Uncertainty Quantification and Propagation in Structural Dynamics

In simulations of complex physical systems, uncertainties arise from the imperfections in the mathematical models introduced to represent the systems and its interactions with the environment. Such uncertainties lead to significant uncertainties in the predictions made using simulations. Since predictions form the basis for making decisions, the knowledge of these uncertainties is very important. This project takes on the challenge of Quantifying Uncertainties in complex structural Dynamics (UQ-Dynamics) simulations using monitored data collected from a sensor network strategically placed over the structure.

A comprehensive Bayesian probabilistic framework will be developed for (a) quantifying the uncertainties that arise in the selection of models of systems and external actions (e.g. loads) based on monitoring data, and (b) propagating the uncertainties through the computational models for the prediction of the uncertainties of response quantities that are representative of system performance, reliability and safety. Conceptual and computational challenges will be addressed for managing uncertainties in simulations. Fast and reliable high performance computing techniques will be developed to drastically reduce to manageable levels the excessive computational demands arising when treating complex systems. Model reduction techniques will substantially reduce the model order of substructures; automated multi-level sub-structuring techniques will speed up system analyses; and parallel computing algorithms will optimally distribute computing to GPUs & multi-core CPUs.

The developed framework will be applied to solve challenging structural health monitoring and residual lifetime prognosis problems. Attention will focus on critical structural maintenance problems, developing novel techniques to predict fatigue accumulation at unmeasured hotspot locations of a structure using sensors at a limited number of measured locations. A series of laboratory experiments and field tests on selected structures will validate and improve theoretical and computational developments.