Fault diagnosis of pump-jack based on neural network

Abstract

Since the present backwardness of pump-jack fault diagnosis method and the waste of time and labor, we adopt wavelet network to transform the working current of pump-jacks and get the detail coefficient and then make it fault characteristics. To adjust the parameters of neural network, we adopt the self-tuning learning rate conjugated gradient method to optimize the object function. The application on thirty five pump-jacks indicates that this method can be used on the fault diagnosis of pump-jacks with the accuracy over ninety five percent.