Research On Acoustic Emission Signal Diagnosis Method Of Axle Fatigue Crack Based On WT-DBN

Axles are a critical component for train operation,and due to their harsh working conditions,surface wear and even fatigue fractures are prone to occur.Therefore,it is necessary to prevent fatigue cracks in train axles during railway transportation.Acoustic emission technology is a new type of non-destructive testing technology that has been widely used in many fields of research.Studying the acoustic emission signal detection technology for train axles can help achieve an accurate grasp of the health condition of train axles.This article first analyzes the theoretical and experimental methods for diagnosing axle crack acoustic emission signals,including three aspects:1.parameter analysis,analyzing the characteristics of fatigue crack signals using the relationship between amplitude,energy,and duration;2.waveform analysis,using wavelets to transform and extract features of the three types of signals,focusing on eight feature parameters under the three types of signals;3.a method based on the combination of wavelet transform(WT)and deep belief network(DBN)to classify and identify the characteristics of the three types of signals and compare and analyze the identification results with the DBN model,wavelet,and one-dimensional convolutional neural network(CNN)combined method.The results show that the WT-DBN-based method has a higher diagnostic accuracy.Then,to address the issue of how different frequencies of acoustic emission waves affect the accuracy of axle fault diagnosis,an experiment was conducted on the attenuation of acoustic emission waves based on the WT-DBN network proposed.The experimental results verify the reliability of the network proposed in fault diagnosis and indicate that the acoustic emission waves gradually attenuate with the increasing distance between the sensor and the sound source,and the diagnostic accuracy of the network gradually decreases.Finally,a monitoring system for acoustic emission signals of axle fatigue cracks based on WT-DBN was developed,including an acquisition module,model selection module,database display module,waveform display module,and result display module,which can realize the acquisition and analysis of acoustic emission signals of axle fatigue cracks.In conclusion,based on the comprehensive experimental results and analysis,the fatigue crack diagnosis model based on WT-DBN can effectively improve the diagnostic accuracy and efficiency,has high reliability in fault diagnosis,and the position of the acoustic emission relative to the sound source has a significant impact on feature extraction and fault recognition.To further improve the detection accuracy and reliability of acoustic emission signals of axle fatigue cracks,it is necessary to arrange the position of acoustic emission sensors reasonably.