PCA-SVM method with sliding window for online fault diagnosis of a small pressurized water reactor

Abstract

Traditional fault diagnosis methods based on signals and operation procedures in small pressurized water reactors (SPWRs) exhibit low efficiency. Complex alarm signals cannot be addressed sufficiently fast via the judgment of the operators alone. Therefore, a data-based fault diagnosis method is proposed in this paper to improve the accuracy and speed of fault diagnosis. First, the principal component analysis (PCA) method was applied for data feature extraction and status monitoring. Second, combining the data features from PCA and support vector machine theory, a three-layer fault classification model was established to diagnose the fault type, location, and degree. The sliding window method was utilized to realize fault diagnosis online. Finally, three typical faults of an SPWR were simulated using the RELAP5 code to generate fault sample data. After training and testing with sample data, the results show that this fault diagnosis method has high accuracy and fast speed.