Prediction of fatigue damage growth in notched composite laminates.

摘要

The damage growth of notched AS4/3501-6 graphite/epoxy quasi-isotropic laminates under long term mechanical fatigue loading is investigated. The effects of load type, load level, load sequence, and spectrum modification are evaluated in this study. X-ray radiography is taken of the test specimens to monitor the fatigue damage, which are in the form of splitting and delamination around the center hole of a specimen. Modifying the compression dominated loading spectrum by omitting the two lowest load levels has been found to have little effect on the propagation of damage while reducing testing time by 99.7%. Linear regression and artificial neural network models are developed to match the test results. Prediction models are developed solely on the basis of constant-amplitude fatigue test and applied in conjunction with a linear cumulative damage rule to predict damage growth, in this study split length, under spectrum loading. Several different artificial neural network structures are trained with different sets of data. Overall, the artificial neural networks prove to be highly effective in predicting the split length growth under fatigue loading even though they are trained on constant-amplitude test results.