Multiscale probabilistic models for manufacturing, performance and failure of composites

Roger Ghanem
University of Southern California (USC)
Viterbi School of Engineering

Composite materials consist of components that provide complementary functionalities under different regimes of behavior. The performance of these components within the composite material system is determined largely during a manufacturing process which includes a combination of mechanical and thermal processes. Uncertainty about environmental and boundary conditions of the manufacturing process is manifested, through the associated physical processes, as spatially-indexed uncertainty about behavior at all scales. This uncertainty is present in experimental observations of both the manufacturing process and composite behavior. Associated numerical predictions are further challenged by modeling errors within each scale as well as while matching scales (multiscale) and physical phenomena (multiphysics). This presentation will present an overview of these uncertainties and current attempt to tackle them using polynomial chaos expansions. Numerical, conceptual, and experimental challenges relevant to credible predictions and qualification will be discussed.

This work includes significant contributions by Dr. Loujaine Mehrez and Mr. Ziad Ghauch of USC. Further, this work is the result of a collaboration with General Motors Research., and contributions from Venkat Aitharaju and William Rodgers at GMR are also acknowledged.

Back to Workshop IV: Uncertainty Quantification for Stochastic Systems and Applications