Research On Remaining Life Prediction Of Parts In Use Based On Magnetic Memory Testing

In order to satisfy the engineering requirement of detecting and remaining life assessment for remanufacturing components, this study focuses on prediction theory and method of remaining life for parts in use based on magnetic memory testing. It takes automobile drive axle housing as the object, according to the results of theoretical derivation, simulation analysis and experiment, the parameters affecting the remaining life of parts are analyzed, and a novel method predicting its residual life is presented. The work of this thesis mainly includes the following four aspects:Firstly, the fatigue dangerous zones are identified based on the simulation analysis and calculation under ABAQUS and FE-SAFE, and the simulation results are compared with residual life results of fatigue testing. It’s found that the simulation results reflect the fatigue dangerous zones accurately, which identifies that the proposed simulation method is reliable and correct.Secondly, fatigue dangerous zones are set as the research objects, and the crack length, stress intensity factor, gradient maximum of magnetic memory signal and degree of stress concentration are extracted and serve as the characteristic parameters, by means of metal magnetic memory testing technique and fracture mechanics theory. Support vector machine (SVM) method is studied because of its outstanding performance in aspects such as needing small samples and showing high prediction precision, and the remaining life prediction model can be established through training and testing the characteristic parameters. It shows that SVM prediction model improves the prediction accuracy, and the research results error is less than10%compared with residual life results in fatigue testing. Meanwhile, the accuracy of model output is related to the degree of damage for parts, the numbers of training samples, the amplitude of loading and the inputs of model and so on. In addition, the proposed method can effectively predict the remaining life of the significant equipment under high-cycle fatigue.Thirdly, based on the wavelet transform, more parameters should be proposed to characterize the damage of the dangerous zones to reduce the prediction error, and improve the accuracy of SVM model further. The magnetic memory signals of ferromagnetic parts detected when the fatigue cracks propagate are processed by wavelet threshold de-noising method. The detail components of de-noised signals are extracted by discrete wavelet transform (DWT), and the interval of regular signals is selected to characterize the damage of the dangerous zone. Reliable features of magnetic memory signals such as the amplitude difference and gradient maximum serve as the parameters, and the defect identification model is established based on signal feature extraction. It shows that the research results error is less than5%compared with the test result, which indicates that the new parameters proposed to characterize the damage of the dangerous zone are reliable.Finally, the remaining life prediction model is established based on the wavelet analysis result mentioned above, considering the common characteristic parameters of part II and the new combined parameters defined in part III. It shows that error between the model results and the fatigue testing result is less than1%, which proves the effectiveness of the new prediction model in improving the prediction accuracy obviously.