Development of diagnostic and monitoring algorithm for the condition of a check valve

The failures of check valves are one of the most important problems to safety in nuclear power plants because the reverse flows through the failed check valve impact on the healthy hydraulic loop. We have suggested the monitoring and diagnostic algorithm for the condition of a check valve by using acoustic emission sensors for detecting the failures of check valves through measuring and analyzing the backward leakage flow without disassembling the check valve. Since the acoustic emission sensor could detect the sound originated from any flow, we have chosen the acoustic emission sensors as the detecting devices for sound originated from backward leakage flow of a failed check valve. For validating the suggested algorithm, we designed a hydraulic test loop installing a typical four inches swing typed check valve. We have assumed that the major failure modes of check valves are disk wear in operation period and insertion of foreign object. The test loop was designed to identify the mechanical failures of the check valve in the case that the reverse backward leakage flows are induced through a failed section in the check valve with disk wear or an inserted foreign object. After experimenting the hydraulic test with various sized disk wear or inserted foreign objects at the various pressures with room temperature, we have developed the monitoring and diagnosis algorithms by using a neural network model in order to identify the type and size of failure in the check valve without any help of expert's knowledge