STATE RECOGNITION METHOD OF PISTON PUMP BASED ON ENERGY FEATURE AND RADIAL BASIS FUNCTION NEURAL NETWORK

The energy of vibration signal would vary in different frequency bands when pist on pump fault occurs.Therefore energy feature parameter extracted from vibratio n signal could be used to identify the state of piston pump.For this purpose,s tate recognition method of piston pump based on energy feature and radial basis function(RBF) neural network is put forward.Firstly,the vibration signal is de-noised by improved wavelet pocket threshold de-noising method.Then de-noise d signals are decomposed into a finite number of stationary intrinsic mode funct ions(IMFs) by empirical mode decomposition(EMD),and IMFs containing main state information are selected for further analysis.Energy feature parameter extract ed from IMFs could be served as input parameter of RBF neural networks to identi fy different states of piston pump.The application examples show that the appro ach of RBF neural network state recognition based on EMD extracting energy param eter of different frequency bands as feature is superior to that based on wavele t packet analysis and can identify piston state effectively.