The Research On Quality Diagnosis And Adjustment In Processing Based On Artificial Intelligence

Product quality is formed in and runs through the whole process of product life cycle. It is the base for the enterprises'participation in market competition, and their existence and development. The product quality forming in processing is the cornerstone of the final quality. With the proposal of "world class quality", the market, customers and enterprises make increasingly stringent demand on product quality, thus the traditional quality control cannot meet their requirements.When product quality in processing is abnormal, only a little time is left for quality monitoring and control, however, about 80 percent of the time is used to diagnose and adjust the abnormal factors. In order to meet requirements of total quality management in processing, many researchers and enterprises focus on the quality diagnosis technology and system applied in processing. Based on previous studies of quality control, the present thesis proposed the control chart pattern recognition arithmetic based on the improved BP neural network and the methods of quality diagnosis and adjustment in processing.The conclusions of the studies are described as follows:(1) An improved BP neural network algorithm for control chart pattern recognition is proposed, which is based on adjustable activation function parameters and dynamic threshold. In order to make the quality characteristics of sample data the same as actual production data, process data Monte Carlo simulation method is optimized. In accordance with the improved iterative formula of network parameters, the sample data is pretreated as input to the neural network recognizer for training. And the training results are used in control chart pattern recognition during production. On the premise of guaranteeing recognition accuracy, the improved BP neural network recognizer, whose topological structure is simple, can increase the recognition speed and improve the generalization ability of neural networks. Finally, the feasibility of this algorithm is validated through computer simulation and application in production. (2) An expert system method based on the fault tree analysis is proposed, which is used in quality diagnosis and adjustment in processing. First of all, the rules related quality in processing are coded and represented with production knowledge. The system automatically generates the fault tree, whose top event is the control chart anomaly pattern. The main abnormal factors subset caused abnormal quality is obtained by fault tree analysis. Then, the system deals with automatic inference and artificial inference according to the monitoring result of quality factor eigenvalue. Finally, the system adjusts the anomaly factors based on the results of quality diagnosis, and the optimal adjustment scheme is fed back to technicians to improve production in time. This expert system for quality diagnosis and adjustment helps to diagnose quality abnormal factors rapidly and implement quality adjustment scheme, so as to shorten the production cycle greatly.(3)An expert system for quality diagnosis and adjustment taken the example of the turning processing is developed. Adopting VB6.0 development environment and SQL Server 2003 database software, the expert system database and the functional modules for man-machine interaction are built. This system realized the control chart pattern recognition in processing, the diagnosis and adjustment to quality anomaly and expert knowledge base maintenance and machine self-learning.