Prediction of remaining fatigue life of metal specimens using data-driven method based on acoustic emission signal

Aiming at the difficulty of predicting the remaining fatigue life of mechanical parts, a similar fatigue life prediction method based on acoustic emission signals is proposed considering the failure of vibration signals. Moreover, this paper innovatively introduces the temperature signal as a degradation feature to assist the acoustic emission signal feature for fatigue life prediction. The proposed fatigue life prediction method involves multiple processes such as feature extraction, feature smoothing, feature selection, feature compression and health index construction. A variational autoencoder structure combined with a long short-term memory neural network is used to achieve feature compression and retain feature trend. A tensile fatigue test bench was developed to collect the degradation signal from health to fatigue fracture to validate the proposed method. The validation results show that the proposed method can accurately predict the remaining fatigue life. In addition, the role of various data processing methods and the applicability of the proposed method in different working conditions are also discussed.