Fault Diagnosis of Chemical Processes Based on PSO-GCN

As industrial production advances, ensuring process safety and minimizing failure-related losses is crucial. Detecting abnormal conditions efficiently is key. To address the complexity, high dimensionality, nonlinearity, and time-varying nature of chemical process data, this paper proposes a fault diagnosis method based on Particle Swarm Optimization (PSO) and Graph Convolutional Networks (GCN). The method denoises the data, constructs a graph using a sliding window, and uses PSO to optimize the diagnostic model's parameters and extract features. These features are then classified via a fully connected layer and classifier. Experimental results show that the method outperforms other machine learning and deep learning approaches in accuracy and precision, and excels at detecting minor faults in chemical processes.