Hierarchical Directed Graph Based On Quantitative Knowledge Fault Diagnosis Approach And Its Application

In thermal power plants, a large proportion failure of the thermodynamic system has a huge impact to safety and economy of the overall unit’s operation, therefore the fault diagnosis of the thermal system plays an important role. Due to the complexity of the internal structure in the thermodynamic system and variability of the system, fault knowledge can’t be easy to establish in the fault diagnosis process. The deep knowledge is used to diagnose in the qualitative model fault diagnosis,which is a ideal fault diagnosis method of thermal system. On the one hand, the method can overcome the difficulties of knowledge acquisition, on the other hand, it can identify some unknown fault information.In this dissertation, based on the qualitative signed directed graph model, a new model combined with the hierarchical and quantitative knowledge is proposed. The new model can reduce the search space when searching for the compatibility paths, and it is much faster. Meanwhile the diagnosis results are described by the probability information, which distinguishes the possibility of the potential failure sources. Firstly, fault diagnosis method based on SDG model has the presence of low efficiency, and it is difficult to distinguish the potential failure sources in the same fault model, then it can occur "combinatorial explosion".To solve the problems of SDG fault diagnosis method, an improved hierarchical directed graph model is proposed. Meanwhile, based on the new model, a fault diagnosis method is proposed, too. This method has the completeness of SDG model. Using the layered strategy and qualitative relationship between nodes, it can find backward the compatibility paths to determine the fault candidate nodes. Finally, this method is applied to the fault diagnosis of deaerator system in the power plant. The results showed that the method effectively reduced the search space of active nodes and improved the speed and accuracy of fault diagnosis.Secondly, for the complexity structure of traditional signed directed graph of complex thermal system, it is easy to produce combinatorial explosion in search paths, meanwhile, it is difficult to distinguish multiple fault sources. Therefore,a new hierarchical graph model based on the quantitative knowledge which is named hierarchical probability signed directed graph is proposed, and on the basis of this mode, a fault diagnosis method is proposed, too. On the one hand, the hierarchical method simplifies the search paths, and it uses a combination of qualitative and quantitative methods to describe causal relationships between failures; on the other hand, the hierarchical directed graph in the mew model can reduce the search spaces, then using compatibility of branches and Bayesian inference to diagnosis the system, it can get the probability of failure and the possibility of fault sources. Finally, take the high pressure heater system in the thermal system for example, its model is established successfully. The experiment show that this method improves the resolution of the fault source and reduces the search spaces avoiding the "combinatorial explosion" occurred.The model and the proposed method is applied to a coal-fired thermal system fault diagnosis, which overcome the complexity structure of the thermal system and causation, slow speed of diagnosis and difficulty to distinguish the possibility of the potential failure sources. The hierarchical method shortens the time of fault diagnosis, meanwhile, it reduces the loss of property caused by thermal system downtime. The quantitative information describes the possibility of each failure source. It reduces the fault repair time, improves the utilization of human and material resources and increases the safety of the thermal system.