Knowledge Graph Embedding With Graph Convolutional Network and Bidirectional Gated Recurrent Unit for Fault Diagnosis of Industrial Processes

Abstract

The stability and reliability of modern industrial processes are key factors in ensuring production safety and achieving high quality and efficiency. Complex industrial processes are characterized by dynamics, nonlinearity, and multivariable coupling, which present challenges for fault diagnosis. To address the limitations of traditional fault diagnosis methods in handling complex industrial processes, a fault diagnosis framework of knowledge graph embedded graph convolutional network and bidirectional gated recurrent unit (KG-GCBiGCN) is proposed in this article. First, a domain knowledge graph of the industrial process is constructed in a top-down manner, leveraging the expertise of domain specialists. Next, node correlation analysis is performed using the maximum information coefficient (MIC) alongside the knowledge graph, resulting in the formation of an adjacency matrix. Then, the knowledge graph is embedded within the graph convolutional network (GCN), and a bidirectional gated recurrent unit (BiGRU) is introduced to capture the dynamic correlations present in time-series data, thereby enhancing the performance of the fault diagnosis model. Finally, fault diagnosis is achieved based on the predicted residuals. By integrating the deep information from the knowledge graph, the framework facilitates inference of the root cause of faults and identifies the propagation paths of the faults. The effectiveness of the proposed method is validated using actual data from the hot strip mill process (HSMP), achieving a fault detection accuracy of 99.23%.

Topics & Concepts