Research On Distributed Multi-fault Diagnosis Method For Primary Loops Of Nuclear Power Plants

Fault diagnosis is an important part of operating support technology, while multi-faultdiagnosis (MFD) is a key topic of fault diagnosis. It needs to take many aspects intoconsideration. The interaction of different fault symptoms makes the amount of multi-fault alarge number, so that it is difficult to use normal matching method of fault patterns to solvemulti-fault problem. In addition, the complexity of nuclear power plant (NPP) makes MFDmore difficult.After studying on the multi-fault problem, a MFD method based on distributed strategiesis proposed in this paper. According to the structure and operation of system, NPP is dividedinto three-level subunits, and the whole MFD task is decomposed into each subunit. Thesystem subunits use Kernel Principal Component Analysis (KPCA) as basic method tomonitor operating condition of NPP and detect faults. The device subunits use physicalequations to modeling, and use Expert System to replenish those models when it is hard tomodeling or there is a much easier way to analysis. On the basis of the above, the diagnosissubunits combine those models to diagnose their tasks of MFD. Critical safety functions areused as principles of conclusion fusion to get the optimal solution MFD problem.On the basis of research mentioned above, a NPP Distributed Multi-Fault DiagnosisSystem (NPPDMFDS) is developed on the platform of VC++, and tested on full scope NPPsimulator. The result shows that distributed strategies can effectively solve the influenceamong fault symptoms, KPCA is able to detect anomalies rapidly, and the whole MFDmethod is confirmed to be reliable and accurate. The research in this paper lays the foundationfor further study.