In order to solve the difficulty of recognition in analog circuit fault diagnosis, under the two aspects of analog circuit fault feature extraction and fault pattern recognition, combined with the respective characteristic of liquid state machine, this paper present a new analog circuit fault diagnosis method based on liquid state machine. Firstly, the soft of MATLAB and PSpice is used to obtain the large number fault sample data automatically, and then the fault mode is classified by liquid state machine. Simulation show that the accuracy rate of fault identification can be reduced compared with the most widely used BP neural network. However, the training time of this method is much less than that of BP neural network, and the generalization ability is strong. It has a certain practical significance for the fault diagnosis of analog circuits.