Research On Critical-to-quality Characteristics Identification And Final Quality Level Prediction For Complex Product Manufacturing Processes

All the time,product quality is the lifeline of complex product manufacturing enterprises.Quality control in the manufacturing processes of complex products has become the fundamental way to ensure the final quality of products.However,with the continuous innovation and development of the emerging information technology such as the Internet of Things and the Big Data,the quality control of complex products manufacturing process has appeared a technical bottleneck,which has made the quality control of complex product manufacturing process become a hot research point at home and abroad.In this paper,the issue of critical-to-quality(CTQ)identification and final product quality prediction in the quality control of complex product manufacturing process are studied.The main contents are as follows:(1)In order to effectively identify the CTQ sets that affect the final quality of the product,we propose a feature selection algorithm based on an improved DE algorithm and apply it to the CTQ identification of complex products.First of all,the proposed algorithm mainly improves the algorithm's ability both in global search and local search by changing the mutation strategy and crossover strategy.Then,we design a series of comparison experiments,by utilizing the original DE algorithm and IG algorithm as compared algorithm to verify the effectiveness of the improved algorithm in CTQ identification.At last,the experimental results show that the proposed algorithm not only makes the prediction accuracy of the subsequent classification algorithm higher,but also reduces the dimension of the CTQ set,which has a great effect on improving the performance of the classification algorithm.(2)This thesis constructs a product final quality prediction framework for multi-complex manufacturing process.In this framework,we use support vector machine(SVM)as the core prediction algorithm.In addition,in order to optimize the training process of the SVM prediction model,a SVM model parameter selection algorithm based on a two stage heuristic is proposed,and the efficiency of the algorithm is verified by contrastive experiments.To verify the practicability of the proposed framework,we carry out an experiment by using some relevant data.The experimental results show that the proposed final quality prediction framework is effective.