Explainable Fault Diagnosis of Control Systems Using Large Language Models

In control systems, accurate and timely diagnosis of malfunctions can ensure the safe and efficient operation of the systems. Although several methods have been proposed for process anomaly detection, including multivariate statistical process control, most of these models are built on several statistical assumptions that limit their applications. When these models detect faults with high accuracy, questions such as "How did the model achieve this outcome?" "Are the predictions valid?" and "Do the outcomes reveal novel information?" often race through our minds. Therefore, to ensure explainability in fault diagnosis of control systems, we propose a causal-based large language model that can potentially answer cause-and-effect questions within these systems, ensuring transparency, interpretability, and robustness.