基于RBF神经网络和遗传算法的注塑成型质量控制与预测

针对注塑产品容易产生翘曲和缩痕的问题,以某检测仪外壳为研究对象,运用RBF神经网络模型和遗传算法,对注塑成型质量进行控制与预测。基于正交试验方案,运用Moldflow有限元分析软件获得试验结果;利用样本数据建立试验因素与响应值之间的RBF神经网络模型,并用最优拉丁超立方抽样技术,获得样本点对模型精度进行检验;运用带精英策略的非支配排序遗传算法(NSGA-Ⅱ)对注塑成型工艺参数进行多目标优化,达到有效控制和预测翘曲变形、体积收缩率和缩痕指数的目的,并经模拟和试模验证误差较小。结果表明,运用RBF神经网络模型和遗传算法对注塑成型质量进行控制与预测,生产出检测仪外壳最大翘曲变形量为0.394 mm,外观无缩痕。

Injection Molding Quality Control and Prediction Based on RBF Neural Network and Genetic Algorithm

Aiming at the problem that injection products were easy to produce warpage and shrinkage marks,taking the shell of a detector as the research object,injection molding quality was controlled and predicted by RBF neural network model and genetic algorithm.On the basis of orthogonal test scheme,the test results were obtained by Moldflow finite element analysis software.Using the sample data,the RBF neural network model between test factors and response values were established.Using the optimal latin hypercube sampling technique,the sample points were obtained to test the accuracy of the model.Injection molding process parameters were optimized by using NSGA-Ⅱ,to control and predict warpage,volume shrinkage and shrinkage index effectively and it was verified by simulation and test that the error was small.The results show that RBF neural network model and genetic algorithm are used to control and predict the quality of injection molding,the maximum warpage of the detector shell is 0.394 mm,no shrinkage mark on appearance.