Study On Error Synthesis Prediction Model In Multistage Manufacturing Process

At present, with the accelerated upgrading of new products, the diversification of user's need, and the constant emergence of new technologies, manufacturing enterprises are facing frequent and unpredictable changes in market conditions. In this case, more and more enterprises realize the importance of improving product quality. Improving the quality of the product is key for an enterprise to win competition for survival and development. The quality of the product evolves in the various stages of product design, manufacturing and marketing. Among these, manufacturing quality is the insurance of product quality, and quality control in manufacturing is an important part of product quality control for enterprises, that means design quality is never enough without quality assurance in manufacturing.However, compared with industrialized countries, especially European countries, China's machinery products have relatively low level of quality, especially in the manufacturing quality assurance capabilities, quality problem is a key for China's machinery products to access to the world market and increase their competitiveness. At this stage, more mature approach of manufacturing quality control is the process quality control, that is to use various quality tools to analyze quality problems, and to identify causes of quality problems and then make improvements, named control during and after the event. However, the best way to control the manufacturing quality is to make the manufacturing enterprises have the ability to predict in advance and to optimize system performance. And then in order to prevention, we should first predict the manufacturing quality. But at present, the prediction for manufacturing quality is only for one or some of the key operations, and there are few systematic researches on the factors which have influence on quality fluctuation. For these reasons, this paper presented the comprehensive predictive modeling of errors which produce the quality fluctuation in the mechanical parts manufacturing processes to predict quality fluctuation of a part produced in one procedure so far, and to provide the basis and technical foundation to analyze the causes of abnormal fluctuations and improve quality during the entire manufacturing process.First, the paper discussed the basic theory and control methods of errors in multistage manufacturing process, analyzed the relevant theory of the quality fluctuation and multi-process manufacturing process. And then based on the basic theory of the error control, the formation mechanism of quality fluctuation was studied and the part quality fluctuation network model was set up. According to the characteristics of the multistage manufacturing process, the concept of two errors was put forward, and then the error propagation mechanism was researched, finally a multi-process error prediction system was established.Then, in view of the complex relationship between the errors and error sources (5M1E), the paper proposed the prediction modeling method using least squares support vector machine (LS-SVM) for the independent error in single process, and described the method in detail as well as gave the corresponding prediction flow. According to the quality fluctuation network model and the error propagation mechanism proposed above, the error transfer function among the multistage process was founded up by using the characteristic of error normal distribution to realize the separation and synthesis of errors. Thus, combined the above two processes, an integrated prediction model of process error was proposed, which provided the basis support for the subsequent error analysis and prevention and control.Finally, taking the 3- step processes for axis machining as an example, the comprehensive prediction model of small end shaft error was established to verify the validity and feasibility of the model that was proposed by this paper.