基于BP神经网络的汽车内置储物盒注塑成型优化

The Injection Molding Optimization of Built-in Car Storage Box Based on BP Neural Network

以某汽车内置储物盒为研究对象,结合聚合物流变学的基本理论,采用Moldflow软件对汽车内置储物盒注塑成型进行模流分析。建立多输入-多输出的BP神经网络拓扑模型,以模流分析得到的充填时间、体积收缩率和总翘曲量等作为BP神经网络的训练样本,采用拟牛顿算法对网络模型进行训练。通过预测值和模拟值的对比,发现训练后BP神经网络模型的相对误差较小,并具有很好的预测能力。以此网络模型对其它各工艺参数组合进行预测,得到了较优的工艺参数匹配关系。并进行试模加工,得到了合格的汽车内置储物盒制件。

Taking a built-in car storage box as the research target and combined with the basic theories of polymer rheology, the injection molding of the head cannula was subjected to flow analysis by using Moldflow software. The BP neural network topology model of multi-input and multi-output was established, the data about the filling time, the volume shrinkage rate and the total warpage amount which obtained in the mold flow analysis were used as the training samples for BP neural network, and quasi-Newton algorithm was used to train the network model. By comparing the predicted values with the simulated values, it was found that the BP neural network model had small relative error and good predictive ability. The other combinations of process parameters were predicted by this network model, and the matching parameters of the process parameters were obtained. Through the trial processing, the qualified car built-in storage box parts were obtained.