Determining optimal parameters for stereolithography processes via genetic algorithm

Abstract

Current part build accuracy of stereolithography processes needs to be improved because part inaccuracy and distortion still limit the processes' application to other areas. This paper focuses on increasing build accuracy by optimally designing the process parameters. The process is modeled and described by a multilayer perceptron neural network. Based on this modeled process, the genetic algorithm searches the optimal process parameters so that optimal conditions yield minimum part build error. In practice, genetic algorithms find near-optimal conditions since they do not guarantee true optimal condition. The nearly optimized process is validated by actually building H-parts and comparing these results with those obtained by the currently used nominal condition.