DEMONSTRATION OF A MODELING ERROR EXTRAPOLATION METHOD IN VERIFICATION, VALIDATION AND UNCERTAINTY QUANTIFICATION OF A STRESS CONCENTRATION IN A CANTILEVER BEAM

摘要

In order to assert that a complex simulation is an adequate representation of the real world the computational model must be validated against experimental results. Typically this validation is conducted at a small number of specified conditions, and set points. The simulation will differ from the experimental result due to experimental, numerical and simulation input uncertainties as well as un-modeled physics. The latter contributor is the modeling error. At these validation set points the modeling error can be estimated to fall within an interval. This range is the simulation minus experiment, plus/minus the combined experimental, numerical and simulation input uncertainties [1]. However, the purpose of creating a simulation is to predict results at set points where experimental results are unavailable. The uncertainty in this prediction is a combination of numerical and simulation input uncertainties along with the modeling error. In order to specify a prediction uncertainty it is necessary to extrapolate the modeling error interval to these prediction set points.