基于径向基函数神经网络和多岛遗传算法的注射成型质量控制与预测

Quality Control and Prediction of Injection Molding Based on RBF Neural Network and MIGA

针对现有多参数、多目标注塑工艺优化应用的遗传算法、粒子群算法等寻优算法存在的实施难度大、求解时间长等缺点,提出基于最优拉丁超立方试验设计方法并结合径向基函数(RBF)神经网络模型和多岛遗传算法(MIGA)对注射成型质量进行控制与预测。以充电宝下盖塑件的体积收缩率和缩痕指数为优化控制目标,以模具温度、熔体温度、保压时间、保压压力、冷却时间为试验因素,应用最优拉丁超立方试验设计方法获得试验样本,基于模流分析获得试验结果,构建试验因素与优化控制目标之间的RBF神经网络模型,基于MIGA在试验因素给定的水平范围内获得了一组最优注塑工艺参数组合并给出了优化控制目标的预测值。模拟试验结果证明,预测值与模拟试验结果基本吻合,提出的方法能实现注塑成型质量的控制与预测,减少了寻找最优工艺参数组合的时间,提高了塑件的生产效率。

Existing genetic algorithms and particle swarm optimization algorithms used in the optimization of multi-parameter and multi-objective injection molding processes have the disadvantages of difficult implementation and long solution time.The method of quality control and prediction for injection molding based on the optimal Latin hypercube test design method combined with radical basis function(RBF)neural network model and multi-island genetic algorithm(MIGA)was proposed.Taking volume shrinkage rate and shrinkage index of the lower cover of charger as optimization targets,and using the mold temperature,melt temperature,holding time,holding pressure and cooling time as test factors.The optimal Latin hypercube test design method was used to obtain the test samples,the test results were obtained based on the mold flow analysis,and the RBF neural network model between the test factors and the optimization control target was constructed.Based on the MIGA,a set of optimal injection process parameter combinations was obtained within a given range of experimental factors and the predicted values of the optimization goals were given.The simulation test results show that the predicted values are basically consistent with the simulation test results.The method proposed can control and predict the quality of injection molding,shorten the time to find the optimal combination of process parameters,and improve the production efficiency of plastic parts.