Magnetoacoustic fusion life prediction method for retired components based on D-S evidence theory

Abstract

Prediction of the remaining life of remanufacturing blanks is crucial to evaluate their remanufacturability. To overcome the deficiency of obtaining insufficient fatigue damage characteristic information using a single non-destructive testing method, a new magnetoacoustic fusion life prediction method based on Dempster-Shafer (D-S) evidence theory optimised by a weighted fusion algorithm is proposed. The characteristic parameters of metal magnetic memory (MMM) and acoustic emission (AE) signals are first extracted on the basis of a fatigue experiment and data layer fusion is carried out to establish the mapping relationship between MMM and AE characteristic parameters and specimen life based on a back-propagation (BP) neural network. The basic probability distribution of the life is assigned in a fuzzy manner according to the normal distribution and the reliability function of each life interval is obtained by data fusion based on D-S evidence theory. Furthermore, the basic probability distribution value is modified based on a weighted fusion algorithm and the corrected data are fused to obtain a more accurate life prediction result.