A battery of approaches has been applied by researchers and practitioners in the field of inventory optimisation to find optimal inventory policies that can drive the success of businesses of various industries. One such approach is based on the use of genetic algorithms, a multi-purpose subclass of evolutionary algorithms that imitate the prin- ciples of evolution to solve combinatorial problems. In this thesis, we extensively explore the theoretical background of inventory optimisation as well as genetic algorithms before we develop a four-stage serial supply chain model and implement a genetic algorithm for base-stock level optimisation.
Table of Contents
1 Introduction
2 Literature Review
2.1 Literature on Serial Supply Chains
2.2 Literature on Genetic Algorithms
3 Description of the Serial Supply Chain Model
3.1 Model Assumptions
3.2 Mathematical Formulation
4 Theory of Genetic Algorithms
4.1 Optimisation Problem and Solution Representation
4.2 Iterative Process
4.3 Caveats and Limitations
4.3.1 Parameter Tuning
4.3.2 Multimodality
4.3.3 Convergence to the Global Optimum
5 Implementing a GA for Base-Stock Level Optimisation
5.1 Supply Chain Model Implementation
5.2 GA Implementation
5.2.1 Parameters
5.2.2 Iterative Steps
6 Empirical Testing
6.1 Parameter Tuning
6.2 Simulation Runs
7 Conclusion
Research Objectives and Themes
This thesis investigates the effectiveness of Genetic Algorithms (GAs) as a metaheuristic approach for optimizing base-stock levels in a four-stage serial supply chain. The research aims to minimize total supply chain costs through a simulation-based approach, comparing various algorithmic configurations and parameter settings against a benchmark random search process.
- Theoretical exploration of inventory optimization and genetic algorithms.
- Development of a four-stage serial supply chain simulation model.
- Implementation and parameter tuning of different GA configurations.
- Empirical evaluation and benchmarking of GA performance in various cost and lead-time scenarios.
Excerpt from the Book
4.2 Iterative Process
When a GA is used to solve an optimisation problem, a solution x is not considered on its own, instead, a population of candidate solutions is considered in each iteration in analogy to the biological scenario where at a given point in time several individuals co-exist. The candidate solutions in the population are referred to as chromosomes and the elements within each chromosome are its genes or the values taken by the variables of a candidate solution (see Reeves, 1997, p.232). Even though there are numerous ways to set up a GA, an iteration of the algorithm generally runs through the following five-step process. Assume we already have an initial population, that is a set of chromosomes with candidate solutions. This initial population is usually chosen at random from the search space (see Reeves and Rowe, 2002, p.29). The sequence of steps is as follows.
Crossover
The crossover operator combines the genes of different chromosomes according to a pre-defined strategy. A commonly used one is the so-called n-point crossover where two (parent) chromosomes are combined by splitting each of them at n points and swapping the fragments to create two offspring chromosomes (see Kramer, 2017, p.12) and (Reeves and Rowe, 2002, p.38). Figure 4.1 demonstrates a one-point crossover of two parent chromosomes.
Summary of Chapters
1 Introduction: Introduces the trade-off between customer service and inventory costs and presents GAs as a promising heuristic for supply chain optimization.
2 Literature Review: Provides an overview of existing research on multi-echelon serial supply chains and the historical development and application of genetic algorithms.
3 Description of the Serial Supply Chain Model: Defines the mathematical framework, assumptions, and constraints for the four-stage serial supply chain used in the simulation.
4 Theory of Genetic Algorithms: Details the theoretical components of GAs, including encoding, iterative steps, and inherent limitations like local optima.
5 Implementing a GA for Base-Stock Level Optimisation: Explains the algorithmic implementation of the SC model and the GA, including parameter definitions and iterative configurations.
6 Empirical Testing: Presents the setup and simulation results of the GA against different supply chain scenarios, comparing performance metrics for various configurations.
7 Conclusion: Summarizes the findings, highlighting the success of elitist GA configurations over other methods and discussing potential areas for future research.
Keywords
Genetic Algorithms, Inventory Optimisation, Serial Supply Chain, Base-Stock Policy, Metaheuristics, Simulation, Parameter Tuning, Total Supply Chain Cost, Heuristic Solutions, Multi-echelon Inventory, Supply Chain Management, Elitist Selection, Convergence, Operational Research, Stochastic Demand
Frequently Asked Questions
What is the core focus of this research?
The thesis investigates the application of Genetic Algorithms (GAs) to optimize base-stock levels within a four-stage serial supply chain to minimize total costs.
What are the primary themes explored in the work?
The core themes include inventory management, evolutionary computation (specifically GAs), supply chain modeling, parameter optimization, and empirical performance benchmarking.
What is the main objective of the thesis?
The primary goal is to determine if a GA can consistently find optimal or near-optimal base-stock policies for a serial supply chain under varying cost and lead-time conditions.
Which scientific methodology is utilized?
The author uses a simulation-based optimization approach, where a Genetic Algorithm is employed as a metaheuristic to navigate a large solution space and minimize costs through iterative improvements.
What topics are covered in the main body?
The main body covers the mathematical modeling of the supply chain, the theoretical mechanics of GAs, the specific implementation of the GA, and comprehensive empirical testing and analysis of results.
Which keywords best characterize this work?
Key terms include Genetic Algorithms, Inventory Optimisation, Serial Supply Chain, Metaheuristics, Base-Stock Policy, and Simulation.
Why are GAs used instead of exact solutions for this problem?
Exact solutions are often computationally prohibitive and struggle with the stochastic nature of real-world supply chain variables; GAs offer a more efficient heuristic alternative.
How does the choice of selection operator impact performance?
The results show that elitist selection leads to faster and more reliable convergence, whereas roulette-wheel selection tends to inhibit convergence and fails to yield satisfactory results for this specific model.
Is the GA performance sensitive to parameter tuning?
Yes, the thesis demonstrates that finding the right balance of mutation rate, crossover rate, and population size is essential for performance, with the author performing extensive pilot studies to identify optimal settings.
- Quote paper
- Leopold Pfeiffer (Author), 2020, Learning from Nature. Using Genetic Algorithms for Inventory Optimisation, Munich, GRIN Verlag, https://www.grin.com/document/958671