Instrumentation data in Nuclear Power Plants(NPPs) is vital for facility monitoring and integrity assessment. However, factors such as sensor aging and noise can generate error signals, compromising data reliability and decision-making. It is essentia...
Instrumentation data in Nuclear Power Plants(NPPs) is vital for facility monitoring and integrity assessment. However, factors such as sensor aging and noise can generate error signals, compromising data reliability and decision-making. It is essential to effectively detect these error signals to enhance instruments data reliability.
In this study, I propose a 1D CNN-LSTM deep learning model combined with an attention mechanism. Focal Loss was applied to address the class imbalance. By visually identifying critical time points throughout the entire period via the analysis of attention weights, this study aims to reduce the workload of experts required for data verification and contribute to the improvement of data quality.