A study on spot welding quality judgment of stainless steel plates based on quantum generative adversarial network and hidden Markov model

During the process of resistance spot welding (RSW) of stainless steel plates, utilizing welding experiment to obtain samples exists shortcomings such as high cost, poor process repeatability, and imbalanced sample sets, which lead to a high model training cost and poor pattern classification performance for quality judgment model. In view of this, a spot welding quality judgment method based on the combination of quantum generative adversarial network (QGAN) and hidden Markov model (HMM) was presented in this paper.

Firstly, employing a generative adversarial network (GAN) to expand the dataset of unqualified welding points, to address the imbalanced datasets caused by experimental methods. Subsequently, integrating quantum computing into the GAN framework to reduce the number of parameters that require modulating and enhance the quality control capability for generated samples. Finally, applying the proposed method to a practical application of spot welding in the roof of stainless steel rail vehicles. The results demonstrated that the proposed method reduced the number of parameters requiring modulation in the GAN to five; the average training and test times of the model were 8.28 s and 4.68 s, respectively, which were lower than those of GAN-HMM (10.44 s and 7.0 s) and HMM (13.4 s and 9.76 s). Moreover, the classification accuracy across all five quality states exceeded 90 %, outperforming both GAN-HMM and HMM. Therefore, the method proposed in this paper was effective.