深度学习在塑料模具设计优化中的模型构建与分析

打开文本图片集

中图分类号：TG76 文献标志码：A

Abstract： Optimization of plastic mold design is of great significance to modern manufacturing industry. However，the traditional methods are often complicated and inefficient，so it is difficult to meet the increasingly stringent requirements. This paper mainly explores the application potential of deep learning （DL） technology in plastic mold design optimization， and constructs a DL model based on convolutional neural network （CNN）. In the research， experimental data，industrial production data and data set generated by simulation software are integrated，and the data quality is improved by feature engineering and pretreatment. Then，the DL model optimized for cooling system layout and gate position is designed and trained. It is not difficult to find that this model is obviously superior to traditional methods such as response surface method and genetic algorithm in prediction accuracy （ R2=0.93 ）， calculation efficiency （optimization within 45 s） and actual production qualification rate （2 （95%） . Taking an automobile dashboard die as an example，the filling time is shortened by 18% and the warpage is reduced by 23% after optimization. The research shows that DL can solve the multivariable nonlinear problem to some extent and provide an intelligent scheme for plastic mold design.

Key words： deep learning; plastic mold design; optimization;cooling system;melt flow

0 引言

模具的设计质量对最终产品的性能、外观及生产效率有着直接影响[1]。（剩余9031字）