Simulation models are widely used to predict how products or systems will perform, but their accuracy depends on the quality of the input data. When some input values, such as material properties or operating conditions, are uncertain, the simulation results can also vary. To address this, COMSOL has introduced inverse uncertainty quantification, a workflow that estimates unknown input parameters by comparing simulation results with experimental measurements.
The approach uses Bayesian inference to update the parameter values based on the observed data and can employ surrogate models to reduce the need for repeated simulations. The resulting parameter estimates can then be used in future simulation studies to evaluate uncertainty using inputs that are informed by both prior knowledge and measured data.
Image courtesy: COMSOL