A Fault Detection Method for Gas Pressure Regulators Based on Improved Dynamic Canonical Correlation Analysis

This paper proposes a new data-driven fault detection (FD) approach for a class of gas pressure regulators based on the improved dynamic canonical correlation analysis (DCCA) method. First, the general principle of gas pressure regulators is introduced, and several typical faulty cases are presented accordingly. Using the reference signal, input, and output data, the improved DCCA-based residuals are generated. Based on this, the thresholds are determined using the randomized algorithm to build the FD logic. Compared with existing methods, the proposed approach can extract the fault information from the input and output data instead of only the output signals. To verify the effectiveness of this approach, experiments on the practical data in fault-free and two faulty cases are conducted.