High-Performance Discrete Particle Swarm Optimization (PSO) Algorithm and Software Development of Application on JSSP

Abstract

Conducted research to study industrial Job-shop Scheduling Problem(JSSP), and designed a coding and decoding scheme.
Prototyped and implemented the Cooperative PSO algorithm for optimizing JSSP problem in MATLAB.
Discovered optimal combination of parameters for CPSO algorithm using Response Surface Methodology(RSM).
Designed 8 JSSP testing cases for CPSO algorithm with optimal parameters, and mapped correlated algorithm convergence curve and scheduling Gantt Graphs.
Conducted performance analysis from two aspects.
- Algorithm Improvement: CPSO algorithm reduced at most 17.68% error to theoretical optimal solution compared to that of standard PSO algorithm;
- Parameter Optimization: approximately 87.5% of 8 test cases results made better performances with aforementioned optimal parameters than original parameters did.
Developed a user-friendly desktop client: PSO-related JSSP software system in C# and SQL Server, providing the capability for users to obtain fast results by inputting key parameters.

Algorithm Design

Standard PSO Algorithm

standard PSO algorithm

Cooperative PSO Algorithm

cooperative PSO algorithm

Performance Analysis

Designed 8 JSSP benchmark testing cases(FT06，FT10，FT20，LA01，LA21，LA26，LA31，LA36) for performance analysis:

Algorithm Improvement

Mainly compared the results of JSSP benchmarks solved by PSO algorithm and CPSO algorithm:

makespan

optimization rate

average relative error

standard deviation

Parameter Optimization

Discovered optimal combination of parameters for CPSO algorithm using Response Surface Methodology(RSM), and about 87.5% of test cases had better performances than the original parameters did:

parameter optimization

Software Development

Developed a user-friendly desktop client: PSO-related JSSP software system in C# and SQL Server, providing the capability for users to obtain fast results by inputting key parameters.