Fault Diagnosis For Condensers Based On Integrated Neural Network

As one of the im portant auxiliary equi pment in power plants’ operation, the condenser’ operation condition is related to the safe and economic operation of power units directly, condenser’s fault reflect fo r vacuum reduction ultimately. Therefore, detecting the condenser’s operating condition accurately, understanding the fault type, overhauling timely, keeping the condenser in a good running condition, ensure to achieve the best power vacuum is the important content to realize energy saving and emission reduction in power plant.The essence of fault diagnosis is a stat e recognition and classification, as th e normal operation of the condenser is af fected by many factors, relationship between fault symptoms and type is intricate and complex, therefore, important measures to improve productivity can be done by searchi ng for detection method that can isolate fault symptoms information and diagnos is malfunction hidden trouble timely and effectively.Based on the analysis of the study of domestic and foreign condenser state overhaul and fault diagnosis technology, this paper extensively studied on the failure appeared in the process of modern condenser operation and make a diagnosis timely.First, the condenser’s common failure types and reasons are analyzed. According to the working principle of condenser, result the common factors which influence condenser’s normal operation, focusing on c ondenser thermal performance relational expression, analysis th e relationship between each af fecting factors. Finally the corresponding relations between the fault symptoms and type are summarized.Secondly, the fault diagnosis method is put forward. Through the analysis on the common faults types and its reason, we learn that the congruent relationship between fault type and sign possess lar ge-dimensional, nonlinear and so on characteristics. In view of the good properties of neural network in dealing with nonlinear problem, the methods of integrated neural network classifier fusion were introduced to improve the diagnostic efficiency.Moreover, this paper introduces the com position and principle of the integration neural network emphatically. The BP and improved BP neural network’s essential diagnostic theory are analyzed, and th e principle and m ethod of the m ultiple diagnostic sub-network decision fusion are detailedly studied. The theoretical basis for integration neural network’s use in the condenser fault diagnosis is provided.Finally, the condenser fault state example is used to sim ulate experiment by MATLAB and is com pared with the tradit ional diagnosis m ethod, which verifies integration neural network plays very good diagnosis effect.