Research On Uncertain Problems In Motor Fault Diagnosis Based On Bayesian Network

Motor is a complex power system, whcih results in the complexity and uncertainty in motor fault diagnosis. For this uncertainty existing, it's difficult to set up a qualitative model for fault diagnosis. Now in the field of fault diagnosis, the uncertain problems becomes the chief problem to solve, and the Bayesian networks has become the hotspot because its most efficient theory model for solving uncertain problems and expressing uncertain information.According to the uncertain problems in motor fault diagnosis, this paper has a deep research in comparing fault diagnosis method, analysising fault pattern and causation, Bayesian networks modeling and fault diagnosis illation. The main jobs are followed:Analyse various fault problems technology and methods for solving uncertain problems. Through comparison, Bayesian network is comfirmed on theory. This paper has researched Bayesian network theory, including Bayesian network character, Bayesian network structure learning and parameter learning and Bayesian network algorithm.According to the fault mode and causation, the Bayesian networks model of fault diagnosis is set up, and the netwok structure constructing, knowledge expression are researched deeply.For the feature of Bayesian network structure learning with a high complexity, a optimize algorithm of network structure leaning based on cluster theory is put forward in the paper. This arithmetic make an optimizing search dependding on the original network structure, and combine the transcendental information to search the optimizing network topological structure. The paper introduces the theory of this algorithm, and the diagnosis algorithm based on this algorithm, also analyses the capability of the algorithm through network information error and algorithm complexity. Through the comparing, the result proves that this improved Bayesian network fault diagnosis algorithm has a high advance.A fault diagnosis system based on Bayesian network is set up, and the system is carried out with C#, including network modeling, structure learning, diagnosis illation and decision-making. Through the experimentation and analysis, this method in the paper can depress the algorithm intricacy effectively, and get an upper diagnosis accuracy rate. All of this prove that this fault diagnosis model and algorithm can solve the uncertain problems effectively in motor fault diagnosis.