Quality assessment for resistance spot welding based on Bayesian image recognition technology

A method for converting electrode displacement signal to binary image in resistance spot welding process is proposed. Based on image characteristic analysis,fifteen image features are extracted from the binary image of electrode displacement waveform. The principal component analysis is used to remove the cross correlation among image features,and a series of weld specimens with different welding quality are selected to develop a quality classifier. The test results based on Bayesian image recognition technology of minimum risk show that it is feasible and reliable to utilize the binary image of electrode displacement signal to evaluate the weld quality and the image conserves the information of the weld quality. The algorithm for image feature extraction is simple,efficient and easy to use. At the mean time,the Bayesian image recognition technique with small samples can realize the welding quality assessment rapidly and accurately, and the method has a broad application prospect.