A modified particle swarm optimization algorithm for dynamic multiresponse optimization based on goal programming approach

While many of the previous applications based on the Taguchi method only focus on single-response optimization in static system, dynamic multiresponse optimization has received only limited attentions. Optimization of dynamic multiresponse aims at finding out a setting combination of input controllable factors that will result in optimal solutions for all response variables at each signal level. However, it is often difficult to find an optimal setting when multiple responses are simultaneously considered because of their contradiction among the requirements. Hence, a new robust design optimization procedure based on response surface methodology is proposed in the article. The polynomial models of system sensitivity and the error variance for each response are firstly fitted, and corresponding individual desirability functions based on their respective characteristic are defined. Then, goal programming approach is used to resolve multiresponse optimization problems. Because the problems are often multiobjective optimization problems and are often with multipeak distribution, multiconstraint and high nonlinearity, traditional gradient algorithms are easy to obtain local optimal solutions. So a modified particle swarm optimization algorithm is proposed to search global optimal solution. The example shows that the proposed approach can obtain more effectively solutions for dynamic multiresponse optimization problems.