1 / 31

An Adaptive GA for Multi Objective Flexible Manufacturing Systems

An Adaptive GA for Multi Objective Flexible Manufacturing Systems. A. Younes, H. Ghenniwa, S. Areibi ayounes@uoguelph.ca, hgenniwa@uwo.ca sareibi @ uoguelph.ca July 2002. GECCO 2002, New York. Outline. Introduction Background (Flexible Manufacturing Systems) Motivation/Contributions

freira
Télécharger la présentation

An Adaptive GA for Multi Objective Flexible Manufacturing Systems

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. An Adaptive GA for Multi Objective Flexible Manufacturing Systems A. Younes, H. Ghenniwa, S. Areibi ayounes@uoguelph.ca, hgenniwa@uwo.ca sareibi@uoguelph.ca July 2002 GECCO 2002, New York

  2. Outline • Introduction • Background (Flexible Manufacturing Systems) • Motivation/Contributions • Example • Mathematical Formulation • Genetic Algorithm Implementation • Numerical Testing and Comparison. • Conclusions & Future Work. GECCO 2002, New York

  3. Introduction • Flexible Manufacturing Systems consist of multiple heterogenous machines (robots/computers). • Ultimate goal: is to maximize the FMS throughput. • Several problems such as part type partitioning, assignment and sequencing must be solved before this goal can be achieved. • This work is an initial investigation for the suitability of Genetic Algorithms to solve Dynamic Optimization problems associated with scheduling/sequencing in FMS systems. GECCO 2002, New York

  4. Background • The goodness of an assignment is measured in terms of minimizing part transfer (primary) and balancing the work-load of the machines (secondary). • The aim is to facilitate the creation of machine cells with minimum part transfer while maximizing the utilization of machines. • While minimizing part transfer tends to favor the assignment of the whole of a part to a single machine, balancing work-load tries to make the work-load distribution even among the machines. GECCO 2002, New York

  5. Motivation • A large number of combinatorial problems are associated with Manufacturing Optimization. • Many short comings from current techniques used for dynamic optimization problems. • This work is used as foundation for future work in the area of dynamic scheduling/sequencing of Flexible Manufacturing Systems. • Techniques developed for such problems can be easily adapted for other type of problems that are dynamic in nature. GECCO 2002, New York

  6. Contribution • One of the main contribution of this work is developing an automated technique to generate benchmarks for Flexible Manufacturing Systems (both Static and Dynamic Benchmarks) • Several crossover techniques have been developed and tested for Flexible Manufacturing Systems. • Not too much work has been done in the literature on solving dynamic optimization problems for Flexible Manufacturing Systems. • This work lays the foundation for Evolutionary dynamic optimization strategy for scheduling/sequencing of FMS. GECCO 2002, New York

  7. Mathematical Formulation • F1: Minimization of part transfer (by minimizing the number of machines required to process the part) • F2: Minimization of the number of necessary operations required from each machine over the possible processing choices. • F3: Load balancing by minimizing the cardinality distance between the workload of any pair of machines. • Over multi-objective mathematical model of FMS is to solve for F1, F2, F3 Subject to a part being processed by a single machine. GECCO 2002, New York

  8. P2 P1 M1 P1 M2 P2 M2 can perform three operations O2 O4 O1 O2 O3 O5 O2 O3 O5 O1 O3 O5 O2 O3 O5 O5 M3 M3 can perform two operations FMS Example Two part types & Three Machines Four operations needed to process P1 M1 can perform Four operations Three operations needed to process P2 This choice tries to minimize part transfer between machines GECCO 2002, New York

  9. P2 P1 M1 P1 M2 P2 O5 O1 O2 O3 O5 O2 O3 O5 O4 O5 O3 O1 O2 O3 O5 O2 M3 FMS Example Two part types & Three Machines This choice tries to distribute workload (operations) evenly between machines GECCO 2002, New York

  10. A GA Algorithm for FMS • Genetic Algorithms are well suited for multiple-objective optimization problems. • The basic feature of GA is multiple directional and global search through maintaining a population of potential solutions from generation to generation. • In our implementation we have combined a Pareto-based approach with an adaptive weighted sum technique for tackling the multi-objective flexible manufacturing systems problem. • One of the main issues is determining the fitness value of individuals according to multiple objectives. GECCO 2002, New York

  11. GA Main Componenets Fitness Function Representation Transformation Functions Selection & Deletion GECCO 2002, New York

  12. P2 P1 P1 P2 O2 O5 O1 O3 O5 O3 O2 O2 O3 O3 O1 O5 O4 O5 O5 O2 Chromosome Representation M1 M2 M3 GECCO 2002, New York

  13. normalized Fitness Function Two objectives are considered normalized The weights determine which of the two objectives is favored GECCO 2002, New York

  14. Crossover Simple Crossover Cut points set at the part delimiter ` GECCO 2002, New York

  15. ` ` ` ` ` ` Crossover Uniform Crossover A Cut point for each gene delimiter GECCO 2002, New York

  16. Crossover Structured Crossover Cut points Randomly set in the string ` GECCO 2002, New York

  17. GA Algorithm • initialize pop. ; • evaluate initial pop. ; • while not stopping condition • { • select fittest parents for reproduction; • apply crossover & mutation; • evaluate pop.; • } GECCO 2002, New York

  18. Benchmarks • Several Benchmarks used to evaluate the performance of the GA for FMS. • Randomly generated with different M/P/O. • The Generator can be used for both Dynamic & Static optimization solvers. GECCO 2002, New York

  19. Benchmarks (Statistics) GECCO 2002, New York

  20. Results & Discussion • The Genetic Algorithm code was developed on a Sun Sparc Ultra 10 Workstation running Solaris 8. • The Code was written in C and compiled using GNU g++ version 2.95.2. • Results obtained were first run using the Genetic Algorithm by optimizing each objective function separately. • The Genetic Algorithm was then run by optimizing both objective functions together. GECCO 2002, New York

  21. GA Convergence

  22. Crossover Operator

  23. Machines Involved GECCO 2002, New York

  24. Cont .. Machines Involved GECCO 2002, New York

  25. Cont .. Machines Involved GECCO 2002, New York

  26. Operations Per Machine GECCO 2002, New York

  27. Operations Per Machine GECCO 2002, New York

  28. Operations Per Machine GECCO 2002, New York

  29. Conclusions • In this paper we introduced an adaptive Genetic Algorithm for Flexible Manufacturing Systems. • A random benchmark generator was developed for both static and dynamic problems. • Results obtained indicate that our Genetic Algorithm implementation achieves excellent results with respect to part transfer and balancing the work among the machines. • Currently we are testing the Genetic Algorithm on a Dynamic version for Flexible Manufacturing where one or more machines may fail during optimization. GECCO 2002, New York

  30. Future Work • Compare this Genetic Algorithm implementation with other advanced search techniques (Tabu Search, GRASP) for both static and dynamic problems. • Incorporate local search with the Genetic Algorithm to create a Memetic Algorithm • Include sequencing constraints and tools costs in the objective function. • Integrating Multi Agent Systems with Genetic Algorithms for complex dynamic optimization approaches. GECCO 2002, New York

  31. Thank You! All Source Code is Available at the following web site: http://www.uoguelph.ca/~sareibi GECCO 2002, New York

More Related