基于灰色关联分析的注塑工艺多目标优化及PSO–SVM预测模型的建立

Multi-objective Optimization of Injection Molding Process based on Grey Relational Analysis and Establishment of PSO–SVM Prediction Model

针对某电器活动上盖翘曲变形及体积收缩问题,对相关注塑工艺参数进行正交实验设计,在Moldflow中模拟分析,并对翘曲变形量及体积收缩率进行信噪比优化处理。利用灰色关联分析法得到翘曲变形量和体积收缩率的灰色关联度,通过对灰色关联度进行极差分析得到各注塑工艺参数对塑件综合目标(翘曲变形量及体积收缩率同时较小)的影响程度为:保压时间>注塑时间>模具温度>熔体温度>保压压力>冷却时间,同时由灰色关联度极差分析结果得出最优工艺参数组合,在最优工艺参数组合下的翘曲变形量相对于正交实验水平下最小翘曲变形量降低了11.8%,体积收缩率相对于正交实验水平下最小体积收缩率降低了5.9%。最后采用粒子群优化算法(PSO)优化后的支持向量机(SVM)神经网络模型对该塑件翘曲变形量及体积收缩率进行预测,通过与不优化的SVM神经网络及BP神经网络预测模型相比发现,PSO–SVM神经网络模型预测精度及稳定性都优于SVM及BP神经网络,可以用于塑件翘曲变形量和体积收缩率的协同优化,解决塑件实际翘曲变形及体积收缩问题。

Aiming at the problems of warping deformation and volume shrinkage of the movable upper cover of an electrical appliance,the orthogonal experimental design was carried out on relevant injection molding process parameters,the experimental data simulated and analyzed via Moldflow,and both of the warpage deformation value and volume shrinkage rate were optimized with signal-to-noise ratio.The gray correlation analysis method was used to obtain the gray correlation degree that between the warpage deformation value and the volume shrinkage rate.At the same time,through the range analysis of the gray correlation degree,the comprehensive target of each injection molding process parameter to the plastic part was obtained(the warpage deformation value and the volume shrinkage rate are simultaneously small),the degree of influence is:holding pressure time>injection time>mould temperature>melt temperature>holding pressure>cooling time.Meanwhile,the optimal process parameter combination was obtained from the gray correlation analysis results.Compared with the minimum warpage deformation value and the minimum volume shrinkage rate under the orthogonal experiment level,the warpage deformation value and volume shrinkage rate under the optimal process parameter combination is reduced by 11.8%and by 5.9%respectively.Finally,the warpage deformation and volume shrinkage rate of the plastic part were predicted through the support vector machine neural network model(SVM)optimized via the particle swarm optimization algorithm(PSO).In comparison with the non-optimized SVM neural network and BP neural network prediction model,the prediction accuracy and stability of the PSO–SVM neural network model is better than those of SVM and BP neural network,indicating that the model can be used for the collaborative optimization of the plastic part’s warpage deformation value and volume shrinkage rate,and solve the actual warpage deformation and volume shrinkage of the plastic part.