Data-driven composite damage prognostics by coupling computational modeling with nondestructive evaluation

Carbon fiber reinforced polymer (CFRP) composite materials experience multiple stages of damage evolution from nucleation to catastrophic failure. Attempts have been made to observe this process experimentally and to obtain remaining useful life (RUL) predictions using computational models. Experimental observations are typically used to calibrate computational models that provide descriptions of the damage evolution process by mathematically accounting for the effects of specific damage mechanisms. Rarely, though, there is direct coupling between the experimental observations and the models beyond fitting of computational parameters or a posteri-ori comparisons. This practice makes the task of reliable prediction difficult when future data is unknown. This talk presents how nondestructive evaluation measurements of acoustic emission (AE) and digital image correlation (DIC) are used to investigate the initiation of damage mechanisms in composites and to track their evolution. Such information is then used in a post-processing scheme that involves mathematical and statistical methods that ultimately provide data-trends related to the evolving damage state of the material which is subsequently fed into computational models. The method is capable to refine and update such models as more data is obtained. The success of this data-driven coupling is assessed by the success of the model to estimate material properties as well as the remaining useful life at the specimen and component levels.