AUTHOR=Huang Jiahui , Liu Jie , Yang Xiaohua , Ma Jiayu , Liu Hua 

TITLE=Research on the uncertainty problem of SDG fault diagnosis based on information flow

JOURNAL=Frontiers in Energy Research

VOLUME=Volume 10 - 2022

YEAR=2023

URL=https://www.frontiersin.org/journals/energy-research/articles/10.3389/fenrg.2022.1013255

DOI=10.3389/fenrg.2022.1013255

ISSN=2296-598X

ABSTRACT=The rapid and accurate diagnosis of nuclear accident is the primary link of nuclear emergency.  Signed directed graph (SDG)has good completeness as a common method for fault diagnosis, but if there are feedforward and negative feedback structures in SDG, compensatory responses and inverse responses will occur in the process of fault occurrence and propagation.  The state of the compensation variable and the inverse variable becomes uncertain, thus destroying the compatibility of the fault propagation path, resulting in failure to the fault diagnose.  In this paper, a deterministic diagnosis method for SDG faults based on information flow is proposed. This method compares the goodness of fit between the sub-information flow of all causal pairs related to compensation variables or inverse variables and the total information flow curve, and obtains sub-information flow curve that has the highest .  The causal pair represented by the information flow is located in the compatible path, and the state of the compensation variable or inverse variable node and the fault traceability are determined according to the fault propagation path with the highest probability. At the same time, according to the diagnosis result, the SDG model is structurally extended as the FR-SDG model, which can intuitively describe the steady information for node state and the main fault propagation path.  It realizes the fast and accurate fault diagnosis from the perspective of introducing data fusion method into response and extending SDG model.  Finally, the experiment of the primary loop subsystem of nuclear reactor is carried out in the simulation system PCTran AP1000, which verifies the feasibility of this method.