Fault diagnosis method for Small modular reactor based on transfer learning and an improved DCNN model

Abstract

Deep convolutional neural networks (DCNN) are widely applied in the realm of deep learning. This paper presents a novel approach that combines transfer learning techniques with a hybrid domain attention mechanism module to enhance and refine the DCNN architecture, consequently boosting its performance. The focus of this study is the application of the improved DCNN model to fault diagnosis within small modular reactor. We hope to improve the fault monitoring and diagnosis capabilities of small modular reactors through algorithm improvements, to enhance their safety. Comparative results demonstrate that the improved model surpasses other deep learning models in terms of convergence rate, recognition accuracy, and model size, evidencing robust generalisation capabilities.