Resistance Spot Welding Based On Artificial Neural Network Expert System Design

It has great interest to establish Expert System (ES) for spot welding process. Considering the traditional ES t~ defects such as the bottleneck problem in the knowledge acquisition, this paper presents a mingle intelligence system by integrating ES and artificial neutral networks (ANN) which has learning ability, parallel processing ability and non-linearity, etc. It takes advantage of two technologies complementary abilities and overcomes the shortcomings resulting from using ES or ANN independently. After discussing the BP learning algorithm of ANN, the knowledge base and inference engine are developed by using the BP algorithm in the ES 憇 knowledge acquisition. The database management system of spot welding process is introduced to accomplish the easily management and application of data. In addition, the better uniformity and enough coverage of ANN 憇 learning pattern is obtained with the lesser experiment times because of the use of cross plan in collecting the training data of ANN. Finally, this paper proposes two ANN models for spot welding parameter selection and joint quality prediction. The trial running results show that this system has the merits of ease of use, high prediction accuracy and is proved to be feasible and effective.