Research On Profile Monitoring Of Multi-stage Manufacturing Process Based On Elastic Net

With the continuous improvement of the productivity level of the manufacturing industry,products have gradually developed toward diversification and complexity.The manufacturing process has also changed from simple production to complex production methods represented by multi-stage continuous manufacturing processes.In a multi-stage manufacturing process,fluctuations in process quality characteristics are iteratively transferred between processes.Product quality is not only affected by the process capability of the process,but also by the quality fluctuations of the upstream process.It is reflected in the final product quality after passing through layers.Therefore,how to conduct quality analysis and effective monitoring on the characteristics of the multi-stage manufacturing process is essential to maintain the stability of the manufacturing process and improve the ability of the product manufacturing process.Traditional multi-stage manufacturing process quality monitoring often monitors the quality characteristic values in the process individually or jointly.However,in the actual production process,the stability of the multi-stage manufacturing process cannot be reflected only by statistical process control of one or more quality characteristics.The complex collinearity between the quality characteristics in the multi-stage manufacturing process often leads to the quality characteristics are extremely sensitive,and the ability to recognize the overall abnormal fluctuations of the process is not significant.Therefore,in recent years,many scholars have introduced the idea of Profile monitoring to the quality monitoring of multi-stage manufacturing process.By monitoring the functional relationship between the quality characteristics of each sub-stage and the quality of the final product,the quality monitoring of multi-stage manufacturing process is realized.At the same time,there are many quality characteristics in the multi-stage manufacturing process to reflect the impact of equipment,process detection,process parameters and environment on product quality.Based on the principle of critical minority and non-critical majority,the quality characteristics are sparse and due to the cost and technical reasons,quality control personnel cannot monitor all quality characteristics.Therefore,effective identification of key quality characteristics is the primary task to achieve Profile monitoring of multi-stage manufacturing processes.To sum up,this article follows the following ideas to develop a multi-stage manufacturing process quality monitoring research.First,in order to effectively deal with multi-collinearity between quality characteristics,this paper introduces the state space idea to build a multi-stage manufacturing process quality relationship model for the characteristics of multi-stage manufacturing process;then,the elastic network method is used to fit and analyze the quality relationship model and pass variable selection identifies key quality characteristics.Through simulation analysis to compare the effectiveness of elastic net and ridge regression and Lasso in the identification of key quality characteristics;further,based on the multi-stage manufacturing process Profile monitoring model,the model coefficient parameters are jointly monitored.Finally,the effectiveness of the proposed method is verified by simulation and example analysis.The results show that:(1)the elastic net method is more advantageous in the identification of key quality characteristics of multi-stage manufacturing processes with strong correlation and group effects;(2)when the quality characteristics have strong correlation and group effects,based on the Profile monitoring performance of the elastic net method is better than the Profile monitoring performance based on the ridge regression and Lasso method;(3)under different degrees of process quality fluctuation,the Profile monitoring performance based on the MEWMA control chart is better than the Profile monitoring performance based on the T~2 control chart.