Multi‐mode process fault diagnosis method integrating <scp>GMM</scp> with reconstruction‐based <scp>PCA</scp>

Abstract Multiple local model approaches for multi‐mode process monitoring have garnered significant attention and research interest. The general procedures for fault diagnosis using multiple local models can be divided into two main steps. The first step involves offline mode classification and modelling for various modes, while the second step focuses on online mode recognition and fault diagnosis. However, when a fault occurs during the second step, there is a higher likelihood of misidentifying the mode, which can lead to false alarms or missed detections. To address this issue, this study proposes a novel multi‐mode process fault diagnosis method based on the Gaussian mixture model with reconstruction‐based principal component analysis (GMM‐RBPCA). This method first calculates the fault isolation results from the local models for all modes and then determines the current mode based on the statistics of these isolation results. This approach effectively reduces the risk of misdiagnosis and underdiagnosis due to inaccurate mode recognition. The effectiveness of the method is validated through mathematical simulations and testing in a continuous stirred tank heater (CSTH) process.