Health Monitoring of Pressure Transmitter Sensing Line Based on Convolutional Neural Network

Blockages of pressure transmitter sensing line are a recurring problem in nuclear plants. The traditional solution is to purge them with nitrogen periodically, which is time consuming and radiation intensive. It is thus desirable to monitor the status of sensing lines on-site. The noise analysis method is a promising method for sensing line health monitoring. However, the existing noise analysis methods, which evaluate the sensing line status by analyzing time domain or frequency domain features of process noise signal, are susceptible to working conditions of pressure transmitters. Further, blockage degree of the sensing line can not be determined by existing methods. Therefore, this paper proposes a machine learning based noise analysis method for health monitoring of pressure transmitter sensing line, in which 1-D Convolutional Neural Network (CNN) is used to extract features from process noise signals. Experimental results show that the proposed method achieved satisfactory results in diagnosis accuracy, false alarm rate and robustness of the model under different working conditions. Further, blockage degree of sensing line can also be evaluated