-Physical Systems, Big Data or Cloud Computing predict the possibility to increase productivity, quality and flexibility within the manufacturing industry and thus to generate advantages within the competition.

The Cyber-Physical System of Smart Factory is crucial to support new business models for manufacturers called (Meier et al. 2010): Manufacturing-as-a-Service, Industrial Product-Service Systems, or similar. Idea of Industrial Product-Service Systems is extended product, i.e. product and service integrated into single product for delivering value in use to the customer during the whole life cycle of a product. However, the idea of Manufacturing-as-a-Service is to transform manufacturer of product or part to manufacturing service provider. Both business models incorporate services into manufacturing enterprises (Vancza et al. 2011), and both require usage of state-of-the-art ICT. Because of that these models can only function around an Internet portal. The aim is to automatically generate process plans and quotations are from the technical product data provided over Internet portal. After customer assent, the prices, transportation costs and delivery times determine the choice of the standardized production plant. Hence, the importance of ICT integration into enterprise’s processes and organization is crucial. That is the reason why such a Cyber-Physical Systems are called Smart Factory.

As mentioned, one of the main requirements of today’s enterprise is to be collaborative, resulting with new virtual organizational structure called Virtual Enterprise (Camarinha-Matos et al. 2011). Virtual Enterprises are usually realized through Production Networks or Manufacturing Networks. Production Networks represent the whole value adding process (marketing, product development, production planning, manufacturing, assembly, quality control and service), so they represent vertical integration of enterprises (Sturgeon 2002). And Manufacturing Networks represent only manufacturing and assembly process, so they represent horizontal integration of enterprises (Ferreira et al. 2016).

Finally, the main features of Smart Factory can be summarized into the following:

- Smart personalized product – Requires flexibility and high level of ICT integration into manufacturing system to produce a product which fits the customer’s exact needs and which is uniquely identifiable, may be located at all times and knows its own history, current status and alternative routes to achieving customer. It can be realized through Reconfigurable Manufacturing System (Koren 1999) or Industry 4.0 Smart Factory (Kagermann et al. 2013).
- Product and service provider – Ability to offer extended products: product and service integrated into single product for delivering value in use to the customer during the whole life cycle of a product; or to offer manufacturing as a service and become manufacturing service provider (Meier et al. 2010). It can be realized through specialized Internet portals and Cloud computing (Tao et al. 2014).
- High level of collaboration – Also requires high level of ICT integration to support collaborative product development, collaborative manufacturing and all other value adding processes. It can be realized through vertical integration called Production Networks (Sturgeon 2002), or through horizontal integration called Manufacturing Networks (Ferreira et al. 2016).

1.3 POTENTIAL OF SMALL AND MEDIUM-SIZED ENTERPRISES

Small and medium-sized enterprises (SMEs), which primarily apply new technologies with ease, were recognized by the European Union as the key factors of transformation of the European “knowledge-based economy” (KPMG Special Services, 2003). According to the EU, the enterprise is classified as SME if: it's independent, have fewer than 250 employees and balance sheet total not exceeding €43 million. In addition, SMEs can be parsed to very small (micro) enterprises having fewer than 10 employees. A further reason of EU investment in SMEs is their share in the total number of enterprises: 99.8% (Figure 1.5).

Abbildung in dieser Leseprobe nicht enthalten

Figure 1.5. Structure of industrial enterprises in the EU (Veza et al. 2013)

A particular potential are micro enterprises that have the productivity level of 62% which is up to 25% less than productivity of SMEs (Müller, 2006). This lack of productivity is primarily classified as unused capacity or lack of work. When it comes to the Republic of Croatia, the structure of industrial enterprises is similar (Figure 1.6).

Abbildung in dieser Leseprobe nicht enthalten

Figure 1.6. Structure of industrial enterprises in the Republic of Croatia (Veza et al. 2013)

The only difference is that in the Republic of Croatia half of employees in the industrial sector are in LEs, while in the EU about a third of employees are in LEs.

However, trends in the period 2004-2007 show an increase in the number of SMEs by 39.6% (Table 1.1) and an increase in number of employees by 22.6% (Table 1.2) (Mladineo et al. 2011). While in the same period the number of LEs remained the same, and number of their employees declined by 2.1%.

Table 1.1. Number of industrial enterprises in the period 2004-2007 (Veza et al. 2013)

Abbildung in dieser Leseprobe nicht enthalten

Table 1.2. Number of employees in industrial enterprises in the period 2004-2007 (Veza et al. 2013)

Abbildung in dieser Leseprobe nicht enthalten

The general conclusion is that the Republic of Croatia is catching up with EU trends in the structure of industrial enterprises, as well as in the structure of their employees. Therefore, the EU strategy for the development of SMEs should begin to apply in Croatia.

Similar structure of SMEs can be found in country with some of the world's most advanced production systems: in Japan (Figure 1.7). That clearly shows that networking of SMEs has a global potential and it represents a future of production systems.

Abbildung in dieser Leseprobe nicht enthalten

Figure 1.7. Structure of industrial enterprises in Japan (Veza et al. 2013)

In 2011, to stimulate research and development of SMEs production networks, European Union has funded six FP7 projects with more than 37 million € budget: ADVENTURE, BIVEE, ComVantage, GloNet, IMAGINE, and VENIS. It appears that one of the key strategies of development of SMEs is their networking in regional production networks

1.4 WHAT THE FUTURE HOLDS

No one knows what the future holds, but some forecasts can be made. Especially forecasts regarding the market demand. On the global level there is an increase in the demand on almost every market. But on regional level, for instance in USA or Europe, the demand of many markets is decreasing. It is caused by the population decline. Because, if a market in one area has a 10 million customers, a population decline to 9 million customers leads to decrease of every market in that area. On the other hand, there is a significant increase in consumption and shorter lifecycles of a products, but it cannot compensate significant population decline. Most of the people won’t have two bank accounts instead of one, or three cellphones, or buy dozen of milk bottles per week for themselves.

According to Macunovich, 80% of economic crises, in the past century, were preceded by the decline in trend or negative trend of the population aged 15-24 years (Macunovich 2012). The issue of population aged 15-24 is very delicate, since they represent the future customers and a new workforce that will satisfy the future increase of the demand. If their number is smaller than a previous generation, an economic crisis is triggered by that disturbance. So, the significant population decline, means a significant decrease of demand. The only question is: how much the population is declining?

First of all, it is important to setup a proper long-term measure for changes in population. The difference between number of births and number of deaths is a short-term measure, so something else must be used. The fertility rate, or the average number of children that would be born to a woman over her lifetime, is a proper long-term measure, because it tells what the future holds. For instance, fertility rate of 1.5 births per woman means that, in generational turnover (approx. every 50 years), the population will decrease at least for 25% (Figure 1.8). It means that the whole market in that area will be smaller for 25%!

Abbildung in dieser Leseprobe nicht enthalten

Figure 1.8. Population trend for fertility rate 1.5 (births per woman)

Unfortunately, the global map of fertility rates shows that most of the countries of the world are “fading away”, i.e. their population is decreasing rapidly (Figure 1.9). Only Africa, Middle East, parts of South America and Southeast Asia, plus India, Mexico, Mongolia, Kazakhstan and few other countries have significant population increment (World Bank 2019). So, the next question is: how fast the population is declining? To answer it, a simple simulation has been made starting with the world of 1 man and 1 woman as an initial population (Figure 1.10).

Abbildung in dieser Leseprobe nicht enthalten

Figure 1.9. Fertility rates (births per woman) in 2017 (World Bank 2019)

Abbildung in dieser Leseprobe nicht enthalten

Figure 1.10. Simulation of population trends for different fertility rates

Figure 1.10 shows that fertility rate 1.00 means that population will shrink 8 times in three generational turnovers (150 years). On a real-world scale, it means from 1 billion to 125 million in 150 years. Such an effect would be devastating for the world’s economy. One of the most terrifying population declines was during the Great Plague in the 14th century when about third of Europe’s population died. However, bad demographic trends in Europe could produce more devastating effect for society and economy, in the near future.

Regarding the world, with fertility rate about 2.43 in 2017 (World Bank 2019), it seems that the world is not in the stage of population extinction, but rather in phase of stagnation or expansion. Nevertheless, this is not completely true. The fertility rate 2.09 is needed for mathematical stagnation of population. The two parents are replaced by two child in generational turnover, but, since there is an unequal distribution of sex, the 2.09 is needed instead of 2.00. To have 100 girls, another 109 boys need to be born, that’s why it is 2.09. So, in developed countries where expected lifetime is more than 75 years, the fertility rate needed for stagnation is about 2.10. But, in less developed countries, with huge mortality rate of infants and children, the fertility rate needed for stagnation is about 3.50! Since today’s world’s average fertility rates consists mainly of less developed countries, this number needs to be normalized. It is difficult to calculate exact number of normalized fertility rate, but it could be approximated to about 1.80 - 1.90 (Table 1.3), which means that the world’s population will soon come to extinction phase, if nothing changes.

Table 1.3. Approximation of the normalized fertility rate of the world

Abbildung in dieser Leseprobe nicht enthalten

Since the normalized world’s total fertility is between 1.80 and 1.90, the demographic trend of the world is moving toward extinction. In couple of decades the world population will stop to grow and start to decline. According to Figure 1.10, the decline can be very dramatic. Since the population is an autocatalytic system, if the parameters are changed, it collapses into itself. The main parameter of population is the fertility rate, and it is changing very fast in the negative direction.

The main problem of these dramatic population changes will be economy. It is impossible to have sustainable economic system in the environment where market demand is shrinking each day. Today, most of the enterprises have long-term plans based on increase of sale about 5-6 % each year. But, what if market is decreasing for 1-2 % each year. It means that these enterprises will be missing more than 7 % of their sale increase. That fact can lead them to bankruptcy, and lead some regional economy into crisis. The capitalism itself is based on increase of sales, supported by increase of households’ incomes. And these two increases are mutually dependent (Figure 1.11): the increase of sale means greater revenue for the enterprise; the greater revenue means better wages for the workers; better wages mean greater household income; greater income means greater consumption; and greater consumption means increase of sale, and the whole process repeats. This is the way how capitalism was functioning for the last 100 years. Starting with Henry Ford who gave decent salaries to his workers, knowing that, with decent income, they will become his customers, also.

Abbildung in dieser Leseprobe nicht enthalten

Figure 1.11. Market growth in 20th century and problem of market growth in 21st century

Today, there is a huge problem of multiplication of money without value adding activities through financial engineering (Figure 1.11). It is caused by the lack of market growth, which is caused by negative demographic trends. Unfortunately, the financial engineering was one of the main triggers of the global economic crisis in 2017.

On the other hand, the greatest value adding activity is manufacturing. But, industrial production cannot be based on increase of output, i.e. on increase of sale, as it used to be in the 20th century. Today, it needs to grove in other ways. One way is to reduce manufacturing cost through new technologies, but also through new organizational trends such as production networks. The other way is to design innovative products which will have higher price on the market. Since, enterprises are lacking the innovation capacities, networking with other enterprises into production networks can help them to develop and produce new innovative products.

2 PRODUCTION NETWORKS

Alone we can do so little, together we can do so much.

(H. Keller, writer)

Summary

In this chapter, the alternative for reconfigurable production – the production network – is presented. Some of the production networks theoretical models are described, like the competence-cell-based network, complexity-based model, and core competence cell model.

Production network lifecycle is compared to product lifecycle and the literature review on production network lifecycle is given.

Furthermore, the perspective of the future of production network, emphasizing new production-organizational paradigm “Production-as-a-Service” and “Manufacturing-as-a-Service”, is given together with description of production networks advantages and challenges.

2.1 PRODUCTION NETWORK MODELS

In the late ’90s, following the trend of globalization and development of Internet and communication technology, a new type of networked organization took interest of scientific community. It was an organization which is geographically dispersed and networked through virtual cyber environment. Therefore, this new form of organization was called virtual organization (Hedberg et al. 1997), network of enterprises (Villa 1998), or, simply, virtual enterprise (Leigh Reid et al. 1996). On the other hand, new paradigms in manufacturing, like agile manufacturing, required new type of production systems, and so, for instance, reconfigurable production system was presented in 1999 by Koren et al.

Nevertheless, at the same time, idea of realizing agility and re-configurability through Production Networks was also born (Leigh Reid et al. 1996; Wu, Mao and Qian 1999). Very soon, production networks were recognized as a new industrial model for world leading manufacturing industries, like the one in USA (Sturgeon 1999), and also in EU (Roth et al. 2001; Camarinha-Matos and Afsarmanesh 2002). Furthermore, many scientific researches (Müller 2006) and projects (Markaki et al. 2013) about production or manufacturing networks were made in last 20 years. In these researches many different names are used for different or similar type of production networks: global production networks (Jaehne 2009), regional production networks (Gerber, Dietzsch and Althaus 2004), competence-cell based production networks (Müller 2006; Matt 2007), reconfigurable collaborations (Schuh et al. 2008), dynamic manufacturing networks (Markaki et al. 2013), or just, virtual enterprises (Leigh Reid et al. 1996; Wu, Mao and Qian 1999; Camarinha-Matos and Afsarmanesh 2002). Generally, there is a difference between production and manufacturing networks, because production networks represent vertical collaboration inside value-adding chain (design, prototyping, manufacturing, quality services, etc.), and manufacturing networks represent horizontal collaboration (i.e. only manufacturing and assembly), but there are also many similarities.

Furthermore, the concept of production networks is a research field of scientists all over the world. In EU: Germany (Bölt et al. 2000; Gerber et al. 2004; Neuberta et al. 2004; Roth et al. 2005; Jähn et al. 2006; Müller et al. 2006; Ackermann 2007; Jaehne et al. 2009; Kampker et al. 2010; Ganß et al. 2011; Lau et al. 2011), Belgium (Vancza et al. 2011), Hungary (Schuh et al. 2008), Portugal (Camarinha-Matos et al. 2001; Pinto Leitão 2004), Netherlands (Camarinha-Matos et al. 2007), Spain (Giret et al. 2009), Italy (Villa et al. 1998; Corvello et al. 2007; Manzini et al. 2011), Greece (Assimakopoulos et al. 2003) and Croatia (Mladineo et al. 2011); in USA (Leigh Reid et al. 1996; Sturgeon et al. 2002); in China (Hongzhao et al. 2005), Japan (Yamawaki 2002) and South Korea (Choi 2005); in Columbia (Micán et al. 2011) and Brasil (Lima et al. 2011).

Since only few concepts (models) have been completely developed to be implemented in practice, only three models will be presented: competence-cell-based network, complexity-based model, and core competence cell model.

Competence-cell-based network

According to Müller (Müller et al. 2006) the current mode of cooperation is mostly hierarchical. In most cases, the two components of co-operation, operation and communication, are delegated to different hierarchical levels: operation to the shop floor level and the inter-organizational communication to the management level. The "redirection" of communication regarding the co-operative production process, produces process losses and prevents direct feedback from shop floor to shop floor.

So the idea was to seek for the non-hierarchical cooperation concept of networking of small and medium-sized enterprises. Such a network is called competence-cell-based network (Müller et al. 2006). Each enterprise represents a single competence-cell, since the employees of each enterprise have a specific set of competencies. However, each competence-cell retains its autonomy, because this network is non-hierarchical.

This concept is particularly interesting for application in Croatia, since the economy of Croatia has very similar problems with slow recovery from real-socialist production system, like ex-Eastern Germany.

Competence cell

A competence cell (Müller et al. 2006) is considered as smallest autonomous indivisible performance unit of value adding. The human competences of each cell are obtaining crucial importance, so the human is in the center of the competence cell. There are different types of competence cells, covering the whole value adding process: marketing competence cells, product development competence cells, production planning competence cells, manufacturing competence cells, assembly competence cells and quality control / service competence cells. E. Müller et al. developed generic model of the competence cell (Müller et al. 2006) (Figure 2.1) based on general production theory, specific networking requirements and investigations into the business processes of marketing, product development, production planning, manufacturing & assembly, logistics and quality control & service. The generic model consists of (Müller et al. 2006):

- the competence of humans, arranged according to professional, methodical, social and personnel competences;
- available resources;
- the fulfilled task or executed function.

With this function a business entity is transformed and a certain performance is achieved. For a complete technical description the aspects of dimension and structure were supplemented. Function, competence, resource and marketable performance serve as criteria to further operationalize the required decomposition of the competence cell. Although E. Müller et al. differ several types of competence-cells.

Abbildung in dieser Leseprobe nicht enthalten

Figure 2.1. Generic model of competence cell (Müller et al. 2006)

Networking of competence cells

The vision of competence-cell-based networking is based on the model consisting of three levels (Figure 2.2) (Roth et al. 2005; Ackermann et al. 2007). From loose infrastructural and mental relations in a regional network (level I) there initially emerges an institutionalized competence network, based on competence cells (level II). The actual creation of value takes place in a production network (level III), and the production network in this model represents virtual enterprise. It is initiated by customer needs transformed into business request.

Abbildung in dieser Leseprobe nicht enthalten

Figure 2.2. Three level model of production networks (Roth et al. 2005)

Complexity-based model

According to G. Schuh et al. (Schuh et al. 2008) it is possible to manage dynamic reconfigurable collaborations in industry by defining generic model of complexity. Reconfigurable collaboration is a type of production network. There are several abstract complexity drivers that can cause problems in collaboration networks. The main drivers are as follows:

- uncertainty (e.g., limited information);
- dynamics (e.g., dynamic changes);
- multiplicity (e.g., a large number of participating elements and influencing factors);
- variety (e.g., many types of elements);
- interactions (e.g., communication loads);
- interdependencies (e.g., feedback loops).

Schuh suggests modeling the dynamic behavior of a production network as a Complex Adaptive Systems (CAS) (Schuh et al. 2008). A CAS can be considered a multi-agent system with seven basic elements in which ‘‘a major part of the environment of any given adaptive agent consists of other adaptive agents, so that a portion of any agent’s efforts at adaptation is spent adapting to other adaptive agents’’. Agents may represent any entity with self-orientation, such as cells, species, individuals, enterprises or nations. Environmental conditions change, due to the agents’ interactions as they compete and cooperate for the same resources or for achieving a given goal. This, in turn, changes the behavior of the agents themselves.

Furthermore, computer-based simulations can be applied to evaluate these systems. Simulations can help observing and investigating, e.g., how (potentially simple) individual behavior rules may emerge and give rise to complex (and often unpredictable) collective behavior. Additionally, the stability of these kinds of systems together with the effects of uncertainties (such as the lack of precise market forecasts as well as personal contacts) could also be evaluated by simulations.

Core competence cell model

Matt, like Schuh, is dealing with problem of the structural complexity of growing organizational systems like production networks (Matt 2007). To reduce structural complexity he reduces cells (enterprises) to three basic types: Core Competence Cells (3C). The core competence cells are defined as:

- Dealer (DL) is defined as a person or a enterprise that buys and sells goods or services;
- Producer (PR) aims at the minimization of manufacturing costs and the optimization of flexibility;
- Service Provider (SP) aims at ‘‘selling’’ his collaborators most profitably.

The central success factor of a network cell is to strictly focus on one core competence type and to force and professionalize it by entrepreneurial incentives. The different success mechanisms of DL, PR, and SP show once again that their mixing increases complexity and causes losses in efficiency. To maintain the strict core competence type focus means to inherit a cell’s ‘‘success DNA’’ to its spin-off in the case of a cell division.

According to D. T. Matt (Matt 2007), it can be stated that the proposed 3C model helps to reduce the entire organizational complexity from a structure perspective. It allows an organization to flexibly adapt to changing environmental conditions and thus promotes sustainable business growth within an organizational network.

2.2 PRODUCTION NETWORK LIFECYCLE

As it was mentioned, the idea of virtual enterprise differs from other types of virtual organization. According to Camarinha-Matos virtual organizations can be described as (Camarinha-Matos et al. 2001):

- extended enterprise is the closest to virtual enterprise, however it is better applied to an organization in which a dominant enterprise extends its boundaries to all or some of its suppliers (automotive industry);
- virtual enterprise can be seen as a more general concept including other types of organizations, namely a more democratic structure in which the cooperation is peer to peer (i.e. extended enterprise can be seen as a particular case of virtual enterprises);
- virtual organization is a concept similar to a virtual enterprise, comprising a network of organizations that share resources and skills to achieve its mission / goal, but not limited to an alliance of enterprises, for example virtual organization could be a virtual municipality organization, associating via a computer network, all the organizations involved in a municipality (city hall, municipal water distribution services, internal revenue services, public leisure facilities, cadaster services, etc.);
- networked organization is the most general term referring to any group of organizations inter-linked by a computer network, but without necessarily sharing skills or resources, or having a common goal.

Since the virtual enterprise has been defined as a something non-hierarchical and temporary, it is important to analyze lifecycle of virtual enterprise, i.e. lifecycle of production network.

Abbildung in dieser Leseprobe nicht enthalten

Figure 2.3. Relation between product lifecycle and virtual enterprise lifecycle (Mladineo et al. 2017)

Few researches have made phenomenological research of virtual enterprise lifecycle. In literature can be found virtual enterprise lifecycle of Camarinha-Matos (Camarinha-Matos et al. 2001) and Leigh Reid (Leigh Reid et al. 1996). Generally, virtual enterprise lifecycle consists of: customer request which triggers the creation of virtual enterprise, creation process, operation process and dissolution process. Therefore, virtual enterprise lifecycle is in close relation with product lifecycle, from the Product Lifecycle Management (PLM) point of view. Because, the creation of VE is triggered by the same cause that triggers the development of the new product: customer requests (market demand). When lifecycle of a product has ended, it is also the end of the virtual enterprise lifecycle. The collaboration between enterprises that started to develop and produce new product ends with the end of the product’s lifecycle (Figure 2.3). Additionally, this collaboration can be extended if enterprises decide to develop new version of product, but then new product lifecycle, inside the same virtual enterprise, begins.

In the following table virtual enterprise lifecycle of Camarinha-Matos and Leigh Reid are mutually compared (Table 2.1).

Table 2.1. Virtual enterprise lifecycles comparison (Veza et al. 2013)

Abbildung in dieser Leseprobe nicht enthalten

2.3 CHALLENGES FOR PRODUCTION NETWTORKS

Despite all the research through 20 years, non-hierarchical production networks consisting of autonomous enterprises in practice, i.e. in the real world, are, in the most cases, only R&D projects or living labs. Although some global corporations are using their own production networks across continents (Netland 2014), it is not at all similar concept to the original idea of non-hierarchical production networks. The main reason why production networks did not become operative in practice, probably, lays in low level of ICT integration into production systems of the enterprises. Koren in 1999 pointed out that in order to achieve re-configurability, machines should have universal and standardized interfaces including both, hardware and software. But this begun to happen only few years ago, especially through development of new industrial platforms like Industry 4.0 (Kagermann Wahlster and Helbig 2013). Now, with Industry 4.0, the level of ICT integration into production systems becomes high enough to support fully operative production networks. Since platform Industry 4.0 has its main foundations in cyber-physical interpretation of production systems, that ability allows easy control of networked production systems. Furthermore, this new type of production network is now called Cyber-Physical Production Network (CPPN) (Monostori 2014; 2015), or, similarly, Socio-Cyber-Physical System in Production Network (Morosini Frazzona 2013). It is also important to point out that CPPNs are seen as one of three key elements of the platform Industry 4.0 (Figure 2.4).

Abbildung in dieser Leseprobe nicht enthalten

Figure 2.4. Production Networks as one of three key elements of Industry 4.0 (Roßgoderer and Piepenbrock 2014)

Therefore, it can be stated that production networks represent the future of the manufacturing. They especially represent possible solution for the new production-organizational paradigm “Production-as-a-Service” and “Manufacturing-as-a-Service”. This new paradigm intends to fulfill very specific needs and requirements of modern customer, i.e. to produce one piece of specific product for only one customer.

For instance, if a customer needs a special custom made motorcycle (Figure 2.5), he/she can buy only a similar motorcycle from motorcycle producer for reasonable price. If the customer wants exactly the same motorcycle as imagined one, he/she needs to buy it from custom made motorcycle producer. However, the price will not be reasonable, it will be very expensive. A custom made motorcycle for reasonable price is something that only production network can produce. And it represents main competitive advantage of production networks. But it also shows that production networks can function like “Production-as-a-Service”.

However, production networks are virtual organizations and there is a problem of stability of such an organizational formation. So the key challenge is to have production network formed of good and trustful enterprises, i.e. partners. And that is the reason why is solving of Partner Selection Problem is one of the key challenges for successful management of production networks.

In the future, research focus will be on the design of fast and accurate algorithms for solving Partner Selection Problem, especially for complex production processes with multiple criteria (objectives). Today’s algorithms are taking lot of time to solve complex production processes. In the future this needs to be solved to be as fast as a Web Service used in Web application for management of production network.

Abbildung in dieser Leseprobe nicht enthalten

Figure 2.5. Competitive advantage of production network

It is also important to highlight that the management of production networks requires knowledge about information technology ant it also requires knowledge about some management tools like multi-criteria decision-making and analysis. All these issues need to be taken into account when drawing a path into the future for production networks.

3 PARTNER SELECTION PROBLEM

Good decisions come from experience and experience comes from bad decisions.

(M. Twain, writer)

Summary

This chapter deals with one of the most important optimization problems linked with production networks – the Partner Selection Problem. In the first part, the general evaluation and comparison of the partners (enterprises) is presented, emphasizing the multi-criteria decision-making PROMETHEE method.

Later, the optimization of the virtual enterprise creation, i.e. the selection of partners in the phase of virtual enterprise creation, is presented and problem with two criteria (objectives) is solved.

At the end, the mathematical model for Partner Selection Problem with multiple objectives (criteria) is presented together with proposal for labeling different problem instances.

3.1 EVALUATION AND COMPARISON OF ENTERPRISES

The problem of the selection of enterprises in production network, also known as Partner Selection Problem (Wu et al. 1999; Fischer et al. 2004; Wu et al. 2005; Mourtzis 2010; Ma et al. 2012), arises when the production process is parsed into technological processes (operations) that need to be completed to produce a product. In fact it is very likely that the same technological operations can be done by two or more different enterprises in the network. The question is: which enterprise to choose? Therefore, it is obvious that, before the selection process, enterprises need to be evaluated on the basis of their performances and competences (Fischer et al. 2004; Mladineo et al. 2011). In this way, enterprises with the highest ratings will be selected and they will form new virtual enterprise.

Figure 3.1 shows a production problem, i.e. a production process with possible alternatives, and its optimal solution (Mladineo et al. 2013). The problem can be presented as a network graph that has a beginning or source (order) and end or drain (delivery). The network is formed of enterprises, and each technological operation is presented by enterprise that can perform it. Each enterprise has its rating, and higher rating is better.

Abbildung in dieser Leseprobe nicht enthalten

Figure 3.1. Alternatives of production process and its optimal solution (Mladineo et al. 2013)

According to Figure 3.1, for each technological operation (turning, milling or assembly) an enterprise with higher rating is selected. Hence, the production process will be realized using the best combination of enterprises. The combination of enterprises is one new virtual enterprise. However, the evaluation of enterprises performances is needed to solve the problem of the selection of enterprises in production network, or partner selection problem.

Multi-criteria decision-making approach

Since, the evaluation and comparison of enterprises is a multi-criteria problem, in this chapter a special multi-criteria decision-making (MCDM) method is used: PROMETHEE method. However, to completely solve partner selection problem a combination of metaheurstic optimization algorithms and MCDM methods must be used. In the literature different approaches using different multi-criteria methods or metaheuristics can be found: Fischer (Fischer et al. 2004) and Jung (Jung et al. 2011) are using AHP (Analytic Hierarchy Process) method; Lanza (Lanza et al. 2010) and Mladineo (Mladineo et al. 2013) are using PROMETHEE method; Gao (Gao et al. 2006) is using Particle swarm algorithm; Yu (Yu et al. 2011) is using TOPSIS (Technique for Order of Preference by Similarity to Ideal Solution) method; Chuanga (Chuanga et al. 2009) and Zhao (Zhao et al. 2006) are using combination of DEA (Data Envelopment Analysis) and Genetic algorithm, and many others are using different evolutionary or multi-agent approaches (Choi et al. 2007; Wang et al. 2009; Nayak et al. 2010; Tao et al. 2010; Lanza et al. 2011; Zhang et al. 2012). It is also important to highlight the Partner Selection Problem is more complex than similar optimization problems like the Assignment Problem and the Job-Shop Problem or Job-Shop Scheduling Problem, therefore same algorithms cannot be used.

PROMETHEE method

The problem of the selection or the ranking of alternatives submitted to a multicriteria evaluation is not an easy problem, since, usually, there is no optimal solution (Brans et al. 1984). No alternative is the best one on each criterion. In the recent decades several decision aid methods or decision support systems have been proposed to help in the selection of the best compromise alternatives. In this chapter the PROMETHEE (Preference Ranking Organisation METHod for Enrichment Evaluations) method was chosen for treating multicriteria problem (Brans et al. 1984; 1986; 1994). This method is known as one of the most efficient but also one of the easiest in the field. PROMETHEE method is well accepted by decision-makers because it is comprehensive and has the ability to present results using simple ranking (Brans et al. 1984).

Abbildung in dieser Leseprobe nicht enthalten

Figure 3.2. Input matrix for PROMETHEE method

An input for PROMETHEE method is a matrix consisting of set of potential alternatives (actions) A, where each a element of A has its f(a) which represents evaluation of one criteria (Figure 3.2). Each evaluation fj(ai) must be a real number.

The preference structure of PROMETHEE method is based on pairwise comparisons (Brans et al. 1984; 1986; 1994). The deviation between the evaluations of two alternatives on a particular criterion is considered. For small deviations, the decision-maker will allocate a small preference to the best alternative and even possibly no preference if he considers that this deviation is negligible. The larger the deviation is, the larger the preference is. There is no objection to consider that these preferences are real numbers varying between 0 and 1. This means that for each criterion the decision-maker has in mind a function:

Abbildung in dieser Leseprobe nicht enthalten

where:

Abbildung in dieser Leseprobe nicht enthalten

and for which:

Abbildung in dieser Leseprobe nicht enthalten

In case of a criterion to be maximized, this function is giving the preference of a over b for observed deviations between their evaluations on criterion fj(). It should have the following shape (Figure 3.3).

Abbildung in dieser Leseprobe nicht enthalten

Figure 3.3. Preference function

The preferences equal 0 when the deviations are negative. The following property holds:

Abbildung in dieser Leseprobe nicht enthalten

For criteria to be minimized, the preference function should be reversed or alternatively given by:

Abbildung in dieser Leseprobe nicht enthalten

The pair {fj(), Pj(a,b}) is the generalized criterion associated to criterion fj(). Such a generalized criterion has to be defined for each criterion. In order to facilitate the identification six types of particular preference functions have been proposed (Table 3.1) (Brans et al. 1984; 1986; 1994).

Table 3.1. Types of generalized criteria (preference functions)

Abbildung in dieser Leseprobe nicht enthalten

First, method PROMETHEE I ranks actions by a partial pre-order, with the following dominance flows (Figure 3.4) (Brans et al. 1984; 1986; 1994):

Abbildung in dieser Leseprobe nicht enthalten

Abbildung in dieser Leseprobe nicht enthalten

where a denotes a set of actions, n is the number of actions and Π is the aggregated preference index defined for each couple of actions. The PROMETHEE I method gives the partial relation.

Abbildung in dieser Leseprobe nicht enthalten

Figure 3.4. PROMETHEE I partial pre-order

Then, a net outranking flow is obtained from PROMETHEE II method which ranks the actions by total pre-order (Figure 3.5) (Brans et al. 1984; 1986; 1994):

Abbildung in dieser Leseprobe nicht enthalten

In the sense of priority assessment net outranking flow represents the synthetic parameter based on defined criteria and priorities among criteria. Usually, criteria are weighted using criteria weights wj and usual pondering technique:

Abbildung in dieser Leseprobe nicht enthalten

Furthermore, different sets of criteria weights can be used and then each set represents one scenario. And usually MCDA problems have more than one scenario.

Abbildung in dieser Leseprobe nicht enthalten

Figure 3.5. PROMETHEE II total pre-order

Example of usage of the PROMETHEE method

Here the PROMETHEE method is demonstrated on the problem of selection of location for new power plant. There are 6 different locations (alternatives) and there are 3 criteria: manpower (number of personnel), power of power plant (MW) and cost of construction (M€). For each criterion preference function and all parameters are chosen (Figure 3.6). Problem is solved by PROMETHEE I method (Figure 3.7) and PROMETHEE II method (Figure 3.8) using special software called Visual PROMETHEE (http://www.promethee-gaia.net/). The weight for each criteria is determined by group of experts.

Abbildung in dieser Leseprobe nicht enthalten

Figure 3.6. Input matrix for the problem

Abbildung in dieser Leseprobe nicht enthalten

Figure 3.7. PROMETHEE I partial pre-order of locations

Abbildung in dieser Leseprobe nicht enthalten

Figure 3.8. PROMETHEE II total pre-order of locations

Ranking of enterprises by using the PROMETHEE method

Special case of virtual enterprise evaluation occurs when partners are a priori selected (Mladineo et al. 2013), i.e. some of enterprises are willing to be part of new virtual enterprise, and some are not. In this special case it is possible to have small number of different combinations of partners of new virtual enterprise. So there is need to mutually compare couple of virtual enterprises. It rises following questions: Which virtual enterprise is the best one? How much is one virtual enterprise better than others? The first question is ranking problem, and the second question is sorting problem (Mladineo et al. 2013). However, pre-requisition of virtual enterprise evaluation is the evaluation of enterprises that can become a part of new virtual enterprise.

To evaluate and rank enterprises it is necessary to design a set of criteria that will represent all the important parameters which need to be taken into account when performing ranking. It should be primarily taken into account that there are parameters that change each time when a new production network is formed for a new product, and there are parameters that do not change so often. Therefore, a set of criteria which will be used can be divided into two sets (Mladineo et al. 2011):

- Dynamic criteria: criteria whose values change for each enterprise depending upon the offer for particular product production or development (an example of such criteria is the price of the product);
- Static criteria: criteria whose values do not change so often, or at most a few times a year (an example of such criteria is a technology of enterprise).

A set of dynamic criteria includes offer that enterprise offered when a new production network for a new product is formed. That offer is usually made up of two elements: the price per piece and the day of delivery. Static set of criteria can be further divided onto:

- Competence criteria: criteria covering all the competencies of the enterprise: technical, organizational and human competence;
- Economic criteria: criteria that consider economic feasibility or risk of involving enterprise into production network;
- Sociological criteria: criteria which analyze sociological impact of involving certain enterprise in the production network.

After criteria and theirs parameters have been determined, an input matrix for PROMETHEE method, i.e. criteria evaluation for each action (enterprise), is made using data gathered in special questionnaire. This questionnaire was sent to the production enterprises of Split-Dalmatia County. In the following figures (Figure 3.9 and Figure 3.10) an input matrix for 7 enterprises is shown. However, star names are used instead of real names of enterprises.

Abbildung in dieser Leseprobe nicht enthalten

Figure 3.9. Input matrix for dynamic and competence criteria

Abbildung in dieser Leseprobe nicht enthalten

Figure 3.10. Input matrix for economic and sociological criteria

PROMETHEE method was performed using 4 different predefined scenarios (Figure 3.11). A set of weights for each scenario was determined by experts.

Criteria preference function type and preference thresholds where obtained using in-built function “Preference Function Assistant” of Visual PROMETHEE software. Following results were obtained (Figure 3.12 and Figure 3.13).

Abbildung in dieser Leseprobe nicht enthalten

Figure 3.11. Different scenarios for different complexity of product and/or production

This analysis showed that 3 enterprises (Beta Ursae Minoris, Alpha Ophiuchi and Beta Aquarii) are dominant in comparison with other enterprises. However, in different scenarios these 3 enterprises are taking turns at the top. For example: for simple product and small series the best enterprise to realize that production process is Alpha Ophiuchi. However, for complex product and large series the best enterprise to realize that production process is Beta Ursae Minoris.

Abbildung in dieser Leseprobe nicht enthalten

Figure 3.12. Ranking results and criteria weights for scenario A and B

Abbildung in dieser Leseprobe nicht enthalten

Figure 3.13. Ranking results and criteria weights for scenario C and D

In this chapter, an evaluation and comparison of enterprises has been achieved. It is clearly shown that, using PROMETHEE method, enterprises can be evaluated taking into account their competences, i.e. what enterprise possess in the terms of technology, references, information system, etc. Hence, economic and sociological criteria can also be added into analysis. A special scenario portfolio was created for different complexity of product and/or production process.

On the case study with real enterprises, it is shown that different scenarios will produce different enterprise as the best one. So it is very important for production network manager to carefully choose criteria weights and form proper scenarios. This could be done by interviewing experts. The evaluation and comparison of enterprises was pre-requisition to evaluate, compare and rank virtual enterprise.

But, in this case the enterprises have been compared, and to compare and rank virtual enterprises becomes more complex. If virtual enterprises are a priori selected, than they could be compared and ranked in a similar way, with some additional criteria. However, if virtual enterprises need to be created there could be millions and millions of possible combinations, where every combination represents one virtual enterprise. Therefore, some optimization algorithm must be used in order to compare and rank only the set of the best virtual enterprises. This approach is described in the following sections.

3.2 OPTIMIZATION OF THE VIRTUAL ENTERPRISE CREATION

The most important step in virtual enterprise lifecycle is the process of creation of virtual enterprise. Because, during that process an optimal combination of enterprises from production network must be selected. The selection of optimal partners is called Partner Selection Problem (PSP) and it can be solved using different metaheuristic algorithms and/or multi-criteria decision-making methods.

Here are the some approaches that can be found in the literature: AHP in combination with Ant Colony Optimization (Fischer, Jähn and Teich 2004; Mladineo, Veza and Corkalo 2011); PROMETHEE method (Lanza and Book 2011; Mladineo and Veza 2013); Particle swarm algorithm (Gao et al. 2006); TOPSIS method (Yu et Wong 2011); DEA in combination with Genetic algorithm (Chuanga et al. 2009), and other different heuristic (Wu, Mao and Qian 1999) or metaheuristic approaches (Zhao, Hong and Yu 2006; Wang, Xu, Zhan 2009; Zhang et al. 2012). Generally, some of the approaches are comparing a priori created set of virtual enterprises, and some of the approaches are using algorithms to construct virtual enterprises from set of the enterprises that are evaluated and compared using some fitness function. A different sets of criteria are also used, but the dominating criteria are: price, time, quality, and transportation cost. These criteria are similar to those that are commonly used in supplier evaluation (Ho, Xu and Dey 2010; Babic and Peric 2014): quality, delivery time, price (production cost plus all other costs, including transportation cost). Criteria set also defines complexity of the problem. Many approaches claim that are successfully solving PSP, but they are actually simplifying the problem by avoiding criterion of geographical dispersion of the partners, i.e. transportation cost.

In reality this problem is even more complex, because there are many different parameters that influence the selection process, like negotiations and similar issues. However, the simple example of a production network with 3 enterprises will be used to analyze the problem and its solution. The parameters that affect rating of the virtual enterprise are: distances between enterprises (Table 3.3) and enterprises rating (Table 3.4).

Table 3.3. Mutual distances between enterprises - Lij

Abbildung in dieser Leseprobe nicht enthalten

Table 3.4. Enterprises ratings - Ci

Abbildung in dieser Leseprobe nicht enthalten

The parameters that represent the constraints are: the capacities of enterprises (Table 3.5) and the time needed to fulfill work (lead time). The parameter "lead time" will be ignored in this example.

Table 3.5. Enterprises capacities - Ki

Abbildung in dieser Leseprobe nicht enthalten

The following problem should be solved (Figure 3.14), i.e. select an enterprise for each technological process. A selected enterprise will carry out selected part of the production process.

Abbildung in dieser Leseprobe nicht enthalten

Figure 3.14. Variants of production process

From a given network graph it can be seen that each enterprise is competent to perform any of the 3 required processes, so each process has 3 possible variants.

Each solution represents a possible variant of the production process. In this example, because of limited capacities, there are 14 varieties (solutions) of the same production process. To evaluate solutions and to determine which is the most optimal, it is necessary to establish the fitness function ff. Two parameters affect the value of the fitness function: the sum of distances between the enterprises involved in the production process and the sum of ratings of enterprises involved in the production process. However, in order to determine the fitness of certain enterprise, these parameters should be weighted, and then by summing the weighted values calculate the value that represents fitness of certain enterprise. It is therefore necessary to introduce a weight for the distance TL and weight for rating TC. In addition, since the fitness value should be maximal and total distance should be minimal, it is necessary to use a reciprocal value of the distance.

Taking into account the above, the optimal variant of the production process is presented in figure (Figure 3.15).

Abbildung in dieser Leseprobe nicht enthalten

Figure 3.15. Optimal variant of production process

This solution can be written in the form of a vector:

[E1, E1, E3] à R = [1, 1, 3]

If there were no limitation in capacity, the solution would be a vector [E1, E1, E1]. However, due to limited capacity the entire production process cannot be performed in the same enterprise.

The expression for calculating the value of fitness function would look like this:

ff = v1 + v2 + v3 + v4

ff = (TC × CR1 + TL × 1/LR1, 0) + (TC × CR2 + TL × 1/LR2, R1) + (TC × CR3 + TL × 1/LR3, R2) + 0

However, this expression has certain problems, so it is not convenient to use. The first problem is the possible division by zero, and the second problem is indifference on staying in the same enterprise. Namely, since remain in the same enterprise means to have a reciprocal value of the distance (1 / L) equal to zero; it is possible that solutions with high fitness have very high total distance. But it is illogical, because it has already been mentioned that without restrictions in capacity the best solution is to perform everything in the enterprise E1, due to its highest ratings. To avoid mentioned problems, it is necessary to set the parameters of the problem a bit different. First of all the distances (in the matrix of distances between enterprises) that are 0 need to be converted to 0.1; to be able to calculate the reciprocal value. After that, the reciprocal value should be placed in a relative ratio with respect to the maximum value of the matrix. The values of ratings are also placed in a relative ratio with respect to the maximum, while the values of capacities remain the same. Thus, the predefined matrices now look like the following tables (Table 3.6 and Table 3.7).

Table 3.6. Relative and reciprocal mutual distances between enterprises - Lij

Abbildung in dieser Leseprobe nicht enthalten

Table 3.7. Relative enterprises ratings - Ci

Abbildung in dieser Leseprobe nicht enthalten

Now, it is obvious that the algorithm will prefer to stay in the same enterprise, which means less total distance, and that lowers transportation costs. Putting values in a relative ratio makes possible to compare two different parameters.

For the solution described above, the expression for calculating the value of fitness function qualities should now look like this:

ff = v1 + v2 + v3 + v4

ff = (TC × CR1 + TL × LR1, 0) + (TC × CR2 + TL × LR2, R1) + (TC × CR3 + TL × LR3, R2) + 0

ff = (0.5 × 1 + 0.5 × 0) + (0.5 × 1 + 0.5 × 1) + (0.5 × 0.8 + 0.5 × 0.009) + 0

ff = 2.409

In calculation the both weighting factors (TC and TL) are given a value of 0.5, i.e. they are equally important. But these weights can be defined by the decision-maker. It is called the a priori approach to multi-objective optimization, which means that decision-makers preferences are known.

Optimization of the VE creation by using Ant Colony Optimization

Since the problem may contain parallel processes (or branches) within the network, it cannot be solved easily by any algorithm for solving network problems. One of the metaheuristic algorithms that are easy to adapt to this kind of problem is Ant Colony Optimization – ACO (Dorigo et al. 1996).

ACO algorithm is a bit modified in a way it can solve the problem parallel processes or problem of branching (Fischer et al. 2004). It is done by solving each branch separately. If we ignore the capacity constraints and “lead time”, such an approach is possible, since the branching occurs at the beginning, because it represent intermediate products (parts) which will be assembled in a final product. Therefore, each branch can be solved before solving the complete network, and that will not change the overall solution.

The algorithm will be applied and the results analyzed on a network consisting of 12 enterprises and product consisting of 7 intermediate products (parts). The product is a bench vise, and it consists of parts: vise body, lower plate, moving jaw, jaw, spindle, handle and rail. Each part has its own production process, i.e. the sequence of technological processes needed to be done before the final assembly (Figure 3.16). It is necessary to form a main production process, which contains production processes of each intermediate product, where each intermediate product is one branch of the network (Figure 3.17).

Abbildung in dieser Leseprobe nicht enthalten

Figure 3.16. The product whose production process will be solved

Abbildung in dieser Leseprobe nicht enthalten

Figure 3.17. Main production process

To put it simply, to solve this production process for each technological process the enterprise that will fulfill it need to be selected. The fictional production networks of Split-Dalmatia County with 12 fictitious enterprises will be used. Each enterprise can perform various technological processes and they actually determine their competence. Fictional production network is shown in Figure 3.18.

Abbildung in dieser Leseprobe nicht enthalten

Figure 3.18. Fictional production network of Split-Dalmatia County

All enterprises of network are mutually connected, and that represents a potential cooperation of each enterprise with each. All relevant information about the enterprises is represented in the following tables (Table 3.8 and Table 3.9). This partner selection problem (Mladineo, Veza and Corkalo 2011) consists of 35 activities and 12 partners (see Appendix for more information).

Table 3.8. Enterprises of a network with information about mutual distances, ratings and capacities

Abbildung in dieser Leseprobe nicht enthalten

Table 3.9. Enterprises of a network with information about their competences

Abbildung in dieser Leseprobe nicht enthalten

The tables show that different enterprises have different competences, some very few of them, while some can do all the required technical processes. The assumption is that the smaller enterprises with narrow field of work can be more specialized, technologically advanced and superior. Therefore, these enterprises will probably have a higher rating than the enterprises with wide field of work. To achieve the better variety, range of ratings is equal to the number of enterprises, ranging from 1 to 12, and rating actually represents the rank (but inverted). For each intermediate product a network with alternatives for each process is developed (Figure 3.19). The enterprises are represented by their "ID".

Abbildung in dieser Leseprobe nicht enthalten

Figure 3.19. Production process of intermediate product with alternatives for each technological process

The figure shows the production process of vise jaws. Same network needs to be created for each intermediate product, but also for the entire product. The algorithm will choose optimal enterprise for each technical process, trying to optimize the entire production process. The problem will be solved with the help of the software package MATLAB. Flowchart of the ACO algorithm is presented on the following figure (Figure 3.20).

Abbildung in dieser Leseprobe nicht enthalten

Figure 3.20. Flowchart of the modified ACO algorithm

As mentioned before, the problem is solved by using a modified ACO algorithm written in MATLAB. Achieved results were satisfying, and now they will be analyzed. It is important to note enterprises capacity constraints were ignored, so each enterprise has infinite capacity. The parameters that affect selection were the rating and the distance. The following figure (Figure 3.21) shows one of the optimal solutions, which has weighting factors set to TC = 0.6 and TL = 0.4.

Abbildung in dieser Leseprobe nicht enthalten

Figure 3.21. One of the optimal combinations of enterprises for production of bench vises

Solving this example with different weighting factors for parameters revealed 7 different optimal solutions, or 7 different optimal combinations of the production process. The person responsible for the process will decide for one of solutions taking into account the quality requirements (which are represented by rating) and the permissible amount of transportation costs (which are represented by total distance). Three different optimal solutions are presented on the following figure (Figure 3.22).

Abbildung in dieser Leseprobe nicht enthalten

Figure 3.22. Different optimal solutions for different weighting factors of the parameters

Optima can be presented as curve that represents the Pareto front (Figure 3.23). Below the curve are solutions "worse" than optimal, and above the curve are solutions "better" than optimal. But, with the currently available resources, these "better" are also unrealizable solutions.

This approach proved that it is possible to optimize the selection of partners (cooperators) in a regional production network, with the help of the modified ACO algorithm. He also proved that the assignment of various technological processes (of the same production process) to the various enterprises achieved higher average rating, which should ultimately lead to greater quality of processes and products.

Abbildung in dieser Leseprobe nicht enthalten

Figure 3.23 Optimal solutions and the approximated Pareto front

However, presented problem is simplified, because only two objectives are used: transportation cost and rating (quality) of the enterprise. It is the reason why the optimal solutions form Pareto form, because two objectives mean two-dimensional plane. But, three objectives would result with three dimensional space, i.e. n-objectives would result with n-dimensional space.

Therefore, better mathematical model of the Partner Selection Problem is needed, which could support more than two criteria and use the original criteria evaluations. Furthermore, there is a need for better optimization algorithm than one presented on this example.

3.3 MATHEMATICAL MODEL OF THE PARTNER SELECTION PROBLEM

There is no clear mathematical definition of Partner Selection Problem (PSP), like, for instance, there are definitions of Job-Shop Scheduling Problem, Assignment Problem, etc. Therefore, beside an algorithm for solving Partner Selection Problems, a mathematical definition or mathematical model of the Partner Selection Problem should be given. In this chapter, the mathematical model of PSP with four objectives (criteria) and some guidelines for labelling different PSP instances have been proposed.

Partner Selection Problem is a problem of assigning set of activities to set of partners, where predefined subset of partners is assigned to each activity. It means that PSP consists of a set of activities A where each activity ai element of A has its set Si which is a subset of set of partners P. The objective is to assign one partner pj from each subset Si to activity ai (Figure 3.24).

Abbildung in dieser Leseprobe nicht enthalten

Figure 3.24. Partner Selection Problem (Mladineo et al. 2017)

However, due to practical reason, it is better to say that pj represents partner’s bid instead of partner. Because, a partner (i.e. enterprise) could be interested in more than one activity, and it can offer different price and lead time for different activities, i.e. different bids. So, set P is actually set of partners’ bids, but to avoid changing name of this problem into Bid Selection Problem, it is not so important to insist on this fact.

Abbildung in dieser Leseprobe nicht enthalten

Figure 3.25. Example of Partner Selection Problem with parallel activities (Mladineo et al. 2017)

Furthermore, Fischer, Jähn and Teich (2004) identified problem of parallel, converging production (or parallel activities) in the Partner Selection Problem. It makes PSP difficult to solve using classical ACO algorithm. Because, in this kind of problems the solution isn’t single path on the graph, but path that is branching across the graph (Figure 3.25). Therefore, this fact must also be taken into account in the definition of the mathematical model of PSP.

Usually, PSP has more than one objective. In this model, four objectives, which are most often founded in the literature, are used: price (activity's cost), time (activity's lead time), quality (of a partner) and transportation cost (between partners). In Figure 3.26, schematic and mathematic representation of simple PSP instance with one objective (price) is presented and objective function and constraints are defined.

Abbildung in dieser Leseprobe nicht enthalten

Figure 3.26. Schematic and mathematic representation of simple PSP instance

Generally, for PSP instances with up to four objectives, mentioned above, following problem matrices and objective functions are defined:

Minimization of total cost:

Abbildung in dieser Leseprobe nicht enthalten

Maximization of total quality (or, optionally, average quality):

Abbildung in dieser Leseprobe nicht enthalten

Minimization of total transportation distance (or transportation cost):

Abbildung in dieser Leseprobe nicht enthalten

Minimization of total lead time:

Abbildung in dieser Leseprobe nicht enthalten

where cij, qij, sij, and tij represent criterion evaluation for each partner, n represents total number of partners (i.e. partners’ bids), xij represents selection decision (is one partner selected, or not), and K represents correction factor which eliminates sums of lead times of activities that are executed parallel. Factor K is difficult to calculate, but it can be obtained algorithmically like in project management techniques (Critical path method or Precedence method), or it can equal zero if there are no parallel activities. Additionally, in reality if a quality is defined as some kind of score or grade then equation (3.11), which summarizes quality score or grade, doesn’t have so much sense. Instead of equation (3.11), total quality score or grade, an average quality score or grade can be used:

Abbildung in dieser Leseprobe nicht enthalten

where total quality is divided with number of partners to get average value of quality score or grade.

Furthermore, following constraints are needed in order for correct mathematical definition of the problem:

Only one partner (i.e. partner’s bid) can be assigned to each activity:

Abbildung in dieser Leseprobe nicht enthalten

Each xij can equal 0 or 1:

Abbildung in dieser Leseprobe nicht enthalten

Each xij ∙ cij must equal 0 or greater than 0:

Abbildung in dieser Leseprobe nicht enthalten

Each xij ∙ qij must equal 0 or greater than 0:

Abbildung in dieser Leseprobe nicht enthalten

Each xij ∙ sij must equal 0 or greater than 0:

Abbildung in dieser Leseprobe nicht enthalten

Each xij ∙ tij must equal 0 or greater than 0:

Abbildung in dieser Leseprobe nicht enthalten

Since PSP is multi-objective (or multi-criteria) problem a Multi-Objective Optimization (MOO) must be used. When considering MOO, there are three approaches that can be identified (Talbi 2009):

· A priori approach – decision-maker provides preferences before the optimization process.

· A posteriori approach – the optimization process determines a set of Pareto solutions, and then the decision-maker chooses a solution from the set of solutions provided by the algorithm.

· Interactive approach – there is progressive interaction between the decision-maker and the solver, i.e. knowledge gained during the optimization process helps the decision-maker to define the preferences.

A posteriori and interactive approaches are usually applied on problems with two or three criteria, no more than that. However, a priori approach could be applied on problems with four, five, or more than ten criteria in theory, because a priori approach is using predefined decision-maker's preferences. In the practice, it means it is using some multi-criteria decision-making method to scalarize objectives, and that actually transforms multi-criteria problem into single-criterion problem. So, optimization process is based on some usual (single-objective) metaheuristic optimization algorithm in combination with MCDM.

In this research, a combination of metaheuristic Ant Colony Optimization (ACO) (Dorigo, Maniezzo and Colorni 1996) and MCDM method called PROMETHEE (Preference Ranking Organization METHod for Enrichment of Evaluations) (Brans, Mareschal and Vincke 1984) has been turned into the new algorithm: HUMANT (HUManoid ANT) algorithm (Mladineo, Veza and Gjeldum 2015). Originally, it was designed by Mladineo in 2014 to solve Partner Selection Problem. More details on the HUMANT algorithm will be given in the next chapter.

Labeling Partner Selection Problem instances

Since no one defined how to label PSP instances, following simple labelling can be used: name of the instance - optimization objectives (abbreviated: T, C, S, and/or Q) - total number of activities - total number of partners. This approach results with the following label for the instance presented in Figure 3.26: PSP-C-2-4. Other labeling examples are given in Table 3.10.

If the PSP instance is from the literature, as 'name of the instance' an abbreviated name of the first author can be used, as it is used for the TSP (Traveling Salesman Problem) instances. This kind of labelling also denotes the complexity of the problem, because problems with high number of activities and partners are the complex ones.

For instance, if there are 30 activities and 5 same partners for each activity, with objective to minimize the price, that problem would be labelled as PSP-C-30-5 and it has more than 10[20] potential solutions. So, this PSP instance is a complex one and with increment of number of criteria, it is becoming more and more complex. However, if criteria of geographic dispersion of partners, represented as transportation distance or cost, is not used, PSP is not NP-hard problem. On the other hand, if criteria of geographic dispersion of partners is used, PSP becomes a NP-hard problem, and it can be solved only using some metaheuristic algorithm.

In this research, PSP instances from the literature used to experimentally test HUMANT algorithm are labelled as: wu-CS-7-19 (Wu, Mao and Qian 1999), ve-CSQ-4-10 (Veza and Mladineo 2013), and mla-SQ-35-12 (Mladineo, Veza and Corkalo 2011).

Table 3.10. Example of labeling of Partner Selection Problem instances

Abbildung in dieser Leseprobe nicht enthalten

4 SOLVING PARTNER SELECTION PROBLEM BY USING THE HUMANT ALGORITHM

I believe that in every God’s creature has more than what is being perceived, even if it is a little ant.

(S. Teresa de Jesús, mystic)

Summary

In this chapter, new metaheuristic algorithm for Single-Objective and Multi-Objective Optimization is presented: HUMANT (HUManoid ANT) algorithm. It was designed to be Multi-Objective Optimization algorithm with a priori approach as a combination of metaheuristic algorithm Ant Colony Optimization and multi-criteria decision-making method PROMETHEE.

The proof of concept of HUMANT algorithm is presented on solving different instances of Traveling Salesman Problem. Furthermore, using multi-criteria approach in single-objective optimization HUMANT algorithm can become more intelligent and more efficient as demonstrated in solving single-objective Shortest Path Problem.

At the end, it has been presented how HUMANT algorithm is able to solve multi-criteria Partner Selection Problem (PSP) instances. On three simple PSP instances (PSP-TCSQ-3-9, wu-CS-7-19 and ve-CSQ-4-10) it has been proven that HUMANT algorithm can successfully find optimal solution of a PSP. For one complex PSP instance (mla-SQ-35-12), HUMANT algorithm has found better optimal solution than the, so far, known optimum.

4.1 HUMANT ALGORITHM

Metaheuristic Optimization

Metaheuristic optimization deals with optimization problems in different areas and applications, from engineering design to economics and everyday life (telecommunication routing, trip planning, etc.) (Yang 2011). Most real-world optimizations are very complex and are accompanied by numerous constraints where such optimizations are solved using a highly efficient metaheuristic algorithm. Different objectives are often conflicting, so finding an optimal solution or even near-optimal solutions can be a hard task (Talbi 2009). There are many applications of metaheuristic algorithms to different multi-objective optimization problems, but no known universal multi-objective metaheuristic optimization algorithm is able to solve most of the multi-objective optimization problems (Talbi 2009).

Generally, multi-objective optimization methods are classified (Figure 4.1) based on three different approaches (Talbi 2009):

- A priori approach – decision-maker provides preferences before the optimization process.
- A posteriori approach – the optimization process determines a set of Pareto solutions, and then the decision-maker chooses a solution from the set of solutions provided by the algorithm.
- Interactive approach – there is progressive interaction between the decision-maker and the solver, i.e. knowledge gained during the optimization process helps the decision-maker to define the preferences.

Abbildung in dieser Leseprobe nicht enthalten

Figure 4.1. Classification of multi-objective optimization algorithms (Talbi 2009)

There are evolutionary a posteriori approach multi-objective optimization algorithms with excellent results in solving problems with two or three criteria, such as: Non-dominated Sorting Genetic Algorithm II (NSGA-II) (Deb et al. 2002), S Metric Selection Evolutionary Multiobjective Optimisation Algorithm (SMS-EMOA) (Beume et al. 2007), Strength Pareto Evolutionary Algorithm 2 (SPEA2) (Zitzler et al. 2001) and Scatter Tabu Search Procedure for Non-Linear Multiobjective Optimization (SSPMO) (Molina et al. 2007).

There are also some successful interactive approaches (Deb et al. 2007; Luque et al. 2007; Phelps et al. 2003) to multi-objective optimization.

However, in this research the focus is on an a priori approach to multi-objective optimization, meaning that a particular single-objective metaheuristic algorithm is combined with a MCDM method, such as Multi Objective Ant Colony Optimization (MOACO) (Eppe 2009) or Progressive Multi-Objective Optimization (PMOO) (Sörensen et al. 2014).

Generally, MCDM methods use some form of a priori knowledge concerning decision-maker preferences, such as criteria weights. Using these a priori decision-maker preferences, alternatives are compared and ranked i.e. they solve complex multi-criteria problems subject to a limited number of alternatives. However, if this limited number of alternatives becomes unlimited or very large (10[20] alternatives or more), then practically applying these methods is impossible, as the computational time becomes too long. Consequently, these methods cannot be applied to complex multi-objective optimization problems because number of possible solutions (alternatives) is too large, and sometimes an infinite number of possible solutions exist. Thus, if a metaheuristic algorithm is combined with a Multi-Criteria Decision-Making method, then only near-optimal solutions are submitted for multi-criteria evaluation i.e. compared and ranked using these a priori decision-maker preferences.

For several years, the idea has existed of combining metaheuristic algorithms, such as the Ant Colony Optimization (ACO), Genetic Algorithm (GA), Particle Swarm Optimization (PSO) or Simulated Annealing (SA) with the Multi-Criteria Decision-Making (MCDM) method (like AHP, ELECTRE, MAUT, PROMETHEE, or TOPSIS) to scalarize multi-objective problems into single-objective problems (Talbi 2009). So, when considering the a priori approach, any of the popular metaheuristic algorithms can be used for multi-objective optimization – Ant Colony Optimization (ACO) (Dorigo et al. 2006), Cuckoo Search (CS) (Yang et al. 2010), Firefly Algorithm (FA) (Yang et al. 2009), Genetic Algorithm (GA) (Holland 1975), Particle Swarm Optimization (PSO) (Kennedy et al. 1995) and Simulated Annealing (SA) (Kirkpatrick et al. 1983) – but has to be combined with a MCDM method in the appropriate manner.

Multi-Criteria (Multi-Attribute or Multi-Objective) Decision-Making (MCDM) consists of selection of the best alternative, comparison and ranking of alternatives, or comparison of alternatives with some referent points (sorting of alternatives). Generally, MCDM methods can be divided into following groups based on their characteristics: based on utility functions – MAUT (Farquhar 1977), outranking methods – AHP (Saaty 1980), ELECTRE (Roy 1968), PROMETHEE (Brans et al. 1984), TOPSIS (Yu et al. 2011), and interactive methods – VIMDA (Olson 1996). The main disadvantage of all these MCDM methods is that their scope is Multi-Criteria problems with finite number of alternatives, usually 10-30 alternatives, and very rare more than 100. So, in combination with metaheuristic algorithm, a limited number of near-optimal solutions found by metaheuristic algorithm should be submitted to some MCDM method. According to (Coello el a 2007), few papers describe methods with a priori approach in which the user's preferences are integrated into a multi-objective metaheuristic algorithm.

Multi-Objective Ant Colony Optimization

There is no single optimal solution to the multi-objective problem, but instead a set of solutions defined as the Pareto optimal solutions (Talbi 2009). A solution is Pareto optimal if a given objective (criterion) cannot be improved without degrading other objectives (criteria). This set of solutions represents a compromise between different conflicting objectives (criteria). The main goal is to obtain the Pareto optimal set and, subsequently, the Pareto frontier. However, there can theoretically be a single solution to some combinatorial optimization problems. There is no optimal solution based on each criterion, but an overall optimal solution is based on decision-maker preferences, obtained using a MCDM method. Such approach is called the MCDM a priori approach to Multi-Objective Optimization.

In this research, a priori approach to multi-objective optimization combines the PROMETHEE method and Ant Colony Optimization (ACO) in a metaheuristic manner into a specialized single-objective and multi-objective metaheuristic algorithm to solve complex optimization problems.

As already mentioned, PROMETHEE method is well accepted by decision-makers because it is comprehensive and has the ability to present results using simple ranking (Brans et al. 1984). An input for PROMETHEE method is a matrix consisting of set of potential alternatives (actions) A, where each a element of A has its f(a) which represents evaluation of one criteria. Each evaluation fj(ai) must be a real number. Method PROMETHEE I ranks actions by a partial pre-order with the following dominance flows: the leaving flow (equation (3.6)) and the entering flow (equation (3.7)). These two flows give the partial relation, and then a net outranking flow is obtained from PROMETHEE II method which ranks the actions by total pre-order calculating net flow (equation (3.8)). In the sense of priority assessment net outranking flow represents the synthetic parameter based on defined criteria and priorities among criteria. Usually, criteria are weighted using criteria weights wj and usual pondering technique.

On the other hand, Ant Colony Optimization (ACO) is a population-based metaheuristic algorithm (Dorigo et al. 1996)0. It is based on set of virtual agents called artificial ants. They search for solutions to given optimization problem. However, optimization problem must be presented as mathematical graph, and ants build solutions by moving on the graph and leaving pheromone trail. The solution construction process is stochastic and it is affected by pheromone update system. Hence, main algorithm procedures are: construction of solutions by ants and pheromone update.

Ants construct solution by moving on the graph from node to node until all nodes are visited or some condition has been met, i.e. until solution is constructed. Every single node is chosen using some probabilistic rule, like one from the Ant System (Dorigo et al. 1996):

Abbildung in dieser Leseprobe nicht enthalten

where pij represents probability that ant k moves from node i to node j, and where η is reciprocal weight of edge connecting two nodes, τ is level of pheromone on edge connecting two nodes, α and β are control parameters of algorithm.

Pheromone level is updated after each iteration n, using rule (Dorigo et al. 1996):

Abbildung in dieser Leseprobe nicht enthalten

where τij(t) represents level of pheromone on edge connecting nodes i and j in time t, and ρ represents pheromone evaporation rate. Furthermore, Δτij value depends on quality of solution (Dorigo et al. 1996):

Abbildung in dieser Leseprobe nicht enthalten

where Lk represents solution found by ant k (e.g. total length) and Q is constant parameter.

Generally, at the beginning of search pheromone level is on each edge set to some minimal value, it is increased over time on edges that are connecting good solutions and it is evaporating on unvisited edges (bad solutions). Eventually, there is a convergence toward optimal or near-optimal solution (Dorigo et al. 1996).

Convergence of ACO was proven in its modification called MAX-MIN ant system (MMAS) (Stützle et al. 2000). It is different from original ACO, because it is using limitation of pheromone level on edges using τmin and τmax parameters, and initially pheromone level on each edge is set to maximal value (τmax). So, MMAS has convergence based on evaporation of bad solutions, instead of encouraging good solutions. It has achieved better results than all others ACO algorithms on many optimization problems (Stützle et al. 2000).

Since first application of ACO in area of multi-objective optimization (Iredi et al. 1993) several approaches to Multi-Objective Ant Colony Optimization (MOACO) have been developed (Eppe et al. 2014):

- Evaluation of objectives – the decision-makers preferences are a priori known (Eppe 2009), so solutions can be scalarized, i.e. problem can be converted from multi-objective to single-objective (Mladineo et al. 2011). This is the same approach as generic a priori approach to Multi-Objective optimization.
- One or several pheromone matrices – each objective is presented by its own pheromone matrix to keep as much as possible information on each criterion. However, at the end scalarization must be made to compare solutions.
- One or several colonies – usually means each colony is constructing solution only for one objective. At the end, Pareto-set of optimal solutions is constructed. This is the same approach as generic a posteriori approach to Multi-Objective optimization.
- Different pheromone update approaches – some specific rules for pheromone update based on set of ants or set of colonies.

As mentioned, in this research a priori approach to Multi-Objective Optimization is used, so the decision-makers preferences will be known and solutions will be scalarized using the MCDM method.

Concept and methodology of the HUMANT algorithm

This research presents interesting approach to single-objective and multi-objective optimization based on idea to combine Ant Colony Optimization and multi-criteria PROMETHEE method for solving multi-objective optimization problems. This combination allows artificial ants to use multi-criteria decision-making which is actually a human attribute.

Originally, algorithm was designed to solve multi-criteria Partner Selection Problem (PSP) and it was called HUMANT (HUManoid ANT) algorithm (Mladineo 2014). Idea and concept of HUMANT algorithm is presented in Figure 4.2.

Abbildung in dieser Leseprobe nicht enthalten

Figure 4.2. Idea and concept of the HUMANT algorithm (Mladineo et al. 2015)

To realize presented concept ACO equation (4.1) for evaluation of nodes and solutions, and equation (4.2) for pheromone update have to be changed. PROMETHEE method must be implemented in them. In procedure of constructing solution probabilistic rule to evaluate nodes is changed from (4.1) into:

Abbildung in dieser Leseprobe nicht enthalten

where Φ’ij represents modified PROMETHEE II score for edge connecting nodes i and j. It is calculated using following equation:

Abbildung in dieser Leseprobe nicht enthalten

where Xij represents weight of edge connecting nodes i and j. Φ’ij is modified PROMETHEE II score, modified to result with interval [0, 1] instead of interval [-1, 1].

The equation (4.5) can also be used to compare two different solutions of the problem in order to find the optimal one. However, pheromone update can be a problem, since PROMETHEE method is not based on utility functions. It means that PROMETHEE scores of alternatives from two different analysis cannot be mutually compared. If the PROMETHEE score is used as-is, an irregular pheromone update will happen, i.e. it may happen that iteration-best solution gets more pheromone than global-best solution founded in the previous iteration. To avoid this irregular pheromone update, in HUMANT algorithm each solution is compared with ideal solution of the problem, which is additional parameter of the algorithm. So, the following equation is used for the pheromone update:

Abbildung in dieser Leseprobe nicht enthalten

where x represents solution and sid represents known optimal solution (for already solved optimization problems). If optimal solution is unknown then ideal solution is used. For instance ideal solution in Shortest Path Problem (SPP) is Euclidean distance between origin and destination. In Traveling Salesman Problem (TSP) ideal solution is sum of each row’s minimal distance (or cost). Usually, it is not achievable solution, i.e. it is solution better than optimal solution.

Furthermore, HUMANT algorithm was designed to use MAX-MIN approach with pheromone trail limits set to τmin = 0 and τmax = 1. If constructed solution x is same as optimal (not ideal) solution sid then equation (4.6) will result with 1, i.e. edges connecting optimal solution will receive maximal pheromone level. So convergence toward that solution is expected. If optimal solution is unknown and ideal solution is used, convergence will be a bit slower. Parameters of HUMANT algorithm and its comparison to MAX-MIN ant system parameters are shown in Table 4.1.

Table 4.1. Comparison of HUMANT algorithm parameters and MAX-MIN ant system parameters

Abbildung in dieser Leseprobe nicht enthalten

Parameter β from equation (4.4) is missing in Table 4.1, because rule of HUMANT algorithm is:

Abbildung in dieser Leseprobe nicht enthalten

where n represents total number of nodes that are evaluated. It is a solution to a problem of non-dominating solution of PROMETHEE II methods, which occurs when there is a large number of alternatives (more then 25-30 alternatives). Because in ACO it is a mandatory that one or few excellent alternatives are dominating in comparison with others. However, control of importance of weight (cost) of edge is implemented into HUMANT algorithm using parameter γ. To simplify PROMETHEE II calculations only linear preference function is used, and parameter γ is controlling parameters of linear preference function. Preference threshold p for linear preference function is calculated using following:

Abbildung in dieser Leseprobe nicht enthalten

where pa represents automatically calculated linear preference threshold. And indifference threshold q is calculated using following:

Abbildung in dieser Leseprobe nicht enthalten

where qa represents automatically calculated linear indifference threshold. Automatic calculation of parameters pa and qa means that qa is standard deviation of set of mutual differences between actions, and pa is qa plus average value of mutual differences between actions.

To conclude, HUMANT algorithm is MOACO algorithm with MAX-MIN strategy, global-best ant pheromone update strategy, and scalarization approach using simplified PROMETHEE II methods (i.e. using only linear preference function and automatic calculation of its parameters). It was designed to have small number of parameters and fast calculation times. It can be applied to single-criteria and multi-criteria optimization problems.

It is important to mention similar approach of combining ACO with PROMETHEE method by Eppe in 2009. However, they only presented theoretical way of doing it with no practical solution, i.e. algorithm. Their pheromone update strategy is different and experimentally not tested, and they did not propose solution to a problem of non-dominating solution of PROMETHEE II methods nor to a problem of problematic interval of PROMETHEE II method results, i.e. [-1, 1] interval.

HUMANT algorithm and single-objective optimization

Traditionally, the Traveling Salesman Problem (TSP) was chosen to experimentally test HUMANT algorithm on single-objective optimization problems. Problem instances att48, eil51, kroA100 from TSPLIB were used and HUMANT algorithm and its parameters were tested. The best combination of algorithm parameters was obtained: α = 1, γ = 1, ρ = 0.4, τmin = 0 and τmax = 1.

On each instance algorithm was run 30 times with 500 iterations. Algorithm has been written in MATLAB and tests were made on 2.80 GHz CPU. Average run-time of 500 iterations for 48-cities problem (att48) was 703 seconds. Parallelization of code was not used. Average result and average deviation from optimum were compared with other ACO algorithms (Table 4.2). Results of other ACO algorithm are taken from (Stützle et al. 2000).

Table 4.2. Comparison of HUMANT algorithm with other ACO algorithms on solving Traveling Salesman Problem.

Abbildung in dieser Leseprobe nicht enthalten

Table 4.2 is proof of a concept of HUMANT algorithm, because results are satisfying and it is interesting that on 100-cities problem HUMANT algorithm is even better than Ant Colony System and original Ant System. HUMANT algorithm had the worst performance on 51-cities problem, because problem instance eil51 is TSP problem with many local optima and HUMANT algorithm has very strong convergence.

HUMANT algorithm and multi-objective approach to single-objective optimization

Shortest Path Problem (SPP) was chosen to experimentally test ability of HUMANT algorithm to use multi-objective approach to single-objective optimization problem. In SPP main aim is to find shortest path between two nodes on some graph that represents road network or similar. It is not NP-hard problem and there are known heuristic algorithm that can solve it, like Dijkstra’s algorithm or modified Dijkstra’s algorithm, which also uses some constraints, called A-star algorithm.

However, if no constraints are used, algorithm will explore all edges or most of the edges of graph. If an algorithm could find shortest path using no constraints and exploring only the nodes closest to origin and destination nodes, such an algorithm would be intelligent and efficient. But, such an algorithm could be designed only if multi-objective approach is used. In solving SPP, additional criteria (objectives) in search for solution can be: deviation from Euclidean distance between origin and destination. It is a bi-criteria approach to SPP (Figure 4.3).

Abbildung in dieser Leseprobe nicht enthalten

Figure 4.3. Multi-objective (bi-criteria) approach to Shortest Path Problem (Mladineo et al. 2015)

Using multi-objective approach to solve single-objective SPP HUMANT algorithm appeared to be much more intelligent and efficient than usual Dijsktra’s algorithm and usual ACO (Figure 4.4).

Abbildung in dieser Leseprobe nicht enthalten

Figure 4.4. Comparison of performance of Dijkstra’s, ACO and HUMANT algorithm on solving Shortest Path Problem (Mladineo et al. 2015)

4.2 HUMANT ALGORITHM AND PARTNER SELECTION PROBLEM

HUMANT algorithm was originally designed to solve Partner Selection Problem. Since most of the Partner Selection Problems have parallel activities, it needs to be little bit adapted for solving problem with parallel activities. Namely, in this kind of problems the solution isn’t single path on the graph, but path that is branching across the graph. The approach to solve this branching problem, proposed by Fischer, Jähn and Teich (2004), is to divide single ant into multiple sub-ants when it arrives on the branching point. When branches unite again, multiple sub-ants will become single ant that carries information about solution of each branch. At the end, global solution is evaluated taking into account the path of this single ant, but also the paths of each sub-ant for each branch. This issue makes optimization algorithm more complex, but it also allows successful solving of Partner Selection Problem with parallel, converging production (i.e. with parallel activities or processes).

To enable solving of PSP with parallel activities, HUMANT algorithm needs to have procedure for construction of solution which can call itself (Figure 4.5). In that way, it becomes possible to move up and down the assembly levels. However, it is important to point that number of assembly levels is lower or equal than number of levels in Bill of Material (BOM). Each level of BOM that consists of at least one assembly defines one assembly level. But BOMs don't need to have assembly on each level, since part – assembly is not only connection between levels. It can also be connection raw material – part, and this kind of connection is not branching point for HUMANT algorithm, as it is presented in Figure 5.

Abbildung in dieser Leseprobe nicht enthalten

Figure 4.5. (a) Comparison of levels of Bill of Material and assembly levels of PSP; (b) Pseudocode of HUMANT algorithm adapted for Partner Selection Problem with parallel activities (Mladineo et al. 2018)

Second problem of HUMANT algorithm’s adaptation are different aggregation types of criteria. There are four basic aggregation types of criteria: Sum, Mean, Distance and Time. They are characterized with following features:

- Sum aggregation type is used for criteria like cost, because it means that evaluation of this criterion is sum of all evaluations (for this criterion) of each partner in virtual enterprise.
- Mean aggregation type is used for criteria like quality, because it means that evaluation of this criterion is average value of all evaluations (for this criterion) of each partner in virtual enterprise.
- Distance aggregation type is used for criteria like transportation distance, because it means that evaluation of this criterion is sum of all distances between partners and theirs predecessors in production process of virtual enterprise.
- Time aggregation type is used for criteria like lead time, because it means that evaluation of this criterion is sum of all evaluations (for this criterion) of each partner in virtual enterprise. But, Time aggregation type is different than Sum aggregation type in case of parallel production. Because, in that case, for Time aggregation type only maximal evaluation of two or more parallel activities is taken into the sum of all evaluations, similar to project management techniques (Critical path method or Precedence method).

Third problem of HUMANT algorithm’s adaptation is in a priori setting of decision-makers preferences. It is a serious problem in Partner Selection Problem, because some criteria, in the sense of preference values, can be too abstract for decision-maker to define his/her preferences. For instance, if a virtual enterprise needs to manufacture 1 000 crankshafts, an industrial engineer will, perhaps, expect cost between 100 000 and 200 000 euros and lead time between 1 and 2 weeks. Therefore, he could define his preferences for these two criteria. But how much total transportation distance he can expect? Or what level of average quality score (or grade)? These criteria can vary between different production networks, but also between different virtual enterprises of the same production network. Therefore, some procedure for suggestion of decision-makers preferences must be established. In this research, this procedure is inspired by TOPSIS method (Hwang et al. 1981), in which ‘ideal’ and ‘anti-ideal’ alternative are constructed.

Abbildung in dieser Leseprobe nicht enthalten

Figure 4.6. An example of ‘ideal’ and ‘anti-ideal’ alternative creation and proposition of criteria thresholds (Mladineo et al. 2018)

These two alternatives are constructed from criteria evaluations of alternatives of multi-criteria problem. The best evaluations on each criterion form ‘ideal’ alternative, and the worst evaluations form ‘anti-ideal’ alternative. But, these two alternatives don't exist in the real-world. An example of ‘ideal’ and ‘anti-ideal’ alternative creation is presented in Figure 4.6. For criteria evaluations a specific calculation must be performed, based on criteria type. Therefore, for transportation criterion, a minimal value (for ‘ideal’ alternative, or maximal for ‘anti-ideal’ alternative) from origin-destination matrix is multiplied by total number of transports between activities, which is always total number of activities decreased by 1. For cost criterion, a minimal or maximal value from cost evaluations is multiplied by total number of activities. For quality criterion, a minimal or maximal value from quality evaluations is multiplied by total number of activities, and, in case of average quality, it is also divided by total number of activities. Additionally, 'ideal' alternative is also used as ideal optimum, which is one of parameters of the HUMANT algorithm.

Furthermore, decision-makers preferences, or thresholds of linear preference function in case of PROMETHEE method, are calculated from evaluations of ‘ideal’ and ‘anti-ideal’ alternatives. Indifference threshold is set to zero, and preference threshold is difference between evaluations of ‘ideal’ and ‘anti-ideal’ alternatives for each criterion, as it is presented in Figure 4.6. Finally, production network administrator can accept these thresholds as they are, or he/her can modified them.

HUMANT algorithm and simple PSP instance

First of all, HUMANT algorithm needs to be tested on simple PSP instance with small number of alternatives and known optimum. Therefore, PSP instance with 3 activities, 9 partners (3 different partners for each of 3 activities), 4 objectives (minimization of lead time T, minimization of cost C, minimization of transportation distance S, and maximization of average quality Q) and 27 possible solutions, will be used for experimental test. This PSP instance is labelled as PSP-TCSQ-3-9 (see Appendix for more information).

In Table 4.3, all 27 possible solutions, i.e. combinations of partners (virtual enterprises), of the instance PSP-TCSQ-3-9 are mutually compared and ranked using PROMETHEE II method.

Table 4.3. Ranking of 27 possible solutions of the instance PSP-TCSQ-3-9

Abbildung in dieser Leseprobe nicht enthalten

Experimental tests were carried out on a 2.80 GHz CPU (single CPU, because code parallelization was not used) using HUMANT algorithm written in MATLAB software. Every PSP instance was run 30 times with 500 iterations of each run. For solving PSP instance PSP-TCSQ-3-9 following objectives, weights and parameters were used:

- Objectives (criteria): minimization of total lead time T, minimization of total cost C, minimization of total transportation distance S, and maximization of average quality Q.
- Criteria weights: wT = 0.25, wC = 0.25, wS = 0.25, and wQ = 0.25 (equal weights).
- Thresholds (of linear preference function): qT = 0.01, pT = 317.00, qC = 0.01, pC = 4440.00, qS = 0.01, pS = 9620.00, qQ = 0.01, and pQ = 28.00.
- Other parameters: α = 1, γ = 1, ρ = 0.4, τ min = 0.01, τ max = 1.00, and for ideal solution sid a minimal evaluation of each criterion from Table 2 was taken.

Each run of HUMANT algorithm on this instance resulted with the same solution (Figure 4.7). In comparison with the best solution (virtual enterprise) from Table 4.3, it could be stated that HUMANT algorithm successfully finds optimal solution of this PSP instance.

Abbildung in dieser Leseprobe nicht enthalten

Figure 4.7. Solution of instance PSP-TCSQ-3-9 found by HUMANT algorithm (Mladineo et al. 2017)

HUMANT algorithm and PSP instance wu-CS-7-19

PSP instance wu-CS-7-19 (Wu, Mao and Qian 1999) consists of 7 activities and 19 partners (see Appendix for more information), and it is a problem with parallel, converging production, because it represents production of a product consisting of 4 main parts (intermediate products). The objective is to minimize total cost of production by minimizing manufacturing cost C and transportation cost S. Since, both costs have same unit, this is actually single-objective problem, but with HUMANT algorithm it will be solved as multi-objective problem. For solving this PSP instance following objectives, weights and parameters were used:

- Objectives (criteria): minimization of total manufacturing cost C, and minimization of total transportation cost S.
- Criteria weights: wC = 0.50, and wS = 0.50 (equal weights).
- Thresholds (of linear preference function): qC = 0.01, pC = 0.50, qS = 0.01, and pS = 0.50.
- Other parameters: α = 1, γ = 1, ρ = 0.4, τ min = 0.01, τ max = 1.00, and for ideal solution sid values were computed in a way that minimal evaluation of criterion C was multiplied with total number of activities, and minimal evaluation of criterion S was multiplied with total number of activities minus one activity.

Each run of HUMANT algorithm on this instance resulted with the same solution (Figure 4.8). This is the optimal solution of this PSP instance. Wu, Mao and Qian (1999) also found it using Shortest-RPT algorithm.

Abbildung in dieser Leseprobe nicht enthalten

Figure 4.8. Solution of PSP instance wu-CS-7-19 found by HUMANT algorithm (Mladineo et al. 2017)

HUMANT algorithm and PSP instance ve-CSQ-4-10

PSP instance ve-CSQ-4-10 (Veza and Mladineo 2013) consists of 4 activities, 10 partners, and 3 objectives (see Appendix for more information): minimization of total cost C, minimization of total transportation distance S, and maximization of average quality Q. This problem doesn’t have parallel, converging production, but it is interesting since it doesn’t have equal criteria weights. It means that some objectives (criteria) are more important than others. In this instance transportation distance (weight 40%) and quality (weight 45%) are much more important than the cost (weight 15%). For solving this PSP instance following objectives, weights and parameters were used:

- Objectives (criteria): minimization of total manufacturing cost C, and minimization of total transportation cost S, and maximization of average quality Q.
- Criteria weights: wC = 0.15, wS = 0.40, and wQ = 0.45.
- Thresholds (of linear preference function): qC = 0.01, pC = 24.003, qS = 0.01, pS = 276.003, qQ = 0.01, and pQ = 0.473.
- Other parameters: α = 1, γ = 1, ρ = 0.4, τ min = 0.01, τ max = 1.00, and for ideal solution sid values were computed in a way that minimal evaluation of criterion C was multiplied with total number of activities, minimal evaluation of criterion S was multiplied with total number of activities minus one activity, and maximal evaluation of criterion Q was taken.

Each run of HUMANT algorithm on this instance resulted with the same solution (Figure 4.9). This is the optimal solution of this PSP instance. Veza and Mladineo (2013) also found it using Modified-ACO algorithm.

Abbildung in dieser Leseprobe nicht enthalten

Figure 4.9. Solution of PSP instance ve-CSQ-4-10 found by HUMANT algorithm (Mladineo et al. 2017)

HUMANT algorithm and PSP instance mla-SQ-35-12

PSP instance mla-SQ-35-12 (Mladineo, Veza and Corkalo 2011) consisting of 35 activities and 12 partners (see Appendix for more information). It is the most complex PSP instance that can be found in the literature. An estimated number of solutions of this instance is more than 10[20] solutions. An optimal solution of this instance is so far unknown. This instance also represents problem with parallel, converging production. The objective is to minimize total transportation distance S and maximize average quality Q. However, transportation distance in this instance is given as reciprocal value of the distance, so the objective is to maximize total transportation distance S. Furthermore, values of both criteria are normalized. For solving this PSP instance following objectives, weights and parameters were used:

- Objectives (criteria): maximization of total transportation cost S, and maximization of average quality Q
- Criteria weights: wS = 0.50, and wQ = 0.50 (equal weights).
- Thresholds (of linear preference function): qS = 0.01, pS = 34, qQ = 0.01, and pQ = 1.
- Other parameters: α = 1, γ = 1, ρ = 0.4, τ min = 0.01, τ max = 1.00, and for ideal solution sid values were computed in a way that maximal evaluation of criterion S was multiplied with total number of activities minus one activity, and maximal evaluation of criterion Q was taken.

The best solution of this instance found by HUMANT algorithm in 30 runs is presented in (Figure 4.10). Average solution from 30 runs was S = 23.930 and Q = 0.490.

Abbildung in dieser Leseprobe nicht enthalten

Figure 4.10. Solution of PSP instance mla-SQ-35-12 found by HUMANT algorithm (Mladineo et al. 2017)

This solution, S = 25.024 and Q = 0.495, is so far the best known solution of this PSP instance. It is in average 17% better than solution founded by Mladineo, Veza and Corkalo (2011) using Modified-ACO algorithm. They founded S = 20.357 and Q = 0.445 as the best solution.

5 APPLICABLE REAL-WORLD SOLUTION FOR SUSTAINABLE PRODUCTION NETWORKS

It's true, I don't like the real world.

(R. Foreman, playwright)

Summary

In the first part of the chapter, the need for production networks in the era of the 4th industrial revolution is described and two important premises about Virtual Enterprises are given. The management of production networks is addressed through the concept of the Smart Collaborative Platform VENTIS (Virtual ENTerprise Information System) with demonstration of sustainability integration in the Partner Selection Problem on the Case Study.

Second part of the chapter defines the optimization methodology and procedure for Virtual Enterprise creation, emphasizing the adaptation of the HUMANT algorithm for this application. The special procedure for the creation of the Pull-type Virtual Enterprise, based on phenomenological method and representing a special case of a priori approach to multi-objective optimization, is described. The whole concept, procedure and optimization algorithm are demonstrated through two Case Studies using real-world enterprises’ data of potential production network from Dalmatia (Split-Dalmatia County, Croatia).

5.1 SUSTAINABLE PRODUCTION NETWORKS

From the very beginning, industrial production was seen as one of the main applications of the virtual organization. Therefore, new concepts – Agile Manufacturing (Wu et al. 1999), Production Networks (Sturgeon 2002), or Social Manufacturing (Jiang et al. 2016) – were seen as a new industrial model for world leading manufacturing industries, like the one in USA, China, and EU. On the other hand, the agile manufacturing and other new manufacturing paradigms required new type of manufacturing systems, therefore, the reconfigurable manufacturing system was presented by Koren (Koren en al 1999). Nevertheless, at the same time, idea of realizing agility and re-configurability through production networks was also born (Leigh Reid et al. 1996; Wu et al. 1999). The fact is that reconfigurable manufacturing systems are based on workplaces with reconfigurable machines. However, in production network, one workplace can be presented by one enterprise which is one whole manufacturing system itself. Since manufacturing system usually have different machines, the re-configurability and adaptability of such a workplace is guaranteed. This characteristic of production network is schematically presented in Figure 5.1.

Abbildung in dieser Leseprobe nicht enthalten

Figure 5.1. Schematic phenomenological comparison of Reconfigurable Manufacturing System and Production Network (Mladineo et al. 2018)

However, non-hierarchical production networks consisting of autonomous enterprises didn’t become fully operational in practice after more than twenty years of scientific research. In the most cases they are R&D projects or living labs. Although some global corporations are using their own production networks – the hierarchical model (Esposito el al 2014) – across continents (Netland 2014; Agostini et al. 2015) it is not at all similar concept to the original idea of non-hierarchical production networks – holarchical model (Esposito el al 2014). The main reason why production networks did not become operational in practice, probably lies in low level of ICT integration into production systems of the enterprises. Koren et al. pointed out that in order to achieve re-configurability, machines should have universal and standardized interfaces including both, hardware and software. The same thing is with enterprises, they should have standardized protocols and interfaces for information exchange, i.e. data-source interoperability (Pang et al. 2015). But few years ago, this change begun to happen through development of new industrial platform: Industry 4.0 (Kagermann et al. 2013).

Since platform Industry 4.0 has its main foundations in cyber-physical interpretation of production system (i.e. in Cyber-Physical Production System – CPPS), creating a concept known as Smart Factory (Roßgoderer et al. 2014), it has built-in ability to manage networked production systems. This introduces a new type of production network concept called Cyber-Physical Production Network (Monostori 2014), or, similarly, Network of Socio-Cyber-Physical System (Morosini Frazzona et al. 2013), or, recently, Social Manufacturing (Jiang et al. 2016).

Smart Factories have preposition to achieve information exchange in real-time across enterprises borders enabled by CPPS, thus allowing creation of flexible value chain with innovative product development and agile manufacturing. It is the reason why the production networks are seen as one of the most important elements of Industry 4.0. For instance, according to Roßgoderer et al.0, three key elements of Industry 4.0 are: cyber-physical production systems, fusion of virtual and real word in product design, and production networks as flexible value chains.

In this research, a term ‘Collaborative Production Networks’ (Huang et al. 2008) is extended with the attribute ‘Smart’ and used for description of a modern production networks based on Industry 4.0 platform. The ‘Smart Collaborative Production Network’ should be supported by the information system, which becomes quite complex one, as it will be discussed later.

Perhaps the most challenging aspect of the Virtual Enterprise information system is that it should incorporate elements of social network and decision-making. Each time when there is a need to create an new Virtual Enterprise (VE) a decision needs to be made what enterprises will be part of this new VE – the Partner Selection Problem (PSP).

From a practical point of view, three smart collaborative platforms have been developed conceptually and practically to the very high technology readiness level: PlaNet (Müller et al. 2008), ManuCloud (Meier et al. 2010), and IMAGINE (Ferreira et al. 2016). The platform for a production network planning – PlaNet is focused on a whole production process, not just a manufacturing. Special focus has been given to the Net Planning Assistant emphasizing automated negotiation process (Müller et al. 2008). The ManuCloud project and platform introduced Manufacturing-as-a-Service (MaaS) concept (Meier et al. 2010). Furthermore, IMAGINE platform represents the most comprehensive smart collaborative platform for network manufacturing. It is focused on the virtual enterprise lifecycle and partner selection (Ferreira et al. 2016) which is very similar to the supplier selection models. However, all of these researches have put focus on the different aspects of the management of collaborative production (or manufacturing), therefore it is difficult to compare them mutually, or to compare them with this research. Yet, they prove that this is an interesting research topic.

Before any discussion about methodology, algorithm and procedure for solving the Partner Selection Problem, two research premises regarding Virtual Enterprise type and sustainability must be set (Mladineo et al. 2018):

- Virtual Enterprise must be sustainable – there are two types of sustainability in the case of VEs: inner and outer sustainability. Outer sustainability is the VE’s support to sustainable development and it is focused on a harmony among 3 P’s of sustainable development: People (society), Planet (environment), and Profit (economy). Practically, it means that minimization of manufacturing costs is only one criterion among set of criteria in the Partner Selection Problem. Economic criteria, like reduction of transport, and social criteria, like selecting partner from a region with high unemployment, should be considered. Although there are researches in which virtual enterprise is evaluated only through cost (Wu et al. 1999), many researchers agree that virtual enterprise cannot be evaluated only through cost (Fischer et al. 2004; Lanza et al. 2010; Veza et al. 2013; Mladineo et al. 2015) but through set of criteria, therefore, multi-criteria (multi-objective) approach becomes mandatory. On the other hand, inner sustainability represents ability of VE to sustain for a long term. This characteristic is linked with the quality (Veza et al. 2013) of enterprises and (Fischer et al. 2004) among enterprises that are selected to be part of new VE. These kind of criteria set aims to create long-term sustainable virtual enterprise, although it is not easy to define them. Quality of the enterprise represents some kind of the excellence level of the enterprise, therefore, it must be possible to evaluate level of excellence of enterprises, in order to evaluate level of excellence of potential VEs. For evaluation of VE's excellence some multi-criteria decision-making methods have been already used: Fischer et al. used AHP method, Lanza et al. and Veza et al. used PROMETHEE, Yu et al. used TOPSIS, etc.
- Virtual Enterprise can be Push-type or Pull-type – there are two main types of procedures to create VE. First step for both types is to create Partner Selection Problem (PSP) based on customer requests. PSP will consist of activities which arise from customer requests. For each activity a set of enterprises will be created, consisting of enterprises that can execute that activity. An evaluation of criteria data for each enterprise must be made using production network database. First type of procedure is a Push-type and it is based on finding optimal solution of created PSP. Solution represents combination of enterprises with some level of excellence. Therefore optimization algorithm in combination with multi-criteria decision-making method must be used to find optimal virtual enterprise (Fischer et al. 2004; Mladineo et al. 2015; Mladineo et al. 2017). Second type of procedure is a Pull-type and it is based on negotiations in which enterprises are placing bids for activities of their interest. It is a negotiation procedure that can be fully or semi-automated (Neuberta et al. 2004; Cao et al. 2015). After negotiation procedure, a limited number of possible combinations, i.e. a set of potential virtual enterprises is created regarding placed bids. That set of potential VEs is submitted to multi-criteria decision-making method and the best VE is selected (Lanza et al. 2010). There are also some extended procedures (Veza et al. 2013), but generally Pull-type VEs are result of negotiations between enterprises, and Push-type VEs are initiated by VE information system. The difference between a Push-type VE and Pull-type VE is in the fact that Push-type is not easy to implement into real-world, since it is ‘forcing (pushing)’ the creation of new VE. Push-type selects optimal partners (enterprises) but without asking them if they have an interest for that project. On the other hand, Pull-type is more likely to be implemented in the real-world, since only the enterprises that are interested in some project are evaluated and eventually selected to be part of the new VE.

Production network management

Virtual Enterprise information system represents the complex information system (Jeroen et al. 2002; Botezatu et al. 2006), since it has the elements of the multi-agent system, it also has elements of the holonic system (Müller 2006; Shen et al. 2006). It is a multi-agent system, because each enterprise of production network is an independent agent that makes its own decisions and collaborates with other agents (Figure 5.2a). It is as multi-agent system, because each enterprise represents a production system by itself, but when enterprises mutually collaborate as virtual enterprise they represent one virtual production system (Figure 5.2b).

Another important aspect of VE information system is the management, not just of a product lifecycle, but of a virtual enterprise lifecycle. Two different, but very similar, VE lifecycles are described by Leigh Reid et al. and Camarinha-Matos et al. consisting of: customer requests, creation of VE, operation and conduction/evolution of VE, and dissolution of VE. Beside the VE lifecycle management (time tracking), a spatial management of the products is important, because the enterprises are geographically-dispersed.

Abbildung in dieser Leseprobe nicht enthalten

Figure 5.2. (a) Elements of Multi-Agent System in production network; (b) Elements of Holonic System in production network (Mladineo et al. 2018)

All above mentioned extends VE information system into a complex system with following features: extended database (extended with spatial attributes), product / VE lifecycle management, and collaboration and negotiation platform. So, VE information system goes beyond an information system thus creating a platform to support human stakeholders, i.e. to support: negotiations among enterprises, decision-making based on the ranking of the enterprises (Babic et al. 1998) and enabling the sustainable collaboration among enterprises. It needs a cloud computing as an infrastructure, so VE information system becomes Smart Collaborative Platform. It is a platform similar to the idea of cloud manufacturing (Tao et al. 2014) and mainly used for collaborative product development (Wang et al. 2012, Chen et al. 2014) emphasizing decision-making (Girodon et al. 2015; Germani et al. 2012) and negotiations. However, new manufacturing paradigms require decision-making support in preparation of production orders (Elgh 2012) thus creating automated production planning and scheduling (Zhong et al. 2015) extended to the whole supply network (Kuo et al. 2012).

Regarding the decision-making, the main focus is on the selection of the optimal partners for manufacturing, product development, or some other phase of the VE lifecycle. As already mentioned, that optimization problem is called the Partner Selection Problem. In this research, solving of Partner Selection Problem has been put in the context of VE creation. It is not solved as an isolated optimization problem, and it is the reason why two special procedures of VE creation have been defined later in this chapter.

Smart Collaborative Platform VENTIS

In this research a focus has been put on two elements of Smart Collaborative Platform: GIS (Geographic Information System) platform and negotiation platform. Negotiations occur in first phase of VE lifecycle: creation of new virtual enterprise. It is one of the most important phases of VE lifecycle, because it needs to result with sustainable VE. That means that selection of appropriate partners is very important. In this research, above described procedures for creation of VE Push-type and VE Pull type are applied on real production network from Croatian region – Dalmatia, using special Smart Collaborative Platform called VENTIS (Virtual ENTerprise Information System) developed in Laboratory for Industrial Engineering at Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture (FESB), University of Split.

Since spatial aspect of VE is important, it is recommended that Smart Collaborative Platform is based on GIS instead of non-spatial information system. Therefore ESRI ArcGIS for Organization has been used as database platform for VENTIS. Furthermore, the whole VENTIS represents an extended Decision Support System (DSS) with three main features of DSS emphasized: data (GIS), model (collaboration, management and decision-making tools) and user interface (Web application). Its main decision-making tool is negotiation platform based on HUMANT algorithm (Mladineo et al. 2015) used for solving Partner Selection Problem and PROMETHEE method (Brans et al. 1984) used for multi-criteria comparison of VEs, in MATLAB environment. S iemens PLM Teamcenter is used as PLM platform, and Virtual Enterprise Lifecycle Management (VELM) and Collaboration platform has not yet been developed. Schematic design of the VENTIS platform is shown in Figure 5.3. A detailed database with all relevant attributes of enterprises has been created on GIS server. Attributes include spatial, general, technological, organizational and personnel information of each enterprise. An extract of enterprise attributes and GIS representation of production network as Web application (Figure 5.4).

Abbildung in dieser Leseprobe nicht enthalten

Figure 5.3. Schematic design of Smart Collaborative Platform VENTIS and its main elements (Mladineo et al. 2018)

Abbildung in dieser Leseprobe nicht enthalten

Figure 5.4. (a) Extract of enterprise attributes from the database; (b) Web application of VENTIS (Mladineo et al. 2018)

Regarding negotiation platform, it is a Decision Support System (DSS) with main task to support process of creation of new Virtual Enterprise. This task includes partner identification and partner selection, supported by optimization algorithm and MCDM method. In this case, HUMANT algorithm and PROMETHEE method are used for it.

Case Study: Integration of sustainability into Partner Selection Problem

Sustainability is generally seen through three P’s: People (society), Planet (environment), and Profit (economy). Therefore, Partner Selection Problem with these 3 groups of criteria (objectives) is used in this Case Study. The production network consisting of 6 enterprises has been created, and the manufacturing process with 4 steps needs to be processed (Figure 5.5).

Abbildung in dieser Leseprobe nicht enthalten

Figure 5.5. Manufacturing process and production network as the Case Study to solve

Enterprises are geographically dispersed in 6 different European cities. In total, 7 criteria are used, but only 6 of them have known evaluations (Table 5.1). The seventh criterion is transport, but its evaluation is unknown before optimization, so it varies depending on the selected partners (enterprises). Evaluation of transport is made using Euclidean distance (Table 5.2). Criteria information and parameters for PROMETHEE method are given in Table 5.3. Criteria evaluations for the cities are criteria evaluations for the countries to which the particular city belong, so they are not precise. These data have been collected from Eurostat database (Eurostat 2017). Some of the data are for 2017, and some of the data are for 2015 or 2016. The manufacturing process with potential partner (enterprise) for each technological step (sub-process) is presented in Table 5.4. It is the PSP that needs to be solved by using HUMANT algorithm, and taking into account all input data (Table 5.1, 5.2, 5.3).

Table 5.1. Evaluations of enterprises (partners) on economic, environmental and social criteria

Abbildung in dieser Leseprobe nicht enthalten

Table 5.2. Euclidean distance matrix between cities (enterprises)

Abbildung in dieser Leseprobe nicht enthalten

Table 5.3. Criteria information and parameters for PROMETHEE method

Abbildung in dieser Leseprobe nicht enthalten

Table 5.4. Steps of manufacturing process with potential partner (enterprise) for each step

Abbildung in dieser Leseprobe nicht enthalten

Since HUMANT algorithm, i.e. PROMETHEE method, allows setting the values for criteria weights, a set of 7 scenarios has been made, where each scenario represents different set of weights (Table 5.5).

Table 5.5. Scenario set for partner selection problem

Abbildung in dieser Leseprobe nicht enthalten

HUMANT algorithm has been used to solve proposed PSP based on 7 different scenarios. Algorithm was found 7 solutions, one for each scenario. However, some of the solutions, found for different scenarios, are identical. At the end, there are 3 different combinations of partners (3 virtual enterprises) for these 7 scenarios. They are presented in Table 5.6 and Figure 5.6.

Table 5.6 – Scenario results with values for each criterion

Abbildung in dieser Leseprobe nicht enthalten

Abbildung in dieser Leseprobe nicht enthalten

Figure 5.6. (a) Optimum with map representation for Scenario 1, 2, 4 and 6; (b) Optimum with map representation for Scenario 3; (c) Optimum with map representation for Scenario 5 and 7
(Mladineo, Veza, Gjeldum and Crnjac 2018)

A visual scenario comparison is given in Figure 5.7. It is important to note that six criteria have minimization as an objective, and only one criterion (C5 – Unemployment rate) has maximization as an objective. Because, that criteria belongs to the group of social criteria, and social criteria are used as social sensibility toward areas (countries) with higher unemployment (maximization as an objective) and lower GDP per capita (minimization as an objective).

Abbildung in dieser Leseprobe nicht enthalten

Figure 5.7. Scenario comparison on radar chart (Mladineo, Veza, Gjeldum and Crnjac 2018)

It is also important to note that each criterion has its own aggregation type. In this PSP instance, six criteria have ‘mean’ as aggregation type, which means that a mean value will be used to evaluate (aggregate) total value for that criterion. A mean value is calculated, since the PSP instance has four technological steps for which four partners need to be selected. So, it is a mean value of criterion evaluations for selected partners. For transport, a sum is used as an aggregation value. It means that Euclidean distance between selected partners is summed taking into account transportation steps defined by the manufacturing process.

At the end, it is very important to highlight that in the term of multi-criteria decision-making, three solutions found by the algorithm are not three different optimal solutions, but three optimal compromises. Namely, in multi-criteria decision-making it is always selection of the best compromise, because optimal solution would mean that it is the solution better than all others on each criterion, or better on at least one criterion and equal on all other criteria. In the reality, such a kind of optimal solution usually does not exist. Therefore, in sustainability optimization, it is always search for the best compromise, like in case of production networks.

Aim of this Case Study was to demonstrate that sustainability of production network could be maximized by selecting optimal combination of partners (enterprises) based on multi-criteria approach which focuses on the three P’s of the sustainable development: People (society), Planet (environment), and Profit (economy).

An instance of Partner Selection Problem, consisting of six enterprises (partners), manufacturing process with four steps and seven criteria (objectives), has been successfully solved using the HUMANT algorithm. Seven criteria were grouped into three groups that represent sustainability: economic criteria, environmental criteria, and social criteria. Furthermore, since the HUMANT algorithm uses PROMETHEE method for multi-criteria decision-making, seven different scenarios were defined representing seven different sets of criteria weights. Some scenarios resulted with same solution, so, at the end, three solutions were found for seven scenarios. These solutions have elements of a Pareto set of optimal solutions, but from the perspective of multi-criteria decision-making these solutions are the best compromises. It is clear that approach presented in this Case Study have a potential for further research and development of information system for the sustainable management of production networks.

5.2 PHENOMENOLOGICAL APPROACH TO THE VIRTUAL ENTERPRISE CREATION

In this research, solving of the Partner Selection Problem (PSP) has been put in the context of VE creation. It means it is not solved as an isolated optimization problem, but in the context of the real-world application. Therefore, two special procedures of the VE creation have been made for two different types of the VE: Push-type VE and Pull-type VE. These procedures are mandatory to have a real-world applicability of the solving of Partner Selection Problem.

Procedure for creation of Push-type VE

A process consisting of a set of activities must be realized in order to satisfy customer request. In manufacturing, these activities are technological processes which form one side of Partner Selection Problem. Other side of PSP is formed by set of enterprises which possess technology for proposed technological processes. Creation of PSP is called partner identification (Leigh Reid 1996), and it is not an easy task without Smart Collaborative Platform. Once the PSP has been created, it is submitted to multi-objective HUMANT algorithm. Since a priori approach to multi-objective optimization is used, decision-makers preferences must be a priori defined (Talbi 2009). But, if multiple criteria, like cost, lead time, transportation cost and quality, are used, it becomes difficult to completely define decision-makers preferences. Because the decision-makers preferences consist of two aspects: perception of scale of the decision-maker (parameters of preference function in case of PROMETHEE method by Brans et al.) for each criterion, and weight of each criterion. Weights assignment can be much easier task than defining perception of scale. If customer is ready to pay an expensive high-quality product, then criterion quality will get higher weight than criterion cost. But, how to define perception of scale for each criterion? An experienced decision-maker can have a perception of scale for criteria like cost or lead time, but what about criteria like transportation distance or quality level? They can vary for different PSP instances and different production networks. Therefore, in this research, an automatic procedure based on analysis of PSP input data suggests perception of scale, i.e. parameters of preference function of PROMETHEE method, for each criterion. These suggested parameters can be accepted by decision-maker or modified a little bit. This issue will be discussed more, later.

With a priori data defined, HUMANT algorithm can solve created PSP and find optimal combination of enterprises, optimal VE. The whole above described procedure for creation of Push-type VE is presented in Figure 5.8. Since this kind of production network should be non-hierarchical, an existence of some kind of production network administrator, instead of production network manager, is assumed. However, it is difficult to define production network administrator, since it is an issue that is missing in the literature. Production network needs to be somehow administered, but it has to remain non-hierarchical. Therefore function of production network administrator should only be assurance of the fair processes and fulfilment of the missing data in customer requests. However, the possibility of the misuse remains, since some enterprises could figure out that they will have better chances to be selected if they offer lower prices, and that group of enterprises could make a secret deal that they will mutually offer lower prices just to get all the jobs. At the end they can make different deal with customer, regarding the final price of the product, and with new mutual contracts share income. It is a problem of oligopoly (or cartel) which is present in everyday economy. Non-hierarchical approach of production network cannot solve the problem of oligopoly, but, if the oligopoly is detected, production network administrator could expel these enterprises from smart collaborative network. Unfortunately, non-hierarchical production network can only rely on ‘in bono fide’ principle.

Abbildung in dieser Leseprobe nicht enthalten

Figure 5.8. Procedure for creation of Push-type VE (Mladineo et al. 2018)

Procedure for creation of Pull-type VE

Like in procedure for creation of Push-type VE, customer request that triggers virtual enterprises lifecycle creates Partner Selection Problem. However, in Pull-type a different procedure follows since customer request creates an internal offer inside VENTIS. This offer, which is intended for manufacturing of some parts, is sent to enterprises of the production network. Enterprises interested in the offer place their bids for the technological processes they have. Generally, VENTIS is selling some kind of job and enterprise is placing bid price to get that job. However, the aim of VENTIS is to choose bid with lowest price. And, it is important to note that the price, i.e. cost for customer, is not, and shouldn’t be, the only criterion that is used to compare and rank enterprises. Usually, criteria like quality, transportation cost, and lead time, are also used. From set of bids a set of potential VEs is constructed by calculating all possible combinations of enterprises which placed bids. Set of VEs is then submitted to some multi-criteria decision-making method and the best VE, from set of potential VEs, is selected.

However, in this research, procedure to create Pull-type VE is extended in order to achieve higher sustainability of VE, similar to Veza et al. (2013). Extended procedure is inspired with phenomenological reduction (epoché), which is a part of phenomenological scientific method (Marx 1987). Its steps are presented in Figure 5.9. Phenomenological reduction is a procedure in which a priori knowledge about object is ‘bracketed’, i.e. set aside (Marx 1987). Main aim of phenomenological reduction is to transform observer of object into uninterested observer, thus reducing observer’s subjectivism (Stein 1989). In this application, ‘bracketing’ of knowledge about potential VEs will assure decision-makers objectivism, thus creating transparent process.

Abbildung in dieser Leseprobe nicht enthalten

Figure 5.9. Phenomenological reduction as first step of phenomenological scientific method (Mladineo et al. 2018)

It is important to highlight that a priori knowledge of potential VEs differs from a priori knowledge of decision-makers preferences. The second one refers to a priori approach to multi-objective optimization (Talbi 2009), in this case applied to Partner Selection Problem, and means that decision-makers preferences are known. However, the first one refers to a priori knowledge about object, in this case potential VE, is something completely different. It is a knowledge about ranks of a set of VEs, where ranks were obtained using some Multi-Criteria Decision-Making (MCDM) method. However, ranking is always insufficient information about excellence of alternatives. For instance, ranks of three alternatives can be known, but all three alternatives could actually be `bad` ones. In that case ranking only defines which alternative is least `bad`. To solve this issue, alternatives must be sorted (Nemery et al. 2008), i.e. compared with at least two reference points. Therefore, in this procedure of creation of Pull-type VE, when set of potential VEs is created based on bids which enterprises placed, set of VEs is submitted to the MCDM method, but results are not taken as granted. Results, or ranks, represent a priori knowledge about potential VEs which is about to be ‘bracketed’, like in phenomenological reduction. It means that VE with 1st rank is not a priori taken as the best one, but it must be tested to prove that this VE is `good` one and to estimate how `good` it really is. To test it, reference points must be found. Figuratively, observer using phenomenological method must discover that stone, which he observes, is actually a stone. The same idea is used in this procedure: discover that potential VEs are ‘good’ or, perhaps, ‘bad’ ones.

The reference points in this case are the theoretical-best solution (optimum) and the theoretical-worst solution (pessimum) of the Partner Selection Problem. Using HUMANT algorithm to find an optimum, original criteria objectives (maximization or minimization) are used, but to find a pessimum, criteria objectives are inverted: maximizing criterion become minimizing, and vice versa. After that, PROMETHEE method is used to compare set of potential virtual enterprises with two reference points (optimum and pessimum). That comparison shows how really ‘good’ one virtual enterprise is. Because, it is possible that virtual enterprises, from the set of potential enterprises, are all ‘good’, or all are ‘bad’, or some are ‘good’ and some are ‘bad’. Without reference points it is impossible to make such a conclusion.

After comparison of set of VEs with reference points, a decision to accept or reject top-ranked VE is made. The whole above described procedure for creation of Pull-type VE is presented in Figure 5.10.

Abbildung in dieser Leseprobe nicht enthalten

Figure 5.10. Procedure for creation of Pull-type VE (Mladineo et al. 2018)

Case Study: VE creation in the Production network of Split-Dalmatia County

Case Study used to evaluate presented algorithm and procedures is a real production network consisting of 9 enterprises from Dalmatia (Split-Dalmatia County, Croatia). This production network is not operational in the reality, but these 9 independent enterprises agreed to be part of a production network for this research. In order to maintain privacy of the data, names of stars are used as code names instead of real names of enterprises. These 9 enterprises form a small Production network of Split-Dalmatia County, which is manufacturing, emphasizing machining, oriented.

In previous research (Mladineo et al. 2013), information were gathered for these 9 manufacturing using detailed Questionnaire on 19 pages. This dataset represents set of competences of production network, which these 9 enterprises form. Excellence level of some competences can be described qualitative or quantitative, so they can be used to evaluate enterprises in term of quality. Some kind of quality score or grade can be assigned to each enterprise (Babic et al. 1998; Fischer et al. 2004). Quality score will be used to mutually compare enterprises and to calculate an overall quality of virtual enterprise. Furthermore, GIS is used to calculate transportation distances based on real traffic routes, not on Euclidean distances.

This real production network is used to experimentally test, not just the HUMANT algorithm, but the whole phenomenological procedure for creation of Pull-type VE. There is no need to experimentally test procedure for creation of Push-type VE solely, since one part of Pull-type VE procedure is the same as the Push-type.

Smart Collaborative Platform VENTIS and its procedure for creation of Pull-type VE are experimentally tested using PSP instance from the literature (Veza et al. 2013). This PSP instance represents manufacturing process of the part ‘Bench-vise jaw’, consisting of 4 different technological processes: milling, drilling, countersinking, and threading. Three criteria (objectives) are used: minimization of total manufacturing cost C, minimization of total transportation distance S, and maximization of average quality Q. However, the partners and their data of original PSP instance are changed, since partners (i.e. enterprises) from real production network will be used.

This PSP is turned into customer request and sent to VENTIS. Inside VENTIS, an internal offer for manufacturing of 2500 parts of the ‘Bench-vise jaw’ is sent to 9 enterprises of the production network (Figure 5.11).

Abbildung in dieser Leseprobe nicht enthalten

Figure 5.11. Placing customer request as an internal offer inside VENTIS platform (Mladineo et al. 2018)

Enterprises interested in bidding place their bids for technological processes they have. Two cases are simulated: in first case bidding process results with set of VEs, which are less or more ‘good’ ones; in second case bidding process results with set of VEs, which are ‘bad’ ones. In the first case, decision-maker would accept the best VE and end negotiation process. In the second case, decision-maker would decline the best VE and repeat negotiation process hopping that some new bids will be placed. In both cases, to simplify input data, same bid price is used for each technological process.

Case 1: ‘The best Virtual Enterprise is a good one’

In this case, placing of bid prices is simulated in a way that it can result with virtual enterprise which is really ‘good’ one. Simulated bids are shown in Table 2. From a set of bids a set of potential virtual enterprises is constructed by creating all possible combinations of enterprises that placed bids (Table 3). It is important to mention that criterion ‘Quality grade’ represents evaluation of the overall quality level of an enterprise. It represents the assessment of all enterprise's competences, including technology, organization, information systems, quality assurance, customer relationship, etc. (Babic et al. 1998; Fischer et al. 2004).

Next step is to evaluate criteria data of the constructed virtual enterprises (Table 5.7). Bid prices are summed for each virtual enterprise to calculate total cost. Average quality is evaluated using quality scores (or grades) of enterprises from the database. Transportation distance is evaluated using GIS tool for the shortest path.

Table 5.7. Enterprises of production network and placed bids (Mladineo et al. 2018)

Abbildung in dieser Leseprobe nicht enthalten

Table 5.8. Virtual Enterprises constructed from all possible combinations of enterprises that placed bids (Mladineo et al. 2018)

Abbildung in dieser Leseprobe nicht enthalten

Table 5.8 clearly shows that the best VE is VE-5, since it has the best evaluations of criteria values in all three criteria, but in this case it is not so important. The a priori fact that VE-5 is the best is ‘bracketed’, according to the phenomenological procedure presented in this research. Because, VE-5 is the best among other a priori created potential virtual enterprises, but is it a ‘good’ one? Is it sustainable? The excellence of the VE-5 needs to be tested to prove its sustainability. So, in the procedure for creation of Pull-type VE, a priori created potential virtual enterprises (Table 5.8) should be compared with reference points, i.e. with the best combination of enterprises (optimum) for this PSP and the worst one (pessimum).

Original criteria objectives (maximization or minimization) are used to find an optimum, but to find a pessimum, criteria objectives are inverted: maximizing criterion become minimizing, and vice versa. In order to make this optimization possible, a missing data must be filled in (Table 5.9). For enterprises that didn’t place bid price, it must be estimated. Estimation is based on the data of some previous projects, or enterprise can be asked to estimate its price (i.e. cost).

Table 5.9. Enterprises of production network and placed bids (Mladineo et al. 2018)

Abbildung in dieser Leseprobe nicht enthalten

Another input is needed for the optimization process: decision-makers preferences, consisting of weight and perception of scale for each criterion. In this example, weight for manufacturing cost C was set to wc = 35%, weight for transportation distance S was set to ws = 15%, and weight for average quality Q was set to wq = 50%. Perceptions of scale, i.e. criteria thresholds, were obtained by HUMANT algorithm.

Experimental tests were carried out on a 2.80 GHz CPU (single CPU, because code parallelization was not used) using HUMANT algorithm written in MATLAB. Average run of 100 iterations was about 12 seconds. This kind of simple Partner Selection Problem with 4 activities and 9 partners can have up to several thousand possible solutions, but problems with more than 20 activities and more than 10 partners can have more than 1020 possible solutions. That’s why metaheuristic algorithm is used, instead of deterministic approach.

In Figure 5.12, the best combination of enterprises (optimum) and the worst one (pessimum) is presented. They form two virtual enterprises VE-Optimum and VE-Pessimum that will be used as reference points.

Abbildung in dieser Leseprobe nicht enthalten

Figure 5.12. Optimum and pessimum of Partner Selection Problem found by HUMANT algorithm (Mladineo et al. 2018)

As it has already been mentioned, it is clear that VE-5 is the best overall on each criterion, but the question is: how ‘good’ or how ‘bad’ all other virtual enterprises are? The reference points will help answer this question. Software Visual PROMETHEE is used to compare all eight virtual enterprises using PROMETHEE II method. The complete input matrix for PROMETHEE II method is shown in Table 5.10.

Table 5.10. Input matrix for PROMETHEE II method (Mladineo et al. 2018)

Abbildung in dieser Leseprobe nicht enthalten

The virtual enterprises (schematically represented in VENTIS Web application) and results of their mutual comparison using PROMETHEE Diamond diagram are presented in Figure 5.13.

Abbildung in dieser Leseprobe nicht enthalten

Figure 5.13. (a) Schematic representation of virtual enterprises in VENTIS (Case 1); (b) Comparison of virtual enterprises on PROMETHEE Diamond diagram (Case 1) (Mladineo et al. 2018)

It is clear that all six virtual enterprises are much closer to VE-Optimum than VE-Pessimum, so it can be concluded that they are all ‘good’. According to their excellence level, they are grouped into three groups: VE-1 and VE-3; VE-2, VE-4 and VE-6; and VE-5. Without reference points, it was not possible to see for instance that VE-5 is much better than VE-1 and VE-3. In this case, top-ranked virtual enterprise (VE-5) is accepted as sustainable VE and process of creation of virtual enterprise will continue.

Case 2: ‘The best Virtual Enterprise is a bad one’

In this second case, placing of bid prices is simulated in a way that it cannot result with virtual enterprise which is really ‘good’ one, but all virtual enterprises are ‘bad’ ones. Simulated bids are shown in Table 5.11.

Table 5.11. Enterprises of production network and placed bids (Mladineo et al. 2018)

Abbildung in dieser Leseprobe nicht enthalten

Like, in the first case, from set of bids a set of potential virtual enterprises is constructed by calculating all possible combinations of the enterprises which placed bids (Table 5.12).

Table 5.12. Virtual Enterprises constructed from all possible combinations of enterprises that placed bids (Mladineo et al. 2018)

Abbildung in dieser Leseprobe nicht enthalten

In the next step, the criteria data of constructed potential virtual enterprises are evaluated (Table 5.12), therefore input matrix for PROMETHEE II method can be completed (Table 5.13).

From Table 5.12 it is unclear which VE is the best one, since none of the enterprises is overall on each criterion. But solving their ranking is not priority, since it is not just a question which VE is the best one, but also which VE is a ‘good’ one, and which is the ‘bad’ one. Again, a priori created VEs need to be compared with reference points (Table 5.13). Schematic representation and results of comparison of virtual enterprises are presented in Figure 5.14.

Table 5.13. Input matrix for PROMETHEE II method (Mladineo et al. 2018)

Abbildung in dieser Leseprobe nicht enthalten

From Figure 17, it is clear that all six virtual enterprises are much closer to VE-Pessimum than VE-Optimum, so it can be concluded that they are all ‘bad’ ones. Furthermore, all six enterprises are below middle point between VE-Optimum and VE-Pessimum, which is taken as threshold of acceptance or reject.

Abbildung in dieser Leseprobe nicht enthalten

Figure 5.14. (a) Schematic representation of virtual enterprises in VENTIS (Case 2); (b) Comparison of virtual enterprises on PROMETHEE Diamond diagram (Case 2) (Mladineo et al. 2018)

So, in this case, top-ranked virtual enterprise (VE-1) is rejected, because it does not represent sustainable VE. Therefore, process of creation of virtual enterprise will not continue. Biding procedure will repeat, in hope that some other enterprises will place their bids in new bidding cycle. If that doesn’t happen, this production network should reject this Partner Selection Problem, i.e. these customer requests.

To conclude, Smart Collaborative Platform called VENTIS for transparent and efficient management of production network has been presented in this research. It is a Smart Collaborative Platform based on phenomenological approach that helps decision-makers to create sustainable virtual enterprise in the process of the partner selection. Since new industrial platforms, like Industry 4.0, are seeking for intelligent, automated and multi-objective algorithms, in this research, the HUMANT algorithm has been adapted to solving of the multi-objective Partner Selection Problem. The PSP instance from the literature has been modified with the real data from 9 enterprises from one Croatian region, and used in two case studies as a proof of concept of developed Smart Collaborative Platform VENTIS. It has been demonstrated how VENTIS successfully uses the Pull-type procedure for creation of sustainable virtual enterprise. But, it is important to note that results of these two cases, presented through PROMETHEE method, cannot be mutually compared. Although the difference between optimum and pessimum looks similar, they have different values and difference between them will vary from comparison to comparison. To have mutually comparable results, each alternative (visual enterprise) should be independently compared with reference points (optimum and pessimum). That can produce more stable and more comparable results. Additionally, to have completely stable results, some sorting method like FlowSort (Nemery et al. 2008) could be used, and this could be the topic for the future research.

6 ANTHROPOLOGICAL ANALYSIS OF PRODUCTION NETWORKS

Then God said: ‘Let us make humankind in our image, according to our likeness, and let them have dominion.’

(Genesis, the Bible)

Summary

In this chapter, the problem of lack of trust among enterprises is addressed. The result of survey among enterprises is presented, showing that enterprises have more complaints on the idea of collaboration, than praises. Since such attitudes are devastating for the production network concept, the anthropological analysis is made in order to find possible solutions of this issue.

Through analysis of anthropological problems of the human being, the anthropological problems of the enterprise that affect collaboration are addressed. The concept of Collaborative Platform is presented as a potential solution for trustful and transparent collaboration.

6.1 CHALLENGES OF THE TRUSTFUL COLLABORATION

According to Adam Smith, every entrepreneur is led by his/her idea to invest money in the business which brings him/her the benefit. However, it seems that some “invisible hand” of the market (Smith 2018) which leads entrepreneurs toward self-interested actions with the most self-benefit, actually creates the unintended social benefits for everyone: entrepreneurs, citizens and state. Usually, it is through growth of economy which actually increases social welfare of everyone.

This idea of “the invisible hand”, which represents the free market concept, was well functioning till the beginning of the 20th century when many things changed and recessions became a common occurrence. In 1930s, during the Great Recession, John Maynard Keynes realized that influence of the government on the market, through different kind of interventions, is mandatory in order to keep the market stable for a long-term. Furthermore in 1950s, John Forbes Nash realized that entrepreneurs, especially corporations, are not led by the idea to maximize their benefits, but instead they are driven by fear of competition, loss of market share, or similar; thus minimizing overall benefits. To conclude, today, the idea of “the invisible hand” is officially out!

The free market on which enterprises maximize their overall benefit would serve as an ideal base for creation of virtual enterprises. The maximization of benefits, for instance through collaboration and joint projects, is an excellent proposition for production network, however, this idea is often not present among enterprises.

On the contrary, enterprises or entrepreneurs usually have more complaints on the idea of collaboration, than praises. The survey among enterprises (Mladineo 2014) showed a huge lack of trust among entrepreneurs in the perspective of collaboration or joint projects through production network (Figure 6.1).

Abbildung in dieser Leseprobe nicht enthalten

Figure 6.1. Some results of survey on how entrepreneurs see the advantages or disadvantages of collaboration through production network (Mladineo 2014)

In order to understand why entrepreneurs do not trust each other very often (Figure 6.1), the anthropological dimension must be taken into account. Behind every enterprise are humans and every entrepreneur is a human being, therefore anthropological analysis is needed in order to understand this phenomena of the collaboration problems. This issue is addressed in the following section in order to find a way how to establish transparent and trustful collaboration.

6.2 ANTHROPOLOGICAL DIMENSION OF THE COLLABORATION PROBLEMS

Influenced by Greek philosopher Plato and some aspects of Aristotle’s philosophy, Christian philosophers Augustine of Hippo and Thomas of Aquino have developed the foundations of the anthropological thought in the medieval Europe. This anthropological thought is very specific, because it has one very important premise, which is also its starting point: human is the wounded being, wounded in his essence (Argüello 2014).

The wounded essence of the human being is manifested through the fact that human can act in his free will, in his freedom, in a way that he loses his freedom at the end. Literally, a man can freely walk into the cage, decide to lock it, and threw the key through the cage bars. So, by acting freely, in complete freedom, this man ended locked-up into the cage by his own free will. He lost his freedom due to his own free will. This paradox explains why some people act badly in a real-world and end up in jail, it explains why some people ruin their lives by alcoholism or drug addiction, but it doesn’t explains what the real cause of such an acting is. There is only conclusion that humans avoid aspiration toward perfection and wisdom, and, instead, they are focused on the survival of their own being in the psychological and physical aspect. They are focused on search for happiness and pleasure in material things, and that search usually doesn’t fill all psychological needs, leaving some kind of the emptiness inside (Argüello and Hernández 2012).

To understand why people act diametrically opposite from the path to perfection it is important to present the human and his challenges and temptations with a simplified sketch (Figure 6.2).

Abbildung in dieser Leseprobe nicht enthalten

Figure 6.2. Human and his most important functions confronted with challenges and temptations

Humans can be presented by their three main and most important functions: the reason, the will, and the experience. There are also other functions – imagination, perception, affection, sensuality, etc. – but they are less important for this analysis. There are also three main challenges for the human (Ratzinger 2007): to become the Divine image i.e. to seek for the perfection, to dominate and rule the Earth, and to be fertile and populate the Earth. These three challenges are described in the Book of Genesis which is the Holy Scripture for Jews and Christians.

However, due to the wounded essence of the human being, the three temptations exists (Argüello and Hernández 2012): the bread – to meet all physical needs, the miracle – the golden fish phenomena or to correct all bad things in the life, and the idol – the mean (usually, it is the money) used to meet all physical needs and to correct all bad things.

Due to existence of these temptations, it is possible to manipulate with the people. The well-known concept of Roman Emperors ‘ panem et circenses’ (bread and circuses) is solving first two temptations: the bread and the miracle (through games and fun in the circus). And, at the end, people adore their new idol – mighty Roman Emperor, who gave them bread and let them forget bad things in their life through fun and games. For instance, two totalitarian leaders of the 20th century – Stalin in the Soviet Union and Tito in Yugoslavia – used the similar approach: they were giving people bread through economically unjustified construction of many factories, and through different kind of games (relay races) and activities (parades) organized by the state. People of their nations adore them, like the idols, although both of them were cruel totalitarian leaders.

Therefore, the temptations are the path that offers itself to the human, and the challenges represent the path that human must take himself. It is also important to note that the matter of the existence is the first temptation (the bread) and the third or the last challenge (to populate the Earth). So, the path of temptations is to meet the physical needs first and at any cost, and the path of challenges is to suppress physical needs and meet them last. The path of challenges puts aspiration toward perfection and wisdom on the first place, so the question is: who can take such a path by himself? The answer is: no one! Many tried through the history, no one succeeded (Argüello 2014). For instance, Greek philosopher Socrates is usually considered as the wisest man in the history of a mankind, but he killed himself when he was falsely accused for spoiling the youth with his teaching. The wisest man committed a suicide, proving that there is a path that no one was capable to take. No one… until “The Light” come to the world, sometime around year 1 AD (Figure 6.3).

Abbildung in dieser Leseprobe nicht enthalten

Figure 6.3. The intervention of “The Light” into the human existence

“The Light”, in Christianity known as Jesus Christ, established the way by revealing the truth and giving the life to everyone. The three additional gifts to protect human against temptations are: faith, hope and love. Today, every person on the planet can take this way, which is opened for everybody: the Way which is the path to perfection, the Truth how to govern the Earth without endangering the planet and sharing resources with everybody, and the Life through populating the planet with new brilliant young minds who will shape the better future for the humanity. The temptation of bread makes every newborn child looks like the hungry mouth to feed, but this child will become a good worker, engineer, or genius who will help the humanity (Wojtyla 2005). Furthermore, it is important to highlight the fact that the will should made first step toward perfection, not the reason. All Greek philosophers thought that the reason can bring them toward perfection, but the faith is the matter of the will, not matter of the reason. For philosopher Blaise Pascal, the faith is not an intellectual certainty, it is “a leap into the dark”, but he had a same wrong attitude, as Greek philosophers, of taking the faith as a matter of the reason. The faith starts with the will which is subjected to Divine’s will, led by the experience of fidelity, thus creating the certainty for the reason, and this certainty is manifested as love (Argüello and Hernández 2012).

To understand presented anthropological dimension in terms of the enterprise, and how it affects the collaboration among enterprises, another sketch is needed (Figure 6.4).

Abbildung in dieser Leseprobe nicht enthalten

Figure 6.4. Enterprise and its most important functions confronted with challenges and temptations

The will of the enterprise, or the entrepreneur (the owner), is manifested through the activities of the enterprise. These activities should aim on becoming the perfect production system (first challenge) by constant striving toward perfection. Similar idea of continuous improvement is the basis of famous Toyota Production System and exists as one of the Lean improvement tools: Kaizen. By moving toward perfection, the products of such a system will also become near perfect, thus allowing the enterprise to dominate the market (second challenge). Since, everyone would like to have a near perfect product, the enterprise should be highly productive (third challenge) to fulfil the demand, but that shouldn’t be a problem, because the near perfect production system will ensure such a productivity.

However, in the reality, the leadership of the enterprise is striving to survive in the market and it is led by the fear of competition, so the activities of the enterprise are not aiming on becoming the perfect system. Strive for survival leads to negative activities of abusing the employees or unnecessary layoffs of employees in order to cut the costs or to gain the income. Fear of competitors, leads enterprise into competitive wars for the domination on the market resulting with stealing of patents and technologies, thus leading to legal cases on the courts. At the end, the high profit idolatry creates unsustainable economy which leads to economic crisis, endangers the environment, and threats safety and health of employees. Enterprises and entrepreneurs which are functioning this way will never be interested in trustful collaboration, nor will ever become a trustful partner.

In order to achieve trustful and transparent collaboration and avoid negative behavior of the enterprises, some kind of the Collaborative Platform needs to be established (Figure 6.5).

Abbildung in dieser Leseprobe nicht enthalten

Figure 6.5. Establishment of the Collaborative Platform for trustful and transparent collaboration among enterprises

The Collaborative Platform could help establishing connections among enterprises through some kind of social network thus minimizing negative competitiveness and competitive wars. The joint projects supported by Collaborative Platform could lead to more sustainable production and help preserve the environment and the workplaces. All this should move away the focus of the leadership from the individual survival in the market to joint survival based on collaboration and better usage of the free manufacturing capacities.

In the previous section, Collaborative Platform called VENTIS (Mladineo et al. 2018) is presented and it represents a first prototype of a platform for trustful and transparent collaboration, which addresses some of anthropological problems presented in this section.

7 CONCLUSION: PRODUCTION NETWORKS IN THE DIGITAL ERA

A goal without a plan is just a wish.

(A. de Saint-Exupéry, writer)

Summary

Instead of the conclusion, this chapter presents some of the new production features brought by digital era, which are essential for production networks. First of all, it is the product traceability based on real-time manufacturing execution system and powered by the RFID technology. The description of the concept and technology, and the demonstration of the vertical integration of production with product traceability are given in the first part of chapter.

In the second part of chapter, the digitalization of the production is presented through description of Smart Enterprise and Lean automation, emphasizing the digitalization of the Lean tools and methods. At the end, the main guidelines for the development of the enterprise toward the Smart Enterprise are presented.

7.1 PRODUCTION NETWORKS AND TRACEABLE SMART PRODUCTS

The Industry 4.0 emphasized the Smart Factory concept (Kagermann et al. 2013), which represents the production based on Smart Products (Uhlemann et al. 2017). Smart Products are unique, so they need to be identifiable, located at any time and they need to know their own history (Gladysz et al. 2016), current status and alternative routes to achieving the customer, thus enabling the product traceability in the sense of spatial location and in the sense of its production process. This is very important for the production networks in order to enable the tracking and managing of the path of the product during its production process.

However, taking into account all features of the Smart Products, it is clear they require more sophisticated technology than usual product tracking and tracing technology like barcodes. The technology that can enable these requirements is radio-frequency identification (RFID) technology (Gladysz et al. 2016). This technology is based on RFID tags for storing data into their memory, and RFID antennas to read data from the tag or write data to the tag. RFID technology is already well-known technology, therefore it could be implemented into Manufacturing Execution System (MES), thus creating RFID-enabled Manufacturing Execution System (Zhong et al. 2015). This kind of live tracking of manufacturing execution connected with Enterprise Resource Planning (ERP) system can significantly improve production planning. The main aim of RFID-enabled Manufacturing Execution System is to have real-time manufacturing execution data, i.e. to have real-time MES.

The main layers of the real-time MES that create MES framework are (Huang et al. 2008): shop-floor layer with various hardware devices (RFID readers, Wi-Fi network, or similar); MES layer; interface layer that aims at real-time intercommunicating with other enterprise information systems (ERP in general); and decision-making layer consisting of the information systems, like the ERP system (Level 4 in ISA-95 Automation Standard, see Brandl 2018). However, MES (also called Manufacturing Operations Management or MOM, Level 3 in ISA-95) cannot be seen as a part of ERP, nor it is completely subordinated to ERP. MES is overlapping with ERP, because it gives an insight into real-time manufacturing execution (Levels 2 and 1 in ISA-95). If the manufacturing plans are not executed as planned, it triggers re-planning and decision-making affecting the changes in ERP and other information system.

Case Study: Demonstration of the vertical integration of production

Three years ago, construction of the assembly line and its product for demonstration of the Industry-4.0-based Smart Factory concept has started in the ‘Lean Learning Factory @ FESB’ at the University of Split (Mladineo et al. 2019). An intelligent assembly line (vertically integrated) and a special product called 'Karet' (children vehicle, popular in Dalmatia) were designed and realized (Figure 7.1).

On a simple way, a complete vertical integration of the production has been demonstrated within laboratory (Figure 7.2). The Industry-4.0-ready equipment has been installed on 4 workplaces on which assembly process takes place. It is the following equipment: RFID system for product, i.e. manufacturing execution, tracking connected to Programmable Logic Controller (PLC); Windows 10 Tablet which serves as the Human Machine Interface (HMI) for controlling the PLC, but it also represents simple MES (RFID-enables) connected to the ERP.

Abbildung in dieser Leseprobe nicht enthalten

Figure. 7.1. Intelligent assembly line: a) Product ‘Karet’ for assembly; b) Assembly line based on the Smart Factory concept (Mladineo et al. 2019)

Abbildung in dieser Leseprobe nicht enthalten

Figure. 7.2. Demonstration of the vertical integration of production (Mladineo et al. 2019)

The ERP system used in this case is ‘Venio ERP’ by Venio Indicium Ltd. This ERP system is based on the MS SQL database thus all the data can be accessed from some other desktop or web application by using authorized connection. This fact allows development of the connection between MES and the ERP system.

The complete Hardware/Software scheme of vertical integration is presented in Figure 7.3. Very important part of the system is the embedded Web Server of the PLC Siemens S7-1200. It allows communication over HTTP with PLC by using any device that supports HTTP protocol (PC, laptop, tablet, smartphone, etc). In this case, PLC Web Server is used for the web page on which user (worker) can input data for writing on the RFID tag, or to read the data from the RFID tag. With the help of the ASP. NET Web Server, these data (from the RFID tag) can be sent to the ERP system, or in other direction (the data can be read from the ERP system and, at the end, written to the RFID tag). Part of this solution represents simple Manufacturing Execution System, although MES is not used just for product tracking only, that’s why this is a very simple MES.

Abbildung in dieser Leseprobe nicht enthalten

Figure. 7.3. Hardware/Software scheme of the presented vertical integration of production (Mladineo et al. 2019)

Regarding the RFID system, there are two main RFID standards for high frequency (HF) readers operating at 13.56 MHz: ISO-14443 for ‘proximity’ reading/writing (range up to 10 cm), and ISO-15693 designed for industry for ‘vicinity’ reading/writing (range up to 100 cm). The ISO-14443 RFID tags are mostly based on FRAM memory and they usually have memory higher than 1024 bytes. However, the industrial ISO-15693 RFID tags are mostly based on EEPROM memory and they usually have memory around 125 bytes. Therefore, it is important to note that industrial RFID tags (ISO-15693) are usually designed with low memory. The reason is in fact that higher memory means higher writing/reading time, and in industrial environment it is important to have fast and efficient reading/writing of the RFID tags, thus low memory is used. However, this can come in contradiction with Industry 4.0 idea of writing complete product data on the RFID tag (Kagermann et al. 2013). The idea was to write all important product data in readable XML format by using some of the industrial standards, like the B2MML (Figure 7.4).

Abbildung in dieser Leseprobe nicht enthalten

Figure. 7.4. Example of material data written in XML format (Brandl 2008)

Since, material data presented in Figure 7.4 are ‘string’ data type and ‘string’ has 405 characters (without ‘white spaces’), it equals 413 bytes. So, it takes a lot of memory and time to write these data on the RFID tag. According to TURCK Company, writing of 413 bytes (using ISO-15693 protocol) to FRAM type RFID tag takes about 0.3 – 0.4 seconds. So, it seems much easier to store only product ID to RFID tag, and link the ID with some database in the cloud where all other product data will be available. But, from the perspective of production networks, it is better to avoid this kind of approach, because than enterprises of the production network should be sharing their databases or having some centralized database with products data, which is hard to realize in the real-world. Centralized database should be avoided in order to keep the non-hierarchical model of the production network and, on the other hand, enterprises are not willing to give access to their information system and cloud to external users from other enterprises. Therefore, the product data should be on the RFID tag, but the balance between the amount of data and R/W speed should be established. Only the most important data should be on the RFID tag, for instance: ID, lot number, customer, etc. With proper product traceability, the production network can be seen as a scalable production system with different levels (Figure 7.5): shop-floor level, regional production network level, or global production network level.

Abbildung in dieser Leseprobe nicht enthalten

Figure. 7.5. Production Network as scalable production system (Mladineo 2014)

The RFID technology has its limitation, but it is very promising technology for the product traceability which is important issue in production networks, but in Industry 4.0, as well. Future research and development are needed to increase the R/W speed of the RFID tags and RFID protocols in order to have more efficient and reliable system.

7.2 PRODUCTION NETWORK OF SMART ENTERPRISES

Among new terms that brought Industry 4.0, the most important one is Cyber-Physical Production System (CPPS). It explains the concept of operations management system, on which the Smart Factory or Smart Enterprise concept is based-on. Generally, CPPS consists of three main elements (Monostori 2014):

- Physical world (Physical layer) – physical assets of production system, like machines and other equipment, plus ICT devices that are part of Internet of Things (IoT);
- Cyber world (Cyber layer) – or digital twin of production system enhanced with the data from IoT devices stored in the computer cloud;
- Interface (Abstraction layer) – an interface between cyber world and physical world.

An example of CPPS is presented in Figure 7.6. It is the real laboratory for production research and its digital representation with digital objects representing IoT devices that mutually communicate within the information system. They report problems, communicate with each other, communicate with humans using human-machine interfaces, etc.

Abbildung in dieser Leseprobe nicht enthalten

Figure 7.6. Main elements of Cyber-Physical Production System (CPPS)

The CPPS, by collecting all important data from production system and using data visualization, becomes smart and supports human decision-making, i.e. it becomes the Smart Factory or Smart Enterprise. It is the reason why the CPPS is, also, seen as one of the key elements of the Industry 4.0.

The main advantages and challenges of the proposed CPPS are (Zühlke 2010; Monostori 2014):

- Smart products – product which fits the customer’s exact needs and which is uniquely identifiable, may be located at all times and knows its own history, current status and alternative routes to achieving customer;
- Single-item production – CPPS should help plan and control of the single-item products, and make it as much economical as serial mass production;
- Production without rigid plans – CPPS should manage production without rigid plans, since it is needs to manage production of single-item products;
- Energy-efficiency – CPPS represents a backbone for more energy-efficient manufacturing and decisions that results with environmental-friendly solutions;
- Cyber Security – CPPS and its computer cloud should be more secure from hacker attacks than IT systems of today, since it is a most weak point of the CPPS;
- Production networks – collaborative product development, collaborative manufacturing and all other value adding processes supported by mutually networked CPPSs.

In context of the Smart Enterprise, CPPS enables not only successful management and real-time control of the production, but also continuous improvement of the business processes.

Smart Enterprises and Lean Automation

On the other hand, Lean management (also known as Lean production, Lean thinking, Lean philosophy, or, simply, Lean) has been on the scene for more than thirty years, becoming the most successful approach to business improvement (Netland and Powell 2017). It was born and developed initially by the Japanese car manufacturer Toyota after World War II, but became known on a global level in the 1990s. It represents a holistic framework for management, production planning and control, and continuous improvement of business processes. The main aspects on which Lean focuses are customer, value stream, and continuous improvement. More precisely, Lean is based on five principles (Womack and Jones 1996): define value for the customer, identify the value stream, create a flow by eliminating waste, establish the pull principle, and seek perfection.

Furthermore, the Lean is often described as “doing what is needed, when it is needed, where it is needed” (Womack and Jones 1996). Even if this might seem intuitive, the Lean reality is a lot more complex, consisting of over more than 40 different tools and methods (Netland and Powell 2017). So, the question is: how does the number one improvement method – the Lean – fit into the new industrial platform Industry 4.0?

The fact is that Industry-4.0-based process management doesn’t exclude Lean as a process improvement platform, so they are complementary, although there are only few researches that combine these two topics. The combined research of Lean management and Industry 4.0 could produce a solution for connecting organization-and-people-oriented Lean and technology-oriented Industry 4.0, thus connecting all three most important aspects of a production system: technology, organization and people. That kind of a research should be focused on mutual integration of Lean and Cyber-Physical Production System.

Abbildung in dieser Leseprobe nicht enthalten

Figure 7.7. Connecting Lean and Industry 4.0

A conceptual approach for connecting Lean with Industry 4.0 is presented in Figure 7.7. To be ready for Industry 4.0, the production system needs to be vertically integrated by connecting all operation, control and management levels of the production. The advantage of Lean is that is already vertically integrated through its principles, methods and tools. However, the challenge remains to horizontally integrate these two different, but vertically integrated, systems or platforms.

The aim of the connecting Lean and Industry 4.0 is to create some kind of a decision support system that will support human decision-making on the management level, production control level, but, also, on the shop-floor lever, by enabling new production concepts, like, for instance, a production without rigid plans. The idea of the production without rigid plans relays on empowering of lower organizational levels, like the shop-floor level, with a decision-making functions.

Although mutual integration of Lean and CPSS seems to be complex, it could be crucial for new business models based on Industry 4.0 platform. Some authors call this integration the Lean automation (Kolberg et al. 2017). An example of Lean automation is presented in Table 7.1.

Table 7.1. Examples of use cases to combine Industry 4.0 with Lean production (Kolberg et al. 2017)

Abbildung in dieser Leseprobe nicht enthalten

To conclude, it is important to emphasize there is a misunderstanding of the word ‘cyber’ in the context of Industry 4.0. Etymologically, word ‘cyber’ means: of, relating to, or involving computers or computer networks – ‘the cyber marketplace’. But, origin and etymology of the word ‘cyber’ is: ‘cybernetic’. The word ‘cybernetic’ means: the science of communication and control theory that is concerned especially with the comparative study of automatic control systems (such as the nervous system and brain and mechanical-electrical communication systems). Origin and etymology of the word ‘cybernetic’ is kybernētēs (Greek) with meaning ‘pilot’, ‘governor’.

Therefore, Industry 4.0 is not just about modern IoT devices and gadgets, it is about creating the Smart Enterprise, i.e. the CPPS that supports human decision-making and has its own intelligence and cognitive capabilities to optimize production. Lean with its principles, methods and tools serves as an excellent toolbox to achieve that aim.

In the context of production networks, this kind of improvement through CPPS and Lean guarantees vertically integrated production system, thus enabling the horizontal integration of such a production systems into the production network.

Why is the Lean particularly important for production networks? Because, the Lean methodology was one of the first organizational improvement methodologies that was promoting horizontal integration among factories in the supply chain. However, supply chains, also known as global production networks, are conceptually different from non-hierarchical production networks (Figure 7.8).

Abbildung in dieser Leseprobe nicht enthalten

Figure 7.8. Comparison of non-hierarchical production network and global production network or supply chain

Namely, in non-hierarchical production networks, the technological or other processes which are sub-processes of production process are out-sourced; and production of parts, components and sub-systems are outsourced in supply chains (global production networks). It is much easier for the enterprise to be part of the supply chain and to supply some component to the OEM. But, to be part of non-hierarchical production network and to do some manufacturing service in particular time for particular single-item product is much harder. The enterprise needs to be able to acquire just-in-time or just-in-sequence principles to be able to react fast, and the easiest way to acquire these principles is through Lean (Kanban for pull-production system, SMED for lead time minimization, etc). It is the reason why the Lean is so much important for enterprises of production networks.

Lean is about establishing the pull system and production network is completely based on the pull system, so the Lean fits in perfectly. And through digitalized Lean tools or Lean Automation, by using Kaizen tool or similar, the enterprises can share information about problems with products or their production system making the management of production networks easier.

Except the Lean, the enterprise should become Smart Enterprise based on Cyber-Physical Production System, and network of CPPSs then creates Cyber-Physical Production Network – CPPN (Monostori 2014). However, in order to achieve this, the enterprises of the production network should take huge improvement steps toward Smart Enterprise. Sometimes it is not completely clear which are the most important steps, so in the following section a brief guidelines for Smart Enterprise development are given and the most important features are described.

Guidelines for Smart Enterprise development

Every organization and institution needs to change constantly, which means it needs to adapt to development of the world, but it also needs to harmonize its inner processes (Holy See Press Office 2016). Regarding the industrial enterprises, the good starting point for the adaptation and transformation is through conceptual definition of the production system model (Figure 7.9), similar to Toyota Production System model (Womack and Jones 1996).

Abbildung in dieser Leseprobe nicht enthalten

Figure 7.9. Model of production system for Croatian Innovative Smart Enterprise (Veza et al. 2016)

After definition of the production system model with its main priorities, values and tools/methods, the concrete steps and objectives of development can be defined. In this research, the six objectives are proposed as important aspects of development toward Smart Enterprise:

- Personal responsibility – People (employees) are most important part of any organization. Therefore, an improvement attempt without support of employees will fail. If each person doesn’t take his/her part of responsibility for the process or part of process in his/her domain, there will be no improvement. While working in Toyota, Rother described special approach toward organizational changes and change of the behavior of employees. He called it Toyota Kata (Rother 2009). He created an approach for employees to adopt new behavior patterns similar to the approach of teaching of new move (kata) in martial arts. The people have natural resistance toward changes, but through Kata approach and established mentorship they are helped to overcome it, thus raising the personal responsibility of the employee. Not just for his tasks, but for the whole process of value adding.
- Customer-focused – According to Koren, the new manufacturing paradigm is the personalized production with two different types (Koren 2010): personalization of the product (product designed to meet customer needs) and regionalization of the product (different design of the product for geographically different markets). Therefore, integrating the ‘Voice of the Customer’ into the product design is crucial. Regarding the personalized product, the Web product configurator is an excellent option to integrate customer needs. However, today’s product configurators need to be more simplified allowing the customer to define his preferences, not to define all technical characteristics of the product (width, height, mass, etc). For instance, a customer must be able to select price as most important factor and his favorite color as less important factor, and product configurator should configure product in selected color, unless it makes the product too expensive. Since this kind of production is the Pull system, the optimal product design is modular design, which means product is designed as a platform which can be customized by adding different modules. The example of such a design is using ‘skateboard’ chassis for a car, instead of classic ‘frame’ chassis.
- Process-oriented – Every organization has its key processes, its core business. To be process-oriented means to identify core business processes and all other supporting processes (process mapping) and to have process performance indicators (process metric). Value Stream Mapping represents an excellent Lean tool (Netland et al. 2017) to identify and map processes and to establish their metric. With Industry 4.0, the Value Stream Mapping diagrams can be enhanced with automated process metric which is collected directly from the shop-floor level through Manufacturing Execution System. Furthermore, responsibility for the process is the horizontal responsibility, which is different from hierarchical vertical responsibility, so the person responsible for the process must be appointed regardless already defined hierarchical responsibilities. These appointments will also raise the awareness of importance of process-orientation among employees.
- Transparent and efficient organization – Every organizational department has its own competences. So, each organizational unit is important, but the organization as a whole is important as well. This awareness of the collective is very important, since processes are mostly shared among different organizational departments or units. The sharing of processes is leading toward collaboration. Therefore, sharing of data and information is crucial to have transparent and efficient processes, thus creating transparent and efficient organization. The Enterprise Resource Planning (ERP) system serves as a platform for sharing all important data, information, and more. But it is important that virtual organization inside ERP system is exactly the same as the real-world organization, thus representing its digital twin. Many enterprises simplify their virtual organization, and that is not a good decision.
- Functional improvement – Since every organizational department has its own functional competences, the local optimization of the organizational unit is equally important as the global optimization of the whole organization. There are several ways to optimize or functionally improve the organizational units. The easiest way is to define each unit as the profit center or the cost center, thus optimize it through profit increase or the cost cutting. A step further is the concept of the fractal, when organization is divided into the smaller organizations (fractals) with their own authority and budget. Regarding the cost cutting, many Lean tools can serve to do this: 5S, SMED, Heijunka, Andon, etc. However, it is important to have in mind that Lean implementation benefits are in form of the S-curve (Netland 2014): at the beginning more effort is needed for not much of a benefit, but during further Lean implementation the benefits are rapidly raising and the effort reduces.
- Technological modernization – At the end, the technological modernization is needed as well in order to keep the pace with the development of the world. Today, the enterprise needs to have vertically integrated production based on the establishment of connection between the ERP system and the Manufacturing Execution System, thus creating the Cyber-Physical Production System. The significant technological changes are needed to achieve this, but, it will enable new type of production planning without rigid long-term plans. All this can result with more efficient and environmentally friendly production based on short-term plans and single-item production.

7.3 FUTURE RESEARCH

The future already started a decade ago with the beginning of the Age of Connectedness (Figure 7.10) which emphasizes the networking and collaboration. It is not just a business paradigm, the people all over the world are mutually networking through web social networks like Facebook, Instagram, Google Plus, etc. and through smartphone apps that support grouping and networking like WhatsApp, Viber, Skype, etc.

Abbildung in dieser Leseprobe nicht enthalten

Figure 7.10. Global business paradigms (Guillory et al. 2004)

Since the networking become a global paradigm, the future research is not about networking, it is about establishing collaboration inside these networks. The future challenge for production network is to have automated negotiations (Neuberta et al. 2004; Cao et al. 2015) and to have automated optimization algorithms (Zhong et al. 2015) for instance to solve Partner Selection Problem in process of creation of Virtual Enterprise. All these challenges represent some simple form of artificial intelligence and intelligent algorithms are needed to support it.

For manufacturing networks, i.e. production networks without product development, the Social Manufacturing as a form of social network for manufacturing enterprises is already taking place (Jiang et al. 2016). The future research challenge in social manufacturing is also automation of negotiations, but it is also the question of the cyber security and trust among partners inside Cyber-Physical Production Network (Monostori 2014) or Network of Socio-Cyber-Physical System (Morosini Frazzona et al. 2013).

Despite the Age of Connectedness and networking as a global paradigm, the problem of trustful collaboration, caused by its anthropological dimension, remains. But, with establishment of collaborative platforms and systems, the future of Production Networks, especially within Industry 4.0, looks bright, very bright.

As a general and final conclusion: the most important moment in the history of Production Networks was a day when they met Industry 4.0.

APPENDIX: SUPPLEMENTAL MATERIAL FOR THE PARTNER SELECTION PROBLEM INSTANCES

Supplemental material

Schematic representations and criteria evaluations of the Partner Selection Problem (PSP) instances used in this research are given in this supplement. For each PSP instance a figure with schematic representation of a problem and geographic map of production network, and two tables with criteria evaluations, are used for complete definition of a problem. The data of these PSP instances are not real-world data, they are made up for illustrative purpose.

PSP instance PSP-TCSQ-3-9

PSP instance PSP-TCSQ-3-9 consists of 3 activities, 9 partners (3 different partners for each of 3 activities), 4 objectives (minimization of lead time T, minimization of cost C, minimization of transportation distance S, and maximization of average quality Q) and 27 possible solutions. Schematic representation of this PSP instance is shown on Figure 8.1, and criteria evaluations are given in Table 8.1 and Table 8.2.

Abbildung in dieser Leseprobe nicht enthalten

Figure 8.1. Schematic representation of the PSP instance PSP-TCSQ-3-9

Table 8.1. Evaluations of criterion S (transportation distance) of the instance PSP-TCSQ-3-9

Abbildung in dieser Leseprobe nicht enthalten

Table 8.2. Evaluations of criterion T (lead time), criterion C (cost) and criterion Q (quality) of the instance PSP-TCSQ-3-9

Abbildung in dieser Leseprobe nicht enthalten

PSP instance wu-CS-7-19

PSP instance wu-CS-7-19 (Wu, Mao and Qian 1999) consists of 7 activities and 19 partners, and it is a problem with parallel, converging production, because it represents production of a product consisting of 4 main parts (intermediate products). The objective is to minimize total cost of production by minimizing manufacturing cost C and transportation cost S. Schematic representation of this PSP instance is shown on Figure 8.2, criteria evaluations are given in Table 8.3 and Table 8.4.

Abbildung in dieser Leseprobe nicht enthalten

Figure 8.2. Schematic representation of the PSP instance wu-CS-7-19

Table 8.3. Evaluation of criterion S (transportation cost) of the instance wu-CS-7-19

Abbildung in dieser Leseprobe nicht enthalten

Table 8.4. Evaluations of criterion C (cost) of the instance wu-CS-7-19

Abbildung in dieser Leseprobe nicht enthalten

PSP instance ve-CSQ-4-10

PSP instance ve-CSQ-4-10 (Veza and Mladineo 2013) consists of 4 activities, 10 partners, and 3 objectives: minimization of total cost C, minimization of total transportation distance S, and maximization of average quality Q. This problem doesn’t have parallel, converging production, but it is interesting since it doesn’t have equal criteria weights. It means that some objectives (criteria) are more important than others. In this instance transportation distance (weight 40%) and quality (weight 45%) are much more important than the cost (weight 15%). Schematic representation of this PSP instance is shown on Figure 8.3, and criteria evaluations are given in Table 8.5 and Table 8.6.

Abbildung in dieser Leseprobe nicht enthalten

Figure 8.3. Schematic representation of the PSP instance ve-CSQ-4-10

Table 8.5. Evaluations of criterion S (transportation distance) of the instance ve-CSQ-4-10

Abbildung in dieser Leseprobe nicht enthalten

Table 8.6. Evaluations of criterion C (cost) and criterion Q (quality) of the instance ve-CSQ-4-10

Abbildung in dieser Leseprobe nicht enthalten

PSP instance mla-SQ-35-12

PSP instance mla-SQ-35-12 (Mladineo, Veza and Corkalo 2011) consisting of 35 activities and 12 partners is the most complex PSP instance that can be found in the literature. An estimated number of solutions of this instance is more than 10[20] solutions. An optimal solution of this instance is so far unknown. This instance also represents problem with parallel, converging production, because it represents production of a product consisting of 7 main parts (intermediate products). The objective is to minimize total transportation distance S and maximize average quality Q. However, transportation distance in this instance is given as reciprocal value of the distance, so the objective is to maximize total transportation distance S. Furthermore, values of both criteria are normalized. Schematic representation of this PSP instance is shown on Figure 8.4, and criteria evaluations are given in Table 8.7 and Table 8.8.

Table 8.7. Evaluations of criterion S (reciprocal normalized transportation distance) of the instance mla-SQ-35-12

Abbildung in dieser Leseprobe nicht enthalten

Abbildung in dieser Leseprobe nicht enthalten

Figure 8.4. Schematic representation of the PSP instance mla-SQ-35-12

Table 8.8. Evaluations of criterion Q (normalized quality) of the instance mla-SQ-35-12

Abbildung in dieser Leseprobe nicht enthalten

BIBLIOGRAPHY

Ackermann J., Müller E. (2007). Modelling, planning and designing of logistics structures of regional competence-cell-based networks with structure types. Robotics and Computer-Integrated Manufacturing 23, pp. 601-607

Agostini L., R. Filippini, A. Nosella (2015). Management and performance of strategic multipartner SME networks, International Journal of Production Economics 169 376-390.

Argüello K. (2014). The Kerygma: In the Shantytown with the Poor. Ignatius Press, USA, 2014.

Argüello K., Hernández C. (2012). Neocatechumenal way – Catechetical directory for teams of catechists. Hope Publishing House, USA, 2012.

Assimakopoulos N.A., Theodosi A.D. (2003). A Systemic Approach for Modeling Virtual Enterprise’s Management Features. Tamkang Journal of Science and Engineering Vol. 6, No. 2, pp. 87-101

Babic Z., N. Tomic-Plazibat (1998). Ranking of enterprises based on multicriterial analysis, International journal of production economics 56-57 29-35.

Babic Z., T. Peric (2014). Multiproduct vendor selection with volume discounts as the fuzzy multi-objective programming problem, International Journal of Production Research 52 (14) 4315-4331.

Beume, N., Naujoks, N., Emmerich, M. (2007). SMS-EMOA: Multiobjective selection based on dominated hypervolume. European Journal of Operational Research, 181, 1653-1669.

Bölt A., Freitag M. (2000). Non-hierarchical Regional Networks - Theories, Models, Methods and Instruments - a Research Agenda. Proceedings of the 7th International Conference on Multi-Organizational Partnerships and Co-operative Strategy, Leuven, Belgium

Botezatu C.P., C. Botezatu, G. Căruţaşu (2006). Virtual Enterprise Information System Requirements, International Conference Competitiveness and Stability in the Knowledge-Based Economy, 2006.

Brandl D. (2018). What is ISA 95? Industrial Best Practices of Manufacturing Information Technologies with ISA-95 Models, http://www.apsom.org/docs/T061_isa95-04.pdf, 03-10-2018.

Brans J.P., Mareschal B. (1994). PROMCALC & GAIA: a new decision support system for multicriteria decision aid. Decision Support Systems 12, pp. 297-210

Brans J.P., Mareschal B., Vincke P.H. (1984). PROMETHEE - a new family of outranking methods in multicriteria analysis. Operational Research IFORS 84, Amsetrdam, Netherlands

Brans J.P., Vincke P., Mareschal B. (1986). How to select and how to rank projects: The PROMETHEE method. European Journal of Operational Research 24, pp. 228-238

Camarinha-Matos L.M., Afsarmanesh H. (2007). A framework for virtual organization creation in a breeding environment. Annual Reviews in Control 31, pp. 119-135

Camarinha-Matos L.M., H. Afsarmanesh (2001). Virtual Enterprise Modeling and Support Infrastructures: Applying Multi-agent System Approaches, ACAI 2001 - LNAI 2086, pp. 335-364.

Cao M., X. Luo, X. Luo, X. Dai (2015). Automated negotiation for e-commerce decision making: A goal deliberated agent architecture for multi-strategy selection, Decision Support Systems 73 1-14.

Chen L., W. Wang, B. Huang, A negotiation methodology for multidisciplinary collaborative product design, Advanced Engineering Informatics 28 (4) (2014) 469-478.

Choi K.H., Kim D.S., Doh Y.H. (2007). Multi-agent-based task assignment system for virtual enterprises. Robotics and Computer-Integrated Manufacturing 23, pp. 624-629

Choi Y., Kim K., Kim C.. (2005). A design chain collaboration framework using reference models. Int. J. Adv. Manuf. Technol. 26, pp. 183-190

Chuanga C.L., Chiangb T.A., Cheb Z.H., Wang H.S. (2009). Using DEA and GA Algorithm for Finding an Optimal Design Chain Partner Combination. Proceedings of the 16th ISPE International Conference on Concurrent Engineering, Taipei, Taivan

Coello, C.A.C., Lamont, G.B., Van Veldhuizen, D.A. (2007). Evolutionary Algorithms for Solving Multi-Objective Problems. New York: Springer.

Corvello V., Migliarese P. (2007). Virtual forms for the organization of production: A comparative analysis. Int. J. Production Economics 110, pp. 5-15

Deb, K., Kumar, A. (2007). Interactive evolutionary multi-objective optimization and decision-making using reference direction method. Proceedings of the 9th annual conference on Genetic and evolutionary computation GECCO '07, 781-788.

Deb, K., Pratap, A., Agarwal, S., Meyarivan, T. (2002). A fast and elitist multiobjective genetic algorithm: NSGA-II. IEEE Transactions on Evolutionary Computation, 6, 182-197.

Demeter K. (2013). Time-based competition – the aspect of partner proximity, Decision Support Systems 54 (4) 1533-1540.

Dorigo, M., Maniezzo, V., Colorni, A. (1996) The Ant System: Optimization by a colony of cooperating agents. IEEE Transactions on Systems, Man, and Cybernetics - Part B, 26, 1-13.

Ekici A. (2013). An improved model for supplier selection under capacity constraint and multiple criteria, International Journal of Production Economics 141 (2) 574-581.

Elgh F. (2012). Decision support in the quotation process of engineered-to-order products, Advanced Engineering Informatics 26 (1). 66-79.

Eppe, S. (2009). Application of the Ant Colony Optimization Metaheuristic to Multi-objective Optimization Problems. Technical report – ULB, Bruxelles.

Eppe, S. (2009). Integrating the Decision Maker's Preferences into Multi Objective Ant Colony Optimization. 2nd Doctoral Symposium on Engineering Stochastic Local Search Algorithms – SLS 2009, Bruxelles, Belgium.

Eppe, S., De Smet, Y. (2014). Approximating PROMETHEE II's net flow scores by piecewise linear value functions. European Journal of Operational Research, 233, 651-659.

Esposito E., P. Evangelista (2014). Investigating virtual enterprise models: literature review and empirical findings, International Journal of Production Economics 148 145-157.

Eurostat: European statistics (2017). http://ec.europa.eu/eurostat/, 15-11-2017.

Farquhar, P.H. (1977). A survey of multiattribute utility theory and applications. Multiple Criteria Decision Making - TIMS Studies in the Management Sciences, 6, 59-90.

Ferreira J., J. Sarraipa, M. Ferro-Beca, C. Agostinho, R. Costa, R. Jardim-Goncalves (2016). End-to-end manufacturing in factories of the future, International Journal of Computer Integrated Manufacturing.

Fischer M., Jähn H., Teich T. (2004). Optimizing the selection of partners in production networks. Robotics and Computer-Integrated Manufacturing 20, pp. 593-601

Ganß M., Baum H., Schütze J., Ivanova R. (2011). Flexibility instruments in SME: an empirical study. 21st International Conference on Production Research - Conference Proceedings, Stuttgart, Germany

Gao F., G. Cui, Q. Zhao, H. Liu (2006). Application of Improved Discrete Particle Swarm Algorithm in Partner Selection of Virtual Enterprise, International Journal of Computer Science and Network Security 6 208-212.

Gerber A., Dietzsch M., Althaus K. (2004). Information based, dynamic quality information system for non-hierarchic regional production networks. Robotics and Computer-Integrated Manufacturing 20, pp. 583-591

Germani M., M. Mengoni, M. Peruzzini (2012). An approach to assessing virtual environments for synchronous and remote collaborative design, Advanced Engineering Informatics, 26 (4). 793-813.

Giret A., Botti V. (2009). Engineering Holonic Manufacturing Systems. Computers in Industry 60, pp. 428-440

Girodon J., D. Monticolo, E. Bonjour, M. Perrier (2015). An organizational approach to designing an intelligent knowledge-based system: Application to the decision-making process in design projects, Advanced Engineering Informatics 29 (3) 696-713.

Gladysz B., Lysiak C. (2016). Light-responsive RFID Tags for Precise Locating of Objects in Manual Assembly Verification Workshops. Procedia CIRP, 41, 951-956

Gong, D., Tang, M., Liu, S., Li, Q. (2017). Reconsidering production coordination: A principal-agent theory-based analysis. Advances in Production Engineering & Management, 12, 51-61.

Guillory W.A., Harding C., Guillory D. (2004). The FuturePerfect Organization: Driven by Quantum Leadership. USA, Innovations International, Inc. 2004.

Hedberg B, G. Dahlgren, J. Hansson, N.G. Olve (1997). Virtual Organization and Beyond, Wiley & Sons, 1997.

Ho W., X. Xu, P.K. Dey (2010). Multi-criteria decision making approaches for supplier evaluation and selection: A literature review, European Journal of Operational Research 202 16-24.

Holland, J. (1975). Adaptation in Natural and Artificial Systems. Ann Arbor: University of Michigan Press.

Holy See Press Office (2016). Greetings to the Roman Curia: permanent personal and structural conversion. https://press.vatican.va/content/salastampa/en/bollettino/pubblico/2016/12/22/161222b.html, 01-07-2019.

Hongzhao D., Dongxu L., Yanwei Z., Ying C. (2005). A novel approach of networked manufacturing collaboration: fractal web-based extended enterprise. Int. J. Adv. Manuf. Technol 26, pp. 1436-1442

Huang C.Y., C.C. Huang, C.Y. Liu (2008). Order confirmation mechanism for collaborative production networks, International Journal of Production Research 46 (3) 595-620.

Huang G.Q., Zhang Y.F., Jiang P.Y. (2008). RFID-based wireless manufacturing for real-time management of job shop WIP inventories. International Journal of Advanced Manufacturing Technology, 36, 752-764.

Hwang C.L., K. Yoon (1981). Multiple Attribute Decision Making: Methods and Applications, Springer-Verlag, New York, 1981.

Iredi, S., Merkle, D., Middendorf, M. (1993). Bi-Criterion Optimization with Multi Colony Ant Algorithms. Computer Science, 359-372

Ivanov, D.; Zschorn, L.; Teich, T.; Sokolov, B. (2007). Design and control of competence-cell-based production networks. // Advanced Engineering. 1, 1, str. 15-26.

Jaehne M., Li M., Riedel R., Müller E. (2009). Configuring and operating global production networks. International Journal of Production Research, 47, 08, pp. 2013-2030

Jähn H., Zimmermann M., Fischer M., Käschel J. (2006). Performance evaluation as an influence factor for the determination of profit shares of competence cells in non-hierarchical region al production networks. Robotics and Computer-Integrated Manufacturing 22, pp. 526- 535

Jeroen W., J. Gijsen, N.B. Szirbik, G. Wagner (2002). Agent Technologies for Virtual Enterprises in the One-of-a-Kind-Production Industry, International Journal of Electronic Commerce 7 (1) 9-34.

Jeschke, S., Brecher, C., Meisen, T., Özdemir, D., Eschert T. (2017). Industrial Internet of Things and Cyber Manufacturing Systems. Industrial Internet of Things. Springer, Cham, 2017.

Jiang, P., Ding, K., Leng, J. (2016). Towards a cyber-physical-social-connected and service-oriented manufacturing paradigm: Social Manufacturing, Manufacturing Letters 7, 15-21.

Jung H. (2011). A fuzzy AHP–GP approach for integrated production-planning considering manufacturing partners. Expert Systems with Applications 38, pp. 5833-5840

Kagermann H., W. Wahlster, J. Helbig (2013). Recommendations for implementing the strategic initiative INDUSTRIE 4.0, Heilmeyer und Sernau, 2013.

Kampker A., Schuh G., Schittny B., Kupke D. (2010). An Approach for Systematic Production Network Configuration. 43rd CIRP Conference on Manufacturing Systems - Conference Proceedings, Vienna, Austria

Kennedy, J., Eberhart, R.C. (1995). Particle swarm optimization. Proceedings of IEEE International Conference on Neural Networks, 1942-1948.

Kirkpatrick, S., Gelatt, C.D., Vecchi, M.P. (1983). Optimization by simulated annealing, Science, 220, 671-680.

Kolberg D., Knobloch J., Zühlke D. (2017). Towards a lean automation interface for workstations. International Journal of Production Research 55, 2845-2856.

Koren Y. (2010). The Global Manufacturing Revolution: Product-Process-Business Integration and Reconfigurable Systems, ISBN 0470583770, John Wiley & Sons, New York, USA

Koren Y. (2013). The rapid responsiveness of RMS, International Journal of Production Research 51 (23-24) 6817-6827.

Koren, Y., F. Jovane, T. Moriwaki, G. Pritschow, G. Ulsoy and H. Van Brussel. (1999). Reconfigurable Manufacturing Systems. Annals of the CIRP 48 (2). 527-540.

KPMG Special Services., EIM Business & Policy Research in the Netherlands., European Network for SME Research (ENSR)., Intomart. (2003). Observatory of European SMEs. ISBN 92-894-5978-6, Belgium

Kuo T.C., C.W. Hsu, K.C. Ku, P.S. Chen, C.H. Lin (2012). A collaborative model for controlling the green supply network in the motorcycle industry, Advanced Engineering Informatics 26 (4). 941-950.

Lanza G., Book J. (2011). Modeling and Simulation of Value Added Networks under Consideration of Individual Target Systems using Software Agents. 21st International Conference on Production Research - Conference Proceedings, Stuttgart, Germany

Lanza G., J. Ude (2010). Multidimensional evaluation of value added networks, CIRP Annals - Manufacturing Technology 59 489-492.

Lau A., Fischer Th. (2011). Cross-sectoral innovation networks for knowledge-intensive products and services. 21st International Conference on Production Research - Conference Proceedings, Stuttgart, Germany

Leigh Reid R., Rogers K. J., Johnson M. E., Liles D. H. (1996). Engineering the virtual enterprise. Automation & Robotics Research Institute – Conference '96, pp. 485-490

Lima A.C.S., Naveiro R.M. (2011). Innovation trends in the Brazilian foundry industry: a survey with small and medium size companies. 21st International Conference on Production Research - Conference Proceedings, Stuttgart, Germany

Liu M., F. Zhang, Y. Ma, H.R. Pota. W. Shen (2016). Evacuation path optimization based on quantum ant colony algorithm, Advanced Engineering Informatics 30 (3). 259-267.

Luque, M., Caballero, R., Molina, J., Ruiz, F. (2007). Equivalent information for multiobjective interactive procedures. Management Science 53, 125-134.

Ma J., Fang Y. (2012). Stability decision models of partner selection problem for supply chain under discrete demand parameter. Advances in Information Sciences and Service Sciences 4, pp. 685-693

Macunovich D.J. (2012). The role of demographics in precipitating economic downturns. Journal of Population Economics 25 (3). 783-807.

Manzini R., Bortolini M., Accorsi R., Montecchi M. (2011). Integrated models and tools for planning logistic networks. 21st International Conference on Production Research - Conference Proceedings, Stuttgart, Germany

Markaki, O., D. Panopoulos, P. Kokkinakos, S. Koussouris and D. Askounis. (2013). Towards adapting Dynamic Manufacturing Networks for Future Manufacturing: Benefits and Risks of the IMAGINE DMN end-to-end Management Methodology. Proceeding of WETICE '13: 305-310.

Marx W. (1987). Die Phänomenologie Edmund Husserls. München, 1987.

Matt D.T. (2007). Reducing the structural complexity of growing organizational systems by means of axiomatic designed networks of core competence cells. Journal of Manufacturing Systems 26, pp. 178-187

Meier M., J. Seidelmann, I. Mezgár (2010). ManuCloud: the next-generation manufacturing as a service environment, European Research Consortium for Informatics and Mathematics News 83.

Micán C., Rubiano O., Orejuela J., Mosquera J., Manyoma P. (2011). Empirical study of manufacturing strategy in colombian SME´s. 21st International Conference on Production Research - Conference Proceedings, Stuttgart, Germany

Mladineo M., Veza I., Gjeldum N., Crnjac M., Aljinovic A., Basic A. (2019). Integration and testing of the RFID-enabled Smart Factory concept within the Learning Factory. Procedia Manufacturing, 31, 384-389.

Mladineo, M., Celar, S., Celent, L., Crnjac, M. (2018). Selecting manufacturing partners in push and pull-type smart collaborative networks. Advanced engineering informatics, 38, 291-305.

Mladineo, M., Veza, I., Gjeldum, N., Crnjac, M. (2018). Web Information System for Sustainability Optimization of Production Networks. ICIL 2018 - Conference Proceedings, Beer-Sheva, Israel.

Mladineo M., I. Veza, N. Gjeldum (2017). Solving Partner Selection Problem in Cyber-Physical Production Networks using the HUMANT algorithm, International Journal of Production Research 55 (9) 2506-2521.

Mladineo, M., Veza, I., Gjeldum, N. (2015). Single-Objective and Multi-Objective Optimization using the HUMANT algorithm, Croatian Operational Research Review 6, 459-473.

Mladineo, M. (2014). Phenomenological approach to the design of production networks. Ph.D. Thesis, University of Split, Split.

Mladineo M., I. Veza (2013). Ranking Enterprises in Terms of Competences inside Regional Production Network, Croatian Operational Research Review 4 65-75.

Mladineo M., I. Veza, A. Corkalo (2011). Optimization of the selection of competence cells in regional production networks, Technical Gazette 18 581-688.

Mladineo, M., S. Takakuwa, N. Gjeldum and I. Veza. (2011). Criteria for selection of cooperators in a regional production network. Proceedings of 13th International Scientific Conference on Production Engineering CIM 2011: 153-158.

Molina. J., Laguna, L., Marti, R., Caballero, R. (2007) SSPMO: A Scatter Tabu Search Procedure for Non-Linear Multiobjective Optimization. INFORMS Journal on Computing, 19, 91-100.

Monostori L. (2014). Cyber-physical production systems: Roots, expectations and R&D challenges, Procedia CIRP 17 9-13.

Monostori, L. (2015). Cyber-physical production systems and logistics systems: Roots, expectations, R&D challenges and results. 37th AIM Conference, Vienna, Austria.

Morosini Frazzona E., J. Hartmannb, T. Makuschewitzb, B. Scholz-Reiter (2013). Towards Socio-Cyber-Physical Systems in Production Networks, Procedia CIRP 7 49-54.

Mourtzis D. (2010). Internet Based Collaboration in the Manufacturing Supply Chain. 43rd CIRP Conference on Manufacturing Systems - Conference Proceedings, Vienna, Austria

Mourtzis D., Doukas M., Psarommatis F. (2012). A multi-criteria evaluation of centralized and decentralized production networks in a highly customer-driven environment. CIRP Annals - Manufacturing Technology 61, pp. 427-430

Müller E., Horbach S., Ackerman J., Schütze J., Baum H. (2006). Production system planning in Competence-Cell-based Networks. International Journal of Production Research 44, 18,19, pp. 3989-4009

Müller E. (2006). Production planning and operation in competence-cell-based networks, Production Planning & Control 17 (2) 99-112.

Müller E., S. Horbach, J. Ackermann (2008). Integrative planning and design of logistics structures and production plants in Competence-cell-based networks, Int. J. Services Operations and Informatics 3 (1). 40-52.

Nayak N., Prasanna K., Datta S., Mahapatra S.S., Sahu S. (2010). A novel swarm optimization technique for partner selection in virtual enterprise. 2010 IEEE International Conference on Industrial Engineering and Engineering Management (IEEM). Singapore, Singapore

Nemery P., C. Lamboray (2008). FlowSort: A flow-based sorting method with limiting or central profiles, TOP 16 (1) 90-11.

Netland T.H., Powell D.J. (2017). A Lean World / The Routledge Companion to Lean Management. Routledge, USA, 2017.

Netland T.H. (2014). Coordinating Production Improvement in International Production Networks: What’s new? International Operations Networks, Springer, 2014.

Neuberta R., Gorlitza O., Teich T. (2004). Automated negotiations of supply contracts for flexible production networks. Int. J. Production Economics 89, pp. 175-187

Olson, D.L. (1996). Decision Aids for Selection Problems. New York: Springer.

Pang L.Y., R.Y. Zhong, J. Fang, G.Q. Huang (2015). Data-source interoperability service for heterogeneous information integration in ubiquitous enterprises, Advanced Engineering Informatics 29 (3) 549-561.

Phelps, S., Koksalan M. (2003). An interactive evolutionary metaheuristic for multiobjective combinatorial optimization. Management Science, 49, 1726-1738.

Pinto Leitão P.J. (2004). An Agile and Adaptive Holonic Architecture for Manufacturing Control. Doktorska disertacija, Polytechnic Institute of Braganca, Portugal

Ratzinger J. (Pope Benedict XVI) (2007). Jesus of Nazareth. Canada, Doubleday Religion, 2007.

Roßgoderer U., G. Piepenbrock (2014). Simulationsgestütztes Value Stream Mapping - Symbiose zwischen IE und Industrie 4.0, Industrial Engineering – Fachtagung in Dortmund, 2014.

Roth S., M. Meyer, M. Moldaschl, R. Lang (2005). How Many Networks Are We To Manage? International Conference on Economics and Management of Networks, EMNet 2005.

Rother, M. (2009). Toyota Kata. McGraw-Hill Professional, USA, 2009.

Roy, B. (1968). Classement et choix en présence de points de vue multiples (la méthode ELECTRE). La Revue d'Informatique et de Recherche Opérationelle (RIRO). 8, 57-75.

Saaty, T.L. (1980). The Analytic Hierarchy Process: planning, priority setting, resource allocation. New York: McGraw Hill.

Schermerhorn J.R., Hunt J.G., Osborn R.N. (2002). Organizational Behavior. ISBN 0471420638, USA

Schlick, C. et al. (2017). Cognition-Enhanced, Self-optimizing Production Networks. Integrative Production Technology. Springer, Cham, 2017.

Schuh G., Monostori L., Csáji B.Cs., Döring S. (2008). Complexity-based modeling of reconfigurable collaborations in production industry. CIRP Annals - Manufacturing Technology 57, pp. 445-450

Shen W., Q. Hao, H.J. Yoon, D.H. Norrie (2006). Applications of agent-based systems in intelligent manufacturing: An updated review, Advanced Engineering Informatics 20 415-431.

Smith A. (2018). The Wealth of Nations. CreateSpace Independent Publishing Platform, USA, 2018.

Sörensen, K., Springael, J. (2014). Progressive multi-objective optimization. International Journal of Information Technology and Decision Making, 13, 917-936.

Stein E. (1989). On the problem of empathy, ICS Publications, Washington, 1989.

Sturgeon T.J. (2002). Modular Production Networks: A New American model of Industrial Organization. MIT Working Paper IPC-02-003, USA

Stützle, T., Hoos, H.H. (2000). MAX-MIN Ant System. Future Generation Computer Systems, 16, 889-914.

Tadic, I., Marasovic, B. (2013). Application of multicriteria decision making through financial, human resources and business process aspect in verification of companies' success. Croatian Operational Research Review (CRORR). 4, 270-282.

Talbi, E.G. (2009). Metaheuristics – From Design to Implementation. New York: John Wiley & Sons.

Tao F., Y. Cheng, L.D. Xu, L. Zhang, B.H. Li, CCIoT-CMfg: Cloud Computing and Internet of Things-Based Cloud Manufacturing Service System, IEEE Transactions on Industrial Informatics 10 (2) (2014). 1435-1442.

Tao F., Zhang L, Zhang Z.H., Nee A.Y.C. (2010). A quantum multi-agent evolutionary algorithm for selection of partners in a virtual enterprise. CIRP Annals - Manufacturing Technology 59, pp. 485-488

Uhlemann T.H.J., Schock C., Lehmann C., Freiberger S., Steinhilper R. (2017). The Digital Twin: Demonstrating the Potential of Real Time Data Acquisition in Production Systems. Procedia Manufacturing, 9, 113-120

Váncza J., Monostori L., Lutters D., Kumara S.R., Tseng M., Valckenaers P., Van Brussel H. (2011). Cooperative and responsive manufacturing enterprises. CIRP Annals - Manufacturing Technology 60, pp. 797-820

Veza, I., Mladineo, M., Gjeldum, N. (2013). Production Networks and Partner Selection Problem, DAAAM International Scientific Book 2013. Vienna, Austria: DAAAM International Vienna, 2013. pp. 511-544

Veza I., M. Mladineo (2013). Sustainability through Production Networks, Management and Production Engineering Review 4 33-39.

Villa A. (1998). Organizing a network of enterprises: an object-oriented design methodology. Computer lntegrated Manufacturing Systems Vol. 11, No. 4, pp. 331-336

Wang T., S. Guo, B.R. Sarker, Y. Li (2012). Process planning for collaborative product development with CD-DSM in optoelectronic enterprises, Advanced Engineering Informatics 26 (2). 280-291.

Wang Z.J., Xu X.F., Zhan D.C, (2009) Genetic algorithm for collaboration cost optimization-oriented partner selection in virtual enterprises. Int. J. Prod. Res. 47, pp. 859-881

Wojtyla K. (Pope John Paul II) (2005). Memory and Identity: Conversations at the Dawn of a Millennium. Random House Incorporated, USA, 2005.

Womack J.P., Jones D.T. (1996). Lean Thinking: Banish Waste and Create Wealth in your Corporation. Simon and Schuster, USA, 1996.

Wu N., Mao N., Qian Y. (1999). Approach to partner selection in agile manufacturing. Journal of Intelligent Manufacturing 10, pp. 519-529

Wu N., Su P. (2005). Selection of partners in virtual enterprise paradigm. Robotics and Computer-Integrated Manufacturing 21, pp. 119-131

Yamawaki H. (2002). The Evolution and Structure of Industrial Clusters in Japan. Small Business Economics 18, pp. 121-140

Yang, X. S., Deb, S. (2010). Engineering optimization by cuckoo search, Int. J. Math. Modelling Num. Optimisation, 1, 330-343.

Yang, X.S. (2009). Firefly algorithms for multimodal optimization. 5th Symposium on Stochastic Algorithms, Foundation and ApFplications, LNCS, 5792, 169-178.

Yang, X.S. (2011). Metaheuristic Optimization. Scholarpedia, 6(8).11472.

Yu C.X., T.N. Wong (2011). A TOPSIS-based pre-selection method supporting multiple products partner selection in a virtual enterprise, 21st International Conference on Production Research, 2011.

Zhang Y., F. Tao, Y. Laili, B. Hou, L. Lv, L. Zhang (2012). Green partner selection in virtual enterprise based on Pareto genetic algorithms, The International Journal of Advanced Manufacturing Technology 66 1-17.

Zhao F., Hong Y., Yu D. (2006). A multi-objective optimization model of the partner selection problem in a virtual enterprise and its solution with genetic algorithms. Int. J. Adv. Manuf. Technol. 28, 1246-1253

Zhong R.Y., G.Q. Huang, S. Lan, Q.Y. Dai, T. Zhang, C. Xu (2015). A two-level advanced production planning and scheduling model for RFID-enabled ubiquitous manufacturing, Advanced Engineering Informatics 29 (4). 799-812.

Zitzler, E., Laumanns, M., Thiele, L. (2001). SPEA2: Improving the Strength Pareto Evolutionary Algorithm for Multiobjective Optimization. Proceedings of the EUROGEN'2001, Athens.

Zühlke, D. (2010). SmartFactory - Towards a Factory-of-Things. IFAC annual Reviews in control, 129- 138.

World Bank (2019). Fertility rate, total. https://data.worldbank.org/indicator/SP.DYN.TFRT.IN, 01-07-2019

ABOUT THE AUTHOR

Abbildung in dieser Leseprobe nicht enthalten

Asst. Prof. Marko Mladineo, Ph. D.

He was born on 7th March 1984 in Croatia, did his undergraduate, graduate and postgraduate studies at Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture, University of Split in 2003-2008 and 2008-2014. He is an Assistant Professor at Chair of Industrial Engineering, Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture, University of Split, Croatia.

Research interests: production systems and production networks, information systems and cyber-physical systems, geographic information systems, smart factories, multi-objective optimization and decision-making, decision support systems, project management, lean management, and business process management.

Scientific projects: Development of integrative procedure for management of production and service improvement process – DEPROCIM (University through Knowledge Fund and Croatian Science Foundation 2017-19). Innovative Smart Enterprise – INSENT (Croatian Science Foundation 2014-18), Know-how Exchange on the Consequences and Challenges of the Integration of Key Enabling Technologies in European Manufacturing for the Danube Region – DanKETwork (Fraunhofer ISI 2014-16), Network of Innovative Learning Factories – NIL (Deutscher Akademischer Austauschdienst 2013-16), Logistics personnel excellence by continuous self-assessment – LOPEC (EU Leonardo da Vinci 2012-14), and Technological and organizational optimization of competence cell (Croatian Ministry for Science, Education and Sport 2008-12).

Author/co-author of 60 scientific texts: 23 scientific journal papers, 32 scientific conference papers, 2 book chapters, and 3 scientific books.

He is married with seven children.

Contact information

University of Split

Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture (FESB)

Research group for Industrial Engineering / Chair of Industrial Engineering

T: +385 21 305 939

E: marko.mladineo@fesb.hr

W: www.fesb.unist.hr

[...]

Ende der Leseprobe aus 180 Seiten

Details

Titel
Production Networks meet Industry 4.0
Hochschule
Sveučilište u Splitu
Autor
Jahr
2020
Seiten
180
Katalognummer
V550025
ISBN (eBook)
9783346183521
ISBN (Buch)
9783346183538
Sprache
Englisch
Schlagworte
production network, Industry 4.0, virtual enterprise
Arbeit zitieren
Marko Mladineo (Autor:in), 2020, Production Networks meet Industry 4.0, München, GRIN Verlag, https://www.grin.com/document/550025

Kommentare

  • Noch keine Kommentare.
Blick ins Buch
Titel: Production Networks meet Industry 4.0



Ihre Arbeit hochladen

Ihre Hausarbeit / Abschlussarbeit:

- Publikation als eBook und Buch
- Hohes Honorar auf die Verkäufe
- Für Sie komplett kostenlos – mit ISBN
- Es dauert nur 5 Minuten
- Jede Arbeit findet Leser

Kostenlos Autor werden