Semisupervised Graph Contrastive Learning for Process Fault Diagnosis

The complexity of unit interactions and the scarcity of labeled samples pose great challenges to effective fault diagnosis of industrial processes. To this end, a semisupervised fault diagnosis model based on graph isomorphism contrastive learning (GICL) is proposed. To model fault propagation, a topology graph with process variables is constructed to guide GICL to model the interactions between process units. Since the topology graph is considered isomorphic under different faults, graph isomorphism embedding is used for contrastive learning to enhance the discrepancy between isomorphic samples, thereby mining the intrinsic information on each sample. For clarity of the decision boundary, the mapping between intrinsic information and partial labels is simply constructed. Additionally, to enhance the understanding of the model’s predictive logic, the score vector and variable importance are calculated. Finally, two cases show the effectiveness of GICL using different label ratios and demonstrate the physical consistency of prediction logic.