Probabilistic Robustness Design Using Support Vector Machine based Metamodel

The probabilistic robustness design is a simulation-based design approach in nature. It is computationally intensive, sometimes even impossible, to perform probabilistic robustness design method on complex time-consuming simulation models. The least squares support vector machine based metamodel is introduced into probabilistic robustness design in order to alleviate the computational burden. Standard particle swarm optimization is employed in two aspects: parameter optimization of the support vector machine metamodel and exploration of the controller parameter space for probabilistic robust solution. An application to a benchmark problem is displayed to demonstrate the feasibility of the proposed method.