UC San Diego

Fatigue-induced damage is one of the most uncertain and extremely unpredictable failure mechanisms for a large variety of structural systems (e.g., aerospace, automotive, offshore, and civil structures) subjected to stochastic and cyclic loading during service life. Among these systems, composite lightweight aerospace structures -- such as fighter aircrafts and unmanned aerial vehicles (UAVs) -- are particularly sensitive to both fatigue- induced and impact-induced damage. Within this scenario, an integrated hardware & software system capable of (i) monitoring the critical components of these systems, (ii) periodically assessing their structural integrity, (iii) predicting their remaining fatigue life (damage prognosis), and (iv) accomplishing a cost-efficient condition-based maintenance (CBM) is ultimately needed. This research contributes to the aforementioned objectives by providing a novel and comprehensive probabilistic methodology for predicting the remaining fatigue life of adhesively-bonded joints in composite structures. According to this methodology, non-destructive evaluation (NDE) techniques and recursive Bayesian inference are repeatedly employed to update the probability distributions of damage extents and damage evolution model parameters at various damage locations after each NDE inspection. The propagation of damage is then stochastically simulated using a probabilistic model for future operational loads and a surrogate model (calibrated and validated at various damage levels using a mechanics-based model) capable of predicting the structural response quantities of interest. Finally, local and global failure criteria are considered simultaneously to compute the probabilities of failure and false-alarm at future times by abstracting the structure (or structural component) into a combination of series and parallel sub-systems. Three benchmark applications are provided in this work to exercise, verify, and validate the proposed framework. The first two benchmark applications analyze the fatigue-driven debonding propagation along a pre-defined adhesive interface in a simply supported laminated composite beam. They demonstrate the efficiency of the proposed recursive Bayesian inference scheme, show the use of the proposed component and system reliability analyses to recursively predict and update the evolution in time of the probabilities of failure and false-alarm of the structure, and illustrate the robustness of the framework. Finally, the third benchmark application validates the proposed damage prognosis methodology by using experimental fatigue test data obtained from the literature