CNN-Based Fault Detection in Nuclear Power Reactors Using Real-Time Sensor Data

Real-time sensors play a vital role in detecting and analyzing faults in nuclear power plants. By combining deep learning techniques with real-time sensor data, the detection of faults in nuclear power reactor systems has achieved remarkable results. This integration has greatly improved the accuracy and efficiency of fault detection, allowing operators to promptly address any potential issues. This paper aims to develop a Convolutional Neural Network (CNN)-based fault detection system for nuclear power reactors using real-time sensor data. The proposed system uses CNN to analyze the sensor data and detect any abnormality or fault in the reactor operation. The system is trained on a dataset of pressurized water reactors, utilizing various measurements from different parts of the reactor such as vibration, pressure, and power levels. Additionally, the paper assesses the performance of the proposed CNN architectures by comparing them to bidirectional long short-term memory (BiLSTM) architectures in detecting faults within nuclear power reactors. The results showed that the CNN architecture outperformed the BiLSTM architecture in terms of accuracy, specificity, precision, sensitivity, false positive rate, F1-score, Matthews Correlation Coefficient, and Kappa coefficient. The CNN architecture achieved an accuracy of 0.9333, while the BiLSTM architecture achieved an accuracy of 0.84. These findings show that the CNN architecture is a promising system for enhancing the safety and reliability of nuclear power reactors. The proposed system has the potential to enhance the safety and reliability of nuclear power reactors by detecting faults early and preventing catastrophic accidents.