Effective use of simulation requires confidence the simulation represents reality well. Challenges to achieving this confidence include model form and parameter uncertainty as well as discretization error. Further, simulations are typically deterministic while the real world they are modeling is stochastic in nature. It is therefore important to understand how real-world uncertainties affect simulation results.
Understanding and accounting for the degree to which a simulation represents reality is the domain of verification, validation, and uncertainty quantification (VVUQ) and helps modelers make statements about the degree of credibility they have in their results, the probability of specific outcomes, and the risk associated with decisions and scenarios.
This course provides an introduction to VVUQ for engineering simulation with a focus on the validation and uncertainty quantification aspects. Analyses that can support VVUQ efforts including sensitivity analysis, uncertainty propagation, and model calibration will be covered. The use of these analyses will be tied back to important guides and standards documents such as NASA STD 7009A and ASME VVUQ 10, 20, and 40. Emphasis will be placed on the use of a surrogate modeling approach, whereby a machine learning model is trained of the simulation to make such analyses more efficient to perform.
Request More Information