The Identification of Fatigue Crack Acoustic Emission Signal of Axle based on Depth Belief Network

The axle is the key component of the running part of the railway vehicle, and it is the key component to hear the dynamic load of the vehicle. Due to its own factors and the bad working environment, it will often lead to cracks, surface damage, and even breakage when the axle is running. In this paper, the acoustic emission signal of fatigue crack of axle based on depth belief network is proposed which gives a new method to help identifying the fatigue crack online by acoustic emission detection technology. Firstly, the fatigue crack signal of axle is processed by using the time domain statistical characteristic parameters, and then the processed characteristic data was input into the depth belief network for learning and training. As a new type of mechanical learning intelligent network, deep belief network can realize the training calculation of multi-hidden layer, and multi-hidden layer can mine the feature of data more autonomously and deeply. Therefore, the accuracy of the results obtained from the operation is high.The experimental results show that the classification and recognition of fatigue crack signals obtained from acoustic emission experiments by depth belief network is very effective.