A Stochastic Dynamic Data-Driven Framework for Real-Time Prediction of Materials Damage in Composites

This chapter develops methods and computational infrastructures that comprise a Dynamic Data-Driven Applications Systems (DDDAS) framework for real-time monitoring and prediction of the behavior of complex physical systems. The developed DDDAS framework for materials damage prediction enables har-nessing imaging data acquired in real-time to train physics-based models of system evolution, allowing computational prediction and dynamic control of the physical system. The targeted physical phenomenon is the failure and damage in composite materials common to modern aircraft. The observational data are obtained from mechanical experiments conducted on laboratory-manufactured carbon nanotube-infused epoxy composites. A family of material models, based on continuum damage mechanics, is employed for simulating the evolution of damage in the composites. The material models are then set in a Bayesian framework for cali-bration, validation, and model selection against a set of observational data while assessing the inherent uncertainties in the data, the model, and the numerical solution. A stochastic DDDAS computational infrastructure is implemented, based on a Bayesian filtering algorithm, which integrates the validated damage model with thedynamicdata,monitorstheevolutionofadamagefieldthroughoutthecomposite specimen, and enables forecasting the failure in a system. The outcomes of this study suggest that not only is DDDAS a beneficial and feasible approach, but it also describes a powerful new technology for developing predictive physics-based models of complex physical phenomena.