Ultrasonic Fast-Identification Expert System of the Auto Stick-Weld Defect based on Echo-characteristics Analyzing

摘要

The current inspection methods of spot-weld quality are difficult to achieve an ideal result especially for the stick-weld defect which is one of the most important types of spot-weld defect in the automotive body. This paper thus developed one fast-identification method to identify the joint defects quickly and efficiently based on quantitative analyzing the echo characteristics of ultrasonic curves that could reflect the spot-welding joint defects. The echo-characteristic parameters were analyzed quantitatively through the decision-making bintree of defects and ART networks. After knowledge acquiring and information processing, a fast-identification expert system (FIES) was developed to identify the automotive body spot-weld defects, through collecting the standard ultrasonic curves and quantitatively analyzing. Finally, a series of experiments were conducted to verify the proposed methods and the results showed that FIES is credible and the identification rate can exceed 95% in total test samples.