A model-based prognostics framework to predict fatigue damage evolution and reliability in composites

In this work, a model-based prognostics methodology is proposed to predict the remaining useful life (RUL) of composite materials under fatigue loads. To this end, degradation phenomena such as stiffness reduction and increase in matrix micro-cracks density are predicted by connecting microscale and macro-scale damage models in a Bayesian filtering framework. The proposed Bayesian filtering framework also allows incorporating various uncertainties in the prediction that are generally associated with material defects, sensing and monitoring noise, modeling errors, etc., to name a few. This, however, results in a explosion of search space due to high dimensionality, and hence a high computational complexity not conducive for real-time monitoring and prediction. To reduce the dimensionality of the problem without significantly compromising on prediction performance (precision and accuracy), a model tuning is first carried out by means of a Global Sensitivity Analysis. This allows identifying and subsequently down selecting the parameters for online adaptation that affect prediction performance the most. Resulting RUL estimates are then used to compute a timevariant reliability index for composite materials under fatigue stress. The approach is demonstrated on data collected from run-to-failure tension-tension fatigue experiments measuring the evolution of fatigue damage in CRFP cross-ply laminates. Micro-cracks are considered as the primary internal damage mode that are estimated from measurements obtained by active interrogation using PZT sensors. Results are presented and discussed for the prediction of growth in micro-cracks…

Fecha de publicación