Study Of Fault Diagnosis Methord Using Data Fusion Integrated ANN And Evidence Theory

Hydraulic system has been extensively used by its merits such as high power-mass ratio in engineering area and it plays a principal role in the control and transmission of many equipments. It is very important to diagnose the fault of hydraulic system in time and accurately in enhancing productivity, decreasing the times of stopping machine and reducing the loss. The pump is the "heart" of the hydraulic system and has the severest working condition. So the fault diagnosis of the pump becomes the crucial factor for the hydraulic system.This paper systemically analyses every fault diagnosis methods of the hydraulic pump and comes to conclusion that traditional fault diagnosis methods, simple diagnosis methods and diagnosis methods based on signal processing and modeling have big defect. And the intelligent diagnosis technology becomes the trend of diagnosis technology now. But the diagnosis accuracy of intelligent diagnosis needs enhancing. Data fusion technology, which develops rapidly nowadays, can make the best of redundancy and complementary information of data from every kind of sources, and can heighten the accuracy and robustness of decision-making system. It provides an available method to get more accuracy fault diagnosis result of the pump.This paper first introduces the basic principle and concept of data fusion, summarizes levels and model structure of data fusion, summarizes common algorithms and sorts them, separately expounds data fusion algorithms, neural network and evidence theory, and lucubrates the structure design of neural network (BP network). And then, the advantages and disadvantages of neural network and the evidence theory are discussed. In order to colligate the two algorithms' strongpoint and increase the fault diagnosis accuracy, fault diagnosis method using data fusion integrated Ann and evidence theory is presented and diagnosis approach of this method is expatiated on. To validate fault diagnosis method using data fusion integrated ANN and evidence theory in pump fault diagnosis, this paper analyzes the common fault mechanism of piston pump, brings forward diagnosis domain, gives diagnosismodel and constructs architecture of diagnosis networks. At last, this method is applied to MCY14-1B type axial piston pump of hydraulic servo system of material experiment machine. Simulative shoe-doffing fault of this pump is diagnosised and this method is validated.