Fault diagnosis of hydraulic pump based on artificial neural network

Artificial neural network (ANN) was adopted to solve the problem of multi-failures diagnosis of the hydraulic pump. The failure modes and mechanisms for the hydraulic pump were analyzed. The multi-failures diagnosis strategy by using integrated back propagational (BP) ANN was proposed. The diagnosis system was decomposed into a control network and three typical failure sub-networks. The training of the control network and sub-networks was carried out respectively, using BP learning algorithm to decrease training time and to reduce scale of the networks. An illustrative example was given. The results have indicated that diagnosis of failure modes of the hydraulic pump using BP ANN is effective and the rate of failure identification increases as the learning sampling gets larger.