Abstract
The importance of condition monitoring technology for equipment has increased with the introduction of condition-based maintenance in nuclear power plants. We are developing a diagnostic system using process signals for plant equipment, such as pumps and motors. It is important to enable the diagnostic system to distinguish sensor drift and equipment failure. We have developed a sensor drift diagnostic method that combines some highly correlative sensor signals by using the MT (Mahalanobis-Taguchi) method. Furthermore, we have developed an equipment failure diagnostic method that measures the Mahalanobis distance from the normal state of equipment by the MT method. These methods can respectively detect sensor drift and equipment failure, but there are the following problems. In the sensor drift diagnosis, there is a possibility of misjudging the sensor drift when the equipment failure occurs and the process signal changes because the behavior of the process signal is the same as that of the sensor drift. Oppositely, in the equipment failure diagnosis, there is a possibility of misjudging the equipment failure when the sensor drift occurs because the sensor drift influences the change of process signal. To solve these problems, we propose a diagnostic method combining the sensor drift diagnosis and the equipment failure diagnosis by the MT method. Firstly, the sensor drift values are estimated by the sensor drift diagnosis, and the sensor drift is removed from the process signal. It is necessary to judge the validity of the estimated sensor drift values before removing the sensor drift from the process signal. We developed a method for judging the validity of the estimated sensor drift values by using the drift distribution based on the sensor calibration data. And then, the equipment failure is diagnosed by using the process signals after removal of the sensor drifts. To verify the developed diagnostic system, several sets of simulation data based on abnormal cases of the plant pumps were used. Through the diagnosis test using the simulation data, the diagnostic system was able to distinguish the sensor drift and the equipment failure. Thus we confirmed the effectiveness of the developed diagnostic system.
