Data-Driven Fault Diagnosis of Chemical Processes Based on Recurrence Plots

A method for the detection and diagnosis of various faults in chemical processes based on the combination of recurrence quantification analysis and unsupervised learning clustering methods is proposed. In the recurrence analysis, determinism and entropy were used to extract the features that influence the process in each fault, and thus fault detection was provided. Different clustering methods including k-means, density-based spatial clustering of applications with noise (DBSCAN), and clustering using representatives (CURE) were used, and comparisons were made based on the accuracy of diagnosis given the type of fault. The Tennessee Eastman process and the four-water-tank process were used to highlight the applicability of the proposed method. The performance of this method was compared with other commonly used methods, such as the principal component analysis (PCA) and kernel-principal component analysis (KPCA) methods. It is shown that the DBSCAN method has a superior performance to the CURE and the k-means method. Also, the recurrence plot method, as a preprocessing method, performs better in combination with DBSCAN and CURE. Performance of the
proposed method was also assessed using online data, and it was demonstrated that the recurrence plot method is capable of detecting and diagnosing faults in chemical processes during operation.