Prediction of residual fatigue life from NDE of corroded components.

摘要

A study was performed to determine how effectively the structural integrity of corroded aircraft components may be predicted from nondestructive evaluation (NDE) of test specimens. Al 2024 specimens were corroded under a wide range of conditions to achieve a range of corrosion morphologies, including a mix of pitting, intergranular, and general corrosion. The specimens were then examined using ultrasonic testing (UT), white light interference microscopy (WLIM), and microradiography to quantify the corrosion damage level and provide inputs to a life prediction model. Image analysis was used to extract metrics to describe the damage state. Finally, the corroded specimens were fatigued to failure. The resulting data was used to obtain a correlation between the NDE metrics and the residual fatigue life. This was accomplished by both multiple linear regression (MLR) and artificial neural networks (ANNs).; A good correlation was achieved by MLR. ANNs also fit the experimental data well, with a slightly lower mean square error. However, for the more heavily damaged specimens, the ANN systematically over-predicted the residual life. Both the MLR model and the ANN model indicate a similar dependence of the fatigue life on the NDE metrics. While there are several important metrics that can be extracted from UT and WLIM, the correlation with microradiography metrics was weak.; The focus of this work is to provide an empirical methodology for quantifying the effects of corrosion damage on fatigue life in situations where deterministic or analytical approaches, such as those based on linear elastic fracture mechanics are not well suited.