Accuracy Improvement of Fault Diagnosis Methods for Small Modular Pressurized Water Reactors Based on Machine Learning Methods

Abstract

The paper proposes a fault diagnosis method for small modular pressurized water reactors (SMPWR). Due to the compactness and complexity of SMPWR, for general fault diagnosis methods, it is difficult to extract the corresponding special complex fault characteristics, resulting in a great increase in the difficulty of diagnosis. Therefore, this paper intends to propose a more applicable diagnosis method in close combination with the characteristics of SMPWR. Traditional machine learning methods based on feature selection and feature extraction work well in other research areas, but when it comes to SMPWR, it is difficult to determine the features that should be extracted, and different feature selection can cause great interference in the diagnosis results. Therefore, this project proposes to use advanced deep learning algorithms to automatically learn fault features of complex systems, reduce human interference, and build up more reasonable and accurate intelligent diagnostic models.