In this book, we propose several modules of diagnosis for complex and dynamic systems. These modules are based on the three algorithms colony of ants, which are AntTreeStoch, Lumer & Faieta and Binary ant colony. These algorithms have been chosen for their simplicity and their vast field of application. However, these algorithms cannot be used under their basal form for the development of diagnostic modules since they have several limitations.
We have also proposed several adaptations in order that these algorithms can be used in diagnostic modules. We have proposed a parallel version of the algorithm AntTreeStoch based on a reactive multi-agents system. This version allows minimizing the influence of initial sort on the outcome of classification. We have also introduced a new parameter called Sid, which allows several ants to connect to the same position, and we have modified the movements of ants by promoting the path of the ant the most similar.
For the algorithm Lumer & Faieta, we have accelerated the speed of construction of classes by adding a speed setting different for each Ant. To reduce the number of movements, we have proposed a new variable that allows saving the identifiers of objects displaced by the same Ant. To improve the quality of classification, we have also added to the algorithm of the indices to report the classes trunks constructed. For the algorithm Binary ant colony, we have proposed a variant called "Hybrid wrapper/filter-based ACO-SVM".
This algorithm allows the selection of parameters. It combines the techniques of filters and enveloping methods in taking advantage of the rapidity of the Fisher report and the adaptation of selected settings to the classifier SVM. It improves the quality of classification according to the data nature in the database for learning and the type of the kernel function used. It also allows adjusting the hyperparameters of the kernel function. We tested these algorithms based on data from two industrial systems, which are the sintering system and the pasteurization system, as well on a few databases of UCI (University of California, Irvine).
Inhaltsverzeichnis
General introduction
1 CONTEXT
2 ORGANIZATION OF THE MANUSCRIPT
3 CONTRIBUTIONS OF THE BOOK
Chapter 1: Fault diagnosis by pattern recognition
1 INTRODUCTION:
2 DEFINITIONS
3 THE DIFFERENT APPROACHES TO THE DIAGNOSIS
4 FAULT FINDING BY PATTERN RECOGNITION
4.1 The basic principle
4.2 Analysis of Observations:
4.3 Reduction of the dimension of the space of representation
4.4 The space of decision:
4.4.1 The hierarchical classification:
4.4.2 The classification by partition:
4.5 The procedure of decision:
4.5.1 Parametric methods:
4.5.2 Non-parametric methods:
4.5.3 Direct calculation of borders:
4.6 Operating Phase
4.6.1 The evaluation criteria
4.6.2 The methods of evaluation
5 ARTIFICIAL INTELLIGENCE TOOLS FOR THE INDUSTRIAL DIAGNOSIS
5.1 Networks of Neurons
5.2 The fuzzy logic
6 CONCLUSION:
Chapter 2: Ant Colony Optimization
1 INTRODUCTION:
2 THE ACTUAL ANTS
2.1 Basic Principle
2.1.1 The resolution of complex problems
2.1.2 Stigmergy
3 THE ARTIFICIAL ANTS
4 MULTI-AGENT SYSTEMS
4.1 The principles of databases
4.2 The agents
4.3 Architecture of agents
4.4 The Reactive architectures:
5 THE COLLECTIVE INTELLIGENCE OF ANTS
5.1 The sharing of tasks
5.2 Self-organization
5.3 The communication
5.3.1 The sound communication
5.3.2 The tactile communication
5.3.3 The visual communication
5.3.4 The chemical communication
6 THE DIFFERENT AREAS OF APPLICATION
7 ANT COLONY ALGORITHMS AND THE ADM
7.1 The algorithm AntTreeStoch
7.1.1 The broad lines
7.1.2 The main algorithm of AntTreeStoch
7.1.3 The case of an Ant on the bracket
7.1.4 The case of an ant placed on another
7.1.5 Criticism and limits of AntTreeStoch
7.1.6 The improvements
7.2 The algorithm of Lumer & Faieta
7.2.1 The broad lines
7.2.2 The distance
7.2.3 The classification
7.2.4 Criticism and limits of the basal algorithm
7.2.5 Improvement of the algorithm
7.3 The BINARY algorithm de colony of ANTS
7.3.1 The broad lines
7.3.2 Criticism and limits of the basal algorithm
7.3.3 Improvement of the algorithm
8 CONCLUSION
Chapter 3 : Support Vector Machine (SVM)
1 INTRODUCTION
2 HISTORY
3 THE SVM
3.1 The optimal hyperplane
4 SVM: PRIMAL FORMULATION
5 SVM: DUAL FORMULATION
6 THE CASE OF A SAMPLE OF NON-LINEARLY SEPARABLE
7 THE FUNCTIONS THE NUCLEI
8 HYPER-PARAMETERS
9 SVM MULTI-CLASSES
9.1 SVM: A against all
9.2 SVM: one against
9.3 DAG-SVM
10 THE REGRESSION BY SVM
11 SOFTWARE TOOLS
11.1 SVMTorch
11.2 LIBSVM
11.3 SVMLight
12 CONCLUSION
Chapter 4: The selection parameters for industrial diagnosis
1 INTRODUCTION
2 THE STEPS FOR THE SELECTION OF PARAMETERS
2.1 The evaluation criteria
2.1.1 The measures of error of classification
2.1.2 The information measures
2.1.3 The measures of consistency
2.1.4 The measures of dependence
2.1.5 Distance measurements
2.2 The methods the filters
2.3 The enveloping Methods
2.4 The integrated methods:
3 THE PROCEDURE OF GENERATION
3.1 The complete generation
3.2 The random generation
3.3 The Sequential generation
3.3.1 The method of constructive generation
3.3.2 The method of destructive generation
4 STOP CRITERION
5 THE VALIDATION
6 EXTRACTION OF THE PARAMETERS
6.1 Principle
6.2 The principal components analysis
7 SELECTION PROCEDURE PROPOSED
8 CONCLUSION
Chapter 5: Application and validation of the proposed approaches
1 INTRODUCTION
2 DESCRIPTION OF THE FIRST PROCESS
2.1 Presentation of SCIMAT
3 THE MANUFACTURE OF CEMENT
3.1 The Sintering
3.1.1 The rotary kiln
3.2 Analysis of operation of a Rotary Kiln
3.3 The parameters of the part Sintering
3.3.1 Modes of Operation studied
4 THE DESCRIPTION OF THE SECOND PROCESS
4.1 Presentation of dairy in the Aurès
4.2 Production of milk
4.3 Pasteurization
4.3.1 Principle
4.3.2 The pasteurizer
4.3.3 The main parameters of the process of pasteurization
4.3.4 Modes of Operations studied
5 THE ALGORITHM ANTTREESTOCH
5.1 Use of Netlogo
5.2 The test data
5.3 The parameters of AntTreeStoch
5.4 The evaluation measures
5.5 The Results
6 APPLICATION OF THE ALGORITHM LUMER & FAIETA
6.1 Configuration of the algorithm
6.2 Results of scenarios :
7 APPLICATION OF THE ALGORITHM HYBRID WRAPPER/FILTER-BASED ACO-SVM
7.1 The test data
7.2 The Heuristic factor FH
7.3 The configuration of the classifier
7.3.1 Adjusting the hyper-parameters
7.3.2 The results of classification
7.4 The Results
8 CONCLUSION
Zielsetzung und thematische Schwerpunkte
Das Buch widmet sich der Entwicklung von Diagnosemodulen für komplexe und dynamische Industriesysteme unter Verwendung von Algorithmen, die auf Ameisenkolonien basieren. Das primäre Ziel besteht darin, bestehende Ansätze (wie AntTreeStoch, Lumer & Faieta und Binary ant colony) so anzupassen und zu verbessern, dass sie effizienter bei der Fehlererkennung und Parameterselektion eingesetzt werden können, um eine bessere Interpretierbarkeit und Leistungsfähigkeit in industriellen Umgebungen zu erreichen.
- Anpassung und Optimierung von Algorithmen der Ameisenkolonie-Optimierung (ACO) für Diagnosezwecke.
- Methoden der Parameterselektion und Dimensionsreduktion von Zustandsvektoren.
- Einsatz von Support Vector Machines (SVM) zur Klassifizierung in industriellen Systemen.
- Integration hybrider Ansätze (Wrapper/Filter-basiert) zur Verbesserung der Klassifizierungsqualität.
- Validierung der vorgeschlagenen Ansätze an industriellen Prozessen wie Zementherstellung und Milchpasteurisierung.
Auszug aus dem Buch
4.3.2 The pasteurizer
It often uses a heat exchanger to plates as pasteurizer. It consists of an assembly of the splined plates installed and tightened on a frame. Each pair of plates form a channel which alternates for the fluid and for the product. The distance between two plates obtained by seals is variable. It can go from 3 to 7 mm for the products most viscous. The exchange surface is obtained by the surface of the Plates and by their number. This type of heat exchanger has the advantage of taking up less space on the ground that the other.
The pasteurizer used operates at large flows and it is completely automated. It represents a device that can heat or cool in a continuous pompable product. It is based on the principle to circulate at the same time the product to a thermal fluid. The latter is generally against the current of the product. It includes several sections as well as a Lodger:
• The lodger: is a tube whose volume allows, depending on the flow rate, maintain the pasteurization temperature during the time necessary.
• Section of heat recovery preheating or pre-cooling: in this section, the incoming product is preheated and the outgoing product is pre-cooled. No external energy is necessary, which allows you to save a lot of energy.
• Heating Section: This section should get the product preheated in the recovery section to the pasteurization temperature. This measure of the pasteurization temperature must be precise and constant.
• Cooling section: once pre-cooled in the section of heat recovery, the product is cooled in this section. The final product should be constantly cooled since pasteurization does not sterile.
Zusammenfassung der Kapitel
General introduction: Stellt den Kontext der industriellen Diagnose, die Bedeutung der zustandsbasierten Wartung und die Zielsetzung des Buches hinsichtlich der Nutzung von Mustererkennung durch Ameisenalgorithmen dar.
Chapter 1: Fault diagnosis by pattern recognition: Definiert grundlegende Konzepte der Diagnose von komplexen Systemen und beschreibt Methoden der Mustererkennung sowie Ansätze der künstlichen Intelligenz.
Chapter 2: Ant Colony Optimization: Bietet eine bibliographische Studie zu Ameisenalgorithmen und stellt die drei untersuchten Basis-Algorithmen sowie Ansätze der Multi-Agenten-Systeme vor.
Chapter 3 : Support Vector Machine (SVM): Erläutert die theoretischen Grundlagen von Support Vector Machines, ihre Anwendung bei linearer und nicht-linearer Trennbarkeit sowie Algorithmen für Multiklassen-Probleme.
Chapter 4: The selection parameters for industrial diagnosis: Diskutiert Methoden zur Reduktion der Dimension von Zustandsvektoren durch Parameterselektion und -extraktion und führt den hybriden ACO-SVM-Ansatz ein.
Chapter 5: Application and validation of the proposed approaches: Präsentiert die Implementierung und Validierung der vorgeschlagenen Ansätze an zwei industriellen Prozessen (Zementwerk und Molkerei) anhand realer und simulierter Daten.
Schlüsselwörter
Diagnose, Klassifizierung, Parameterselektion, Ameisenkolonie-Algorithmen, Multi-Agenten-Systeme, Support Vector Machine, industrielle Systeme, Zementherstellung, Pasteurisierung, Mustererkennung, Metaheuristiken, Optimierung, SVM, Fehlererkennung, künstliche Intelligenz.
Häufig gestellte Fragen
Worum geht es in dieser wissenschaftlichen Arbeit grundsätzlich?
Die Arbeit befasst sich mit der Entwicklung von intelligenten Diagnosemodulen für komplexe und dynamische industrielle Produktionssysteme, um Störungen zuverlässig zu erkennen und zu lokalisieren.
Was sind die zentralen Themenfelder der Publikation?
Die Arbeit kombiniert Methoden der künstlichen Intelligenz, insbesondere Algorithmen aus der Ameisenkolonie-Optimierung, mit Techniken der Mustererkennung und Support Vector Machines.
Was ist das primäre Ziel oder die Forschungsfrage?
Das Ziel ist die Adaptation und Optimierung von Ameisenalgorithmen zur effizienten Parameterselektion und zur Verbesserung der Klassifizierungsqualität bei der Überwachung von industriellen Anlagen.
Welche wissenschaftliche Methode wird primär verwendet?
Die Autoren verwenden bio-inspirierte Optimierungsverfahren (ACO-Varianten wie AntTreeStoch, Lumer & Faieta) in Kombination mit überwachten Lernverfahren (SVM) und hybriden Selektionsmethoden.
Was wird im Hauptteil der Arbeit behandelt?
Der Hauptteil gliedert sich in die theoretische Fundierung der Ameisenalgorithmen und SVMs sowie die praktische Anwendung und Validierung dieser Methoden an konkreten industriellen Fallbeispielen.
Welche Schlüsselwörter charakterisieren die Arbeit?
Zu den wichtigsten Begriffen zählen Diagnose, Klassifizierung, Parameterselektion, ACO, Multi-Agenten-Systeme und Support Vector Machines.
Wie werden die vorgeschlagenen Algorithmen validiert?
Die Validierung erfolgt durch praktische Simulationen an zwei spezifischen industriellen Prozessen: der Zementherstellung in einem Drehrohrofen und der Pasteurisierung in einer Molkerei.
Warum wurde der Hybrid Wrapper/Filter-basierte ACO-SVM-Ansatz entwickelt?
Dieser Ansatz wurde entwickelt, um die Geschwindigkeit von Filtermethoden mit der hohen Klassifizierungspräzision von SVM-Methoden zu vereinen und so eine optimale Auswahl relevanter Parameter zu ermöglichen.
- Citar trabajo
- Kadri Ouahab (Autor), Adel Abdelhadi (Autor), 2016, Ant Colony Algorithm in Fault Diagnosis, Múnich, GRIN Verlag, https://www.grin.com/document/344694
