Research On Quality Prediction Technology Of Injection Molding Products Driven By Dat

Product quality management technology has become increasingly perfect after years of development.However,quality management is based on post-inspection of product quality,which will not reduce the unqualified rate of products.If a reasonable method is adopted to build a model for prediction of product quality,it can provide quality prediction for the production process of products and even provide solutions for product quality control,so as to achieve the purpose of reducing enterprise costs and improving enterprise production efficiency,which plays a positive role in improving the competitiveness of enterprises.Therefore,this paper carries out the following research work based on the production of injection molding products in K Enterprise:1)Taking injection molding production as an example,a quality-oriented intelligent injection molding production system was designed.In view of K enterprise’s demand for injection molding product quality prediction and intelligent production,the overall structure of intelligent injection molding system is designed and the system functions and key technologies are analyzed,from which the importance of the key subsystem-quality predictor is deduced.In view of the key technology of quality prediction studied in this paper,the implementation scheme and theoretical method are analyzed,which provides support for the subsequent construction of neural network structure and system development.2)A sparse one-dimensional convolutional neural network(SP1-DCNN)is proposed for classification.Since the traditional one-dimensional convolutional neural network can effectively extract features from process signals,but there is a lot of irrelevant information in these feature representations,which is useless for classification,a sparse one-dimensional convolutional neural network is proposed to predict product quality conformance in actual industrial processes.An example is given to verify SP1-DCNN and the results are compared with five classical deep learning algorithms.The results show that compared with other deep learning models,the proposed structure can extract higher-order discriminant features from high-dimensional nonlinear process variables of complex industrial processes,and improve the effectiveness of the model classification by eliminating inherent interference.3)Based on the SP1-DCNN structure proposed in this paper,the quality of injection molding products in K enterprise was predicted.The key parameters of 3953 molds in the process of injection molding were obtained,including 24 characteristic parameters and 3product size parameters,which were finally divided into unqualified products and qualified products.Based on the proposed SP1-DCNN model,the test samples were classified and trained,and the trained model was used for product quality prediction,and the prediction results were analyzed.4)Based on the above work,the quality prediction subsystem,the key subsystem of intelligent injection molding system,is designed and developed.This paper introduces the environment needed for the development of quality prediction subsystem,and introduces the quality prediction subsystem in detail from four aspects: system objective,system structure design,system function design and database logical structure design.Finally,the interface of the main function modules of the system is designed and developed.