抄録
The spot-welded joint is widely used as a structural part. In the quality control of the joint, one of the most important issues is to quantitatively estimate connective strength such as tension shearing load from the information detected by nondestructive testing as accurately as possible. In this paper, a system based on an ultrasonic inspection device and a neural network is proposed to evaluate the tension shearing load of each spot-welded joint. The network is configured by learning the pattern sets dealing with the interaction between the scanning graphs of the welded region using ultrasonic testing and tension shearing load of the joints. The input of the network is the echo height at each scanning position of the welded region. The output from the system was examined by comparing it with the experimental results, and the validity of the system was thereby confirmed.
