| As the core component of the hydraulic servo system, electro-hydraulic servo valve is the focus of fault diagnosis. The cause of the fault reasons also appears in uncertain, nonlinear and other complex states. This Paper describes the composition of electro-hydraulic servo valve, characteristic curves,major fault and the fault mechanism, as well as several methods of artificial intelligence fault diagnosis, including expert systems,fuzzy,gray theory,artificial neural networks and so on. We respectively analysis their good and bad points .By the comparison on them, finally I decide to use artificial neural network for fault diagnosis of hydraulic servo valve.I take an experiment on the nozzle flapper electro-hydraulic servo valve in our laboratory. I respectively set the system pressure at 3MPa, 3.5MPa, 4MPa,4.5MPa, 5MPa and set the servo in 5 fault status, which respectively are: 1) normal state; 2) spool in the limit of the end; 3) fixed orifice fouled 4) spool wear; 5) valve not in the zero. Then we get its pressure characteristic curves. Through analysis and quantification on these curves, we get a large number of data which provide samples for the following diagnosis.Then we theoretically study knowledge of B-P neural network of a single hidden layer, which belongs to the artificial neural networks. We design a B-P neural network which contains 32 input neural units,5 output neural units,1 hidden layer unit, variable number of hidden layer unit. We use it for fault diagnosis. Through the training of a large number of samples, the network adjusts the network weights, thresholds and other parameters to meet the training requirements. These parameters will be stored in the computer. At last, we input a set of normalized data to the trained network. The data come from the pressure characteristic curves when the system pressure is 3.5MPa. The results indicate that the outlet errors are within the defined limits, which verifies the correctness of the network.
|
PDF Full Text Request