Relevance variable selection variational auto-encoder network for quality-related nonlinear process monitoring

Yao Ma, Hongbo Shi, Shuai Tan, Bing Song, Yang Tao

Abstract: Highlights•A novel semi-supervised quality-related process monitoring method is proposed.•A novel latent variable selection strategy is designed.•A novel variable contribution plot method is proposed to identify the responsible fault variables.•A data weighting strategy is proposed to enhance the strong quality-related variables.