Multi-objective optimization of injection-molded plastic parts using entropy weight, random forest, and genetic algorithm methods

Yanli Cao,Xiying Fan,Yonghuan Guo,Sai Li,Haiyue Huang

DOI: https://doi.org/10.1515/polyeng-2019-0326

IF: 1.624

2020-04-28

Journal of Polymer Engineering

Abstract: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.

polymer science