Data-Driven Inference Models Of Plastic Injection Molding Process Parameters

Injection molding is the most crucial process for forming plastic products,and process parameters are one of the critical factors affecting the appearance,size,and performance of products.However,the optimization of the process parameters is weakly theoretical and strongly empirical problems,and there is an urgent need to develop a scientific and systematic method.The optimization methods of injection molding process parameters mainly include case-based reasoning,expert system,search based on surrogate models,etc.,but these methods still exist some problems in engineering application,such as process knowledge dependence,low optimization efficiency,and insufficient optimization accuracy.The data-driven inference models of injection molding process parameters are expected to break through the application bottleneck of setting and optimizing process parameters.This paper focuses on these studies,and the main work and results are as follows.Firstly,existing process retrieval methods are lack high-dimensional parts information,leading to a low matching degree of case retrieval.To solve this challenge,the injection pressure curves were proposed to characterize the information of the forming process in the mold.Further utilizing the variational autoencoder reduced the dimension of the pressure curve,mapping the high-dimensional curve space that was difficult to compare to a unified low-dimensional variable space.The results of similarity classification and retrieval cases on 120 products showed that compared with the simple forming features and pure geometric features,the accuracy of feature representation increased from about 73% to 86%,and the process retrieval error reduced from 15% to 10%.Secondly,in response to the problem of the strong dependence of manual experience in product defect correction,we constructed unified fuzzy rules of experience and knowledge and established a Takagi-Sugeno-Kang(TSK)fuzzy rule network model integrating knowledge representation and inference of process parameters optimization.Furthermore,a learning method was proposed to automatically discover optimization rules of process parameters from process datasets.The Dropout strategy and Bagging ensemble learning strategy were adopted to alleviate the problem of rule explosion caused by the growth of process knowledge bases in high-dimensional process data.Thirdly,the influences of the fuzzy rule network parameters and structure on knowledge representation and inference were analyzed.Based on these analyses,two methods,parameters learning and structure learning of the model were developed respectively.The learning method of process data based on experience pool replay was proposed,establishing the incremental learning strategy of process data.The test results of the benchmark datasets showed that compared with the existing method,the average inference accuracy of the proposed method increased by about 5%,and the rule readability was greatly improved.On the injection molding dataset,the number and length of rules were reduced by 50%,realizing high interpretability as well as incremental learning stability.Further,aiming at the problem of insufficient precision of process parameters optimization in existing methods,the accurate process parameters optimization model based on the Markov decision process was established.The simulation data and proxy model were fused,realizing the reinforcement learning of the Markov decision process parameter optimization strategy.The forming experimental results of the high-precision optical lens showed that the number of parameter optimization was reduced by 20% compared with the traditional expert system.It met the requirement of ±5 μm manufacturing accuracy of optical lens,and the process capability index increased by ～ 4 times.Finally,based on the above models,a prototype of an intelligent parameter setting and optimization system for injection molding was developed.The application case of 10 products in the enterprise showed that the success rate of optimizing process parameters was 100%,and the average inference times was 6～7.Compared with the traditional expert system,the number of process parameter optimization of the system was reduced by 30%.In terms of forming stability,the process capability index of the systematically optimized parameters was about 3.5 times that of the manually optimized parameters.This paper studies the data-driven inference models of process parameters and developed a prototype system verified by actual experiments.The research results are valuable for improving the injection molding process level and provide some guidelines for the setting and optimization of forming process parameters in other fields.