基于改进粒子群算法和小波神经网络的高强钢扭曲回弹工艺参数优化*

针对高强钢复杂件冲压后出现的扭曲回弹现象，运用有限元仿真软件DYNAFORM对复杂件的冲压、回弹过程进行数值模拟，提出了评价复杂件扭曲回弹程度的指标，并运用试验设计和小波神经网络代理模型方法对扭曲回弹进行了优化研究。以某弯曲梁为研究对象，以扭曲回弹为成形目标，通过正交试验设计筛选出对扭曲回弹影响较大的工艺参数作为影响因素。利用拉丁超立方对影响因素进行抽样，通过数值模拟获得样本数据，建立影响因素与成形目标之间的小波神经网络代理模型，利用改进的粒子群算法对该模型迭代寻优获得最优参数。结果表明：采用优化后的工艺参数能有效地减小该弯曲梁的扭曲回弹，该方法为减小复杂件的扭曲回弹提供一种有益的指导。

Optimization of Parameters in Twist Springback Process for High-strength Sheets Based on Improved Particle Swarm Optimization Algorithm and Wavelet Neural Network

Aiming at the twist springback appearing after the stamping process of high-strength steel sheet, the stamping and springback processes are numerically simulated based on the finite element analysis software DYNAFORM. Then a new method to evaluate the torsion springback is proposed, and the design of experiments and the wavelet neural network agent model are used to the optimization research of the twist springback. The flex-rail model is taken into account, taking the value of twist springback as forming target, and the process parameters selected through the orthogonal experiment design as influencing factors. The Latin hypercube is used to sample influencing factors, and the simulation is carried out to get the samples. The wavelet neural network agent model of influencing factors and forming target is built. Then the optimal solution is got by the iteration of improved particle swarm optimization algorithm. The results show that the optimized process parameters can effectively reduce the twist springback, and the research provides a useful method to reduce the twist springback of complex parts.