Abstract
Many design problems incorporate computer simulation as design tool and are required to meet several performance criteria. Resin Transfer Molding (RTM) process is one of the examples. In this research, we use Goal Programming (GP) to formulate the RTM problem, use Latin Hypercube Design (LHD) to sample the training data from computer simulation model, apply Kriging to construct an approximation model to replace the simulation model, and develop a new algorithm to identify the good design setting in a reasonable time for the RTM process. The values of gate location, vent location, pressure, and viscosity are determined to minimize the fill time, maximize the quality, and minimize the deviations from budget plan and viscosity setting. The performance of the setting identified by the approximation model is verified by simulation results.In our algorithm, we use the gradients estimated by Kriging model to improve Particle Swam Optimization (PSO) algorithm. We call it PSO-Gradient algorithms. The results show that this approach can effectively identify a good setting for RTM process and the PSO-Gradient algorithms outperforms traditional PSO.