Using COMSOL to Estimate the Heat Losses of Composite Panels Undergoing Repairs Using Bayesian Inference

A. Emery[1]
[1]University of Washington, Seattle, WA, USA

Composite repairs involve applying a prescribed heat flux which must be adjusted to account for heat losses. These losses must be estimated. Parameter estimation requires a knowledge of sensitivities. This is often done using finite differences. As shown in the figure, substantial errors often occur. The paper describes the use of the PDE features of COMSOL to obtain much more accurate values by eliminating the errors due to finite differencing. The sensitivities are then used in conjunction with Bayesian Inference to compute the maximum a posterior probable values of the heat flux along with estimates of the confidence intervals.

Download