Residual Life Prediction Of Remanufactured Components Based On Magnetoacoustic Fusion

With the rapid development of social economy,a large number of waste products eliminated every year in the world not only lead to the waste of resources,but also cause environmental pollution.Therefore,it is very important to make full use of the surplus value.Remanufacturing engineering is a new industry which can prolong the remaining service life of scrapped products by repairing or transforming them.The technology of nondestructive testing and life prediction plays a key role in the remanufacturability evaluation of remanufactured blank and whether the remanufactured products meet the production and use standards.The purpose of this paper is to evaluate the residual life of the remanufactured component based on non-destructive testing and data fusion technology,and then to determine the remanufacturability of the blank.Based on tension-tension fatigue test,metal magnetic memory(MMM)and acoustic emission(AE)monitoring system was used to monitor the key parts in the fatigue cycle.The experimental results show that the variation trend of magnetic memory and acoustic emission signals at each monitoring point shows three obvious stages,which correspond to the stages of crack initiation,steady propagation and unstable propagation during fatigue damage evolution.Three characteristic parameters,namely synthetic magnetic signal,normal amplitude difference and normal deviation,were extracted by analyzing the variation rule of magnetic memory signal during fatigue cycle.Through energy spectrum analysis of acoustic emission monitoring data by wavelet packet,the high-frequency energy ratio of acoustic emission signals at different life stages is obtained.Through wavelet decomposition and reconstruction of monitoring signals,the information entropy of reconstructed signals is extracted to obtain the wavelet entropy of AE signals.Finally,four characteristic parameters,such as AE amplitude,energy,high-frequency energy ratio and wavelet entropy,were determined.The BP neural network was used to fuse the characteristic parameters of magnetic memory and sound emission,and the probability distribution of specimen life was fuzzy assigned according to the normal distribution function,and the probability of each section of specimen life under the single nondestructive testing method of magnetic memory and sound emission was predicted respectively.On this basis,based on Bayes theory,the index layer fusion is performed for the basic probability distribution of the two magnetic memory and acoustic emission methods,and the probability distribution results of each life interval after fusion are obtained.The prediction accuracy of the result is higher than that of the single method.Based on D-S evidence theory,the results of magnetic memory monitoring and acoustic emission monitoring are fused to obtain more accurate prediction results than Bayes method.Weighted fusion algorithm based on D-S evidence presented optimization method,with the introduction of weight coefficient,D-S evidence theory,the basic probability allocation value,then the revised data fusion,the basic probability allocation value to further improve the prediction accuracy,the results show that the precision of fuzzy probability distribution than the original increased by nearly 10%.