Research On Quality Prediction Method Of Production Based On Multi-stage Manufacturing Process Data Analysis

In the multi-stage manufacturing process,there is a complex nonlinear relationship between the product quality and the quality characteristics and process parameters of each manufacturing stage.It is difficult for traditional prediction methods to accurately predict the product quality by using real-time manufacturing process data.Under the background of intelligent manufacturing,the real-time data acquisition of multi-stage manufacturing process is realized,and the data analysis method is applied to mine and analyze the manufacturing process data,which can effectively improve the accuracy of quality prediction.In order to solve the problem that it is difficult to accurately predict the product quality in the multi-stage manufacturing process,the coupling process parameters and quality data are analyzed and the quality rules are formed.The main research contents are as follows.(1)Analysis of product quality problems in the multi-stage manufacturing process.The main characteristics of multi-stage manufacturing process and product quality problems in manufacturing process are studied.The process of multi-stage product quality problem prediction based on data analysis is analyzed.(2)Research on quality prediction considering uncertainty and data imbalance.There are many uncertain factors that affect product quality abnormalities,and the imbalance of quality data leads to the problem of low quality prediction accuracy.Apply the rule-based deep confidence network(RBDBN)to mine the process parameters and quality characteristic data in the multi-stage manufacturing process,form quality classification rules to optimize the traceability of abnormal process parameters;Establish a product quality prediction model based on Cat Boost to reduce the impact of unbalanced quality data on prediction accuracy.(3)Research on quality prediction considering time series and strong correlation.Aiming at the problem of low quality prediction accuracy caused by strong timing and key abnormal process parameters cannot be accurately optimized.According to the quality classification rules and the attention mechanism of quality variables,the key parameters causing product quality anomalies are determined.A quality prediction model based on bi-directional longterm and short-term time series network(Bi-LSTM)is established to make full use of the manufacturing data at t-1 time to predict the product quality at t time.(4)Research on the application of quality prediction method based on multi-stage manufacturing process data analysis.In this paper,injection molding manufacturing is taken as a typical example of multi-stage manufacturing.In the quality anomaly prediction of injection molding products,the quality rules of injection molding manufacturing are mined by data analysis method,and the weight differentiation of abnormal process parameters is realized by using quality rules and quality variable attention mechanism;And input Bi-LSTM prediction model to accurately predict the products with abnormal quality;Based on the quality rules,the optimization strategy of quality anomaly is studied.The research shows that the established product quality prediction model has high prediction accuracy in the case of extremely unbalanced quality data,and the prediction accuracy of the model reaches 98.26%,which can accurately predict abnormal products.