Intelligent multi-severity nuclear accident identification under transferable operation conditions

Nuclear power plants (NPPs) have witnessed significant advancements in intelligent accident identification in recent years. However, comprehensive research on fine-grained analysis of accident severity levels has been lacking, thus limiting its practical application in real operating environments. This study proposes a novel intelligent nuclear accident identification method based on the designed dual-branch domain adaptation neural network (DBDANN). The network employs the multiple kernel maximum mean discrepancy (MK-MMD) to measure the feature distribution discrepancy between the source and target domains. Additionally, a deep residual convolutional neural network (DRCNN) is integrated with two separate classifiers with dual-branch design for accident type and severity identification parallelly. The results from three transferrable scenarios for accident identification in nuclear simulation platforms demonstrate the effectiveness of the proposed method in providing fine-grained severity level recognition. Moreover, the proposed method holds promising potential for future intelligent model expansion and practical deployment.