Abstract The qualities of injection-molded parts are affected by process parameters. Warpage and volume shrinkage are two typical defects. Moreover, insufficient or excessively large clamping force also affects the quality of parts and the cost of the process. An experiment based on the orthogonal design was conducted to minimize the above defects. Moldflow software was used to simulate the injection process of each experiment. The entropy weight was used to determine the weight of each index, the comprehensive evaluation value was calculated, and multi-objective optimization was transformed into single-objective optimization. A regression model was established by the random forest (RF) algorithm. To further illustrate the reliability and accuracy of the model, back-propagation neural network and kriging models were taken as comparative algorithms. The results showed that the error of RF was the smallest and its performance was the best. Finally, genetic algorithm was used to search for the minimum of the regression model established by RF. The optimal parameters were found to improve the quality of plastic parts and reduce the energy consumption. The plastic parts manufactured by the optimal process parameters showed good quality and met the requirements of production.

中文翻译：

---

使用熵权、随机森林和遗传算法方法对注塑成型塑料零件进行多目标优化

摘要 注塑件的质量受工艺参数的影响。翘曲和体积收缩是两种典型的缺陷。此外，夹紧力不足或过大也会影响零件的质量和工艺成本。进行了基于正交设计的实验以最小化上述缺陷。Moldflow软件用于模拟每个实验的注射过程。利用熵权确定各指标权重，计算综合评价值，将多目标优化转化为单目标优化。通过随机森林（RF）算法建立回归模型。为了进一步说明模型的可靠性和准确性，采用反向传播神经网络和克里金模型作为对比算法。结果表明，RF的误差最小，性能最好。最后，利用遗传算法寻找RF建立的回归模型的最小值。找到了最佳参数，以提高塑料零件的质量并降低能耗。经优化工艺参数生产的塑件质量优良，符合生产要求。