The use of deterministic models to predict the behavior of systems is currently a common practice adopted in engineering. However, these predictions always present a certain level of uncertainties related mainly to (1) the uncertain nature of excitations, (2) simplifications in mathematical models, and (3) the lack of information about model parameters. These uncertainties must be considered if more robust predictions are required, especially to identify optimal designs or improve the prognosis of systems already in operation.
The characterization of uncertainties by means of a stochastic framework allows performing analyzes aimed at quantifying the probabilities of having a specific performance (estimates of failure probability and risk assessment), optimizing performance variables (risk minimization), or aimed at improving the prediction of systems with data monitoring (implementations in the monitoring of structures). Motivated by the importance of conducting this type of analysis, the objective is to contribute to the mechanical design of systems and devices through the development of efficient computational methodologies for the quantification of uncertainties, based mainly on the use of stochastic simulation, Bayesian analysis, and implementation of metamodels.
This group is formed in the Department of Mechanical Engineering of the University of Michigan-Dearborn to serve as a platform to generate interdisciplinary collaborations and disseminate progress in the field.