Vibration-based prediction of residual fatigue life for composite laminates through frequency measurements

Abstract

Fiber reinforced polymer (FRP) structures may experience cumulative fatigue damage during their service life which can lead to structural failure. This paper focuses on predicting the residual fatigue life of FRP structures using vibration parameters. The relationship between residual fatigue life and natural frequencies is examined through modal testing and fatigue measurements on FRP beam specimens. Two prediction methods; semi-empirical models and machine learning (ML) algorithms, are utilized. The semi-empirical models are derived from existing “residual stiffness” models based on the relationship between bending stiffness and flexural frequencies. ML algorithms Support Vector Machine (SVM) and Artificial Neural Network (ANN), are developed for fatigue life prediction. Experimental validation is performed using measured frequencies during fatigue testing of FRP beams. The ML algorithms can use multimode frequencies unlike single mode of semi-empirical models. The verification results show that the ML algorithms can be used to predict the residual fatigue life with the selection of the appropriate mode of frequency. The results show that ML algorithms outperform single-mode frequency inputs, and the use of higher modes of measured frequencies improves the precision of fatigue life prediction. An inverse algorithm based on SVM exhibits higher prediction accuracy and stability, even with limited training samples.