Artificial Intelligence-based Fault Detection and Diagnosis: Towards Application in a Chemical Process

As the chemical processes become more and more complicated, it is crucial but also difficult to detect and diagnose faults in the system than ever. With the development of machine learning algorithms and cloud computing, to monitor chemical processes completely with the help of big data become true. However, most machine learning methods suffer from unexplainable problem. By employing Explainable Artificial Intelligence(XAI) techniques, this paper applies two kinds of unsupervised machine learning methods, One-Class Support Vector Machine and Depth-based Isolation Forest Feature Importance, to detect and diagnosis faults in a chemical process. By combining two different fault explanation methodologies, accuracy and reliability of the fault diagnosis results can be achieved by the validation in a simulated continuous stirred tank reactor.