Multi-objective optimization of the hydraulic press crossbeam based on neural network and pareto GA

The structures approximation analysis technology is studied based on neural network. The back-propagation neural network model corresponding to the size parameters of the hydraulic press' crossbeam and its displacement or stress is generated to replace the original finite element model in this paper. Using the saturated multi-level table of orthogonal arrays to choose the trained samples could make the neural network has extensive representations. In order to search the minimization of the crossbeam's volume and displacement, the Pareto GA is used and the detailed technique is described. The optimization result is satisfactory, which shows the combination of the neural network and Pareto GA provides a new scientism method on solving the complex solid structures' multi-objective optimization.