Generalized Zero-Shot Operating Performance Assessment Based on Performance Grade Similarity in Industrial Processes

Process operating performance assessment (POPA) is essential for pursuing the economic benefits of industrial processes. In this study, a generalized zero-shot POPA method is proposed based on the similarity between different performance grades. Firstly, a stacked denoising autoencoder (SDAE) is used to extract the deep features closely related to the operating performance, and then the generalized zero-shot POPA model is established. In addition, a performance grade similarity matrix (PGSM) is constructed by communicating with experts, consulting literature and leveraging accumulated experience. During the online POPA phase, a gating mechanism is introduced to assess whether the performance grades are seen or unseen. Furthermore, the PGSM is used to accurately assess the unseen performance grades. Finally, a hydrometallurgical process is used to verify the superiority of the proposed method.