In real-world industrial processes, it is crucial to avoid damaging faults to ensure steady operations. However, new faulty conditions can arise in variable production environments. Traditional fault diagnosis methods perform poorly with few or no fault data for training. To address this issue, we propose a graph-regularized coupled dictionary learning (GRCDL) method for zero-shot fault diagnosis in industrial processes. Our method enhances the coupled dictionary learning (CDL) framework by integrating class prototype learning, which utilizes the rich information of seen class faults in the data feature space. A feature correlation function is also designed to facilitate the transformation and reconstruction of data and semantic features in the shared space, preventing certain classes from becoming the nearest neighbors of unrelated classes in a low-dimensional semantic space. To mitigate noise interference and preserve local manifold structures, the graph-regularized constraint is developed to enhance the robustness and locality of the coupled dictionaries. Considering the different underlying distributions between unseen and seen faults, a novel domain adaptation function is proposed. This function uses the learned coupled dictionaries as mediums for knowledge transfer and update from seen to unseen classes and effectively learns the class prototypes of unseen faults. Extensive experiments in industrial processes verify the superior performance of our method for zero-shot fault diagnosis.

Aluminum alloy is widely used in transportation, catering, industry, sports, health and other fields, because of their excellent high specific strength and corrosion resistance. Aluminum industry has been an important mainstay industry of a national economy. In the process of electrolytic aluminum, the superheat degree is a very important production target. When an electrolysis cell is working in the appropriate superheat degree state, the life of cell is prolonged and the amounts of aluminum released will enhanced. However, to measure the superheat degree is very difficult and the measured results cannot timely feedback to the process of production. To address the problem, a soft sensing prediction model of superheat degree is proposed in this paper, in which some new concepts such as the decay function of data weight, the credibility of a rule and the rule tree are introduced. Basically speaking, the processing of the soft sensing prediction model is mainly based on the rough set data analysis method and a tree data structure. The static rules are obtained from the history data by using the attribute reduction and value reduction method in the rough sets. The rule tree is updated timely based on the incremental data set accordingly. The effectiveness of the proposed model is verified with the aluminum production data provided by Shandong Weiqiao Aluminum Electrolysis limited company in China.