Thermal parameter estimation in the presence of uncertainty

Thermal properties are inferred from experiments by inverse techniques. In contrast to most mechanical and electrical experiments, thermal experiments usually involve boundary conditions and properties that have a degree of uncertainty. This uncertainty complicates the analysis and may lead to increased errors of estimation. This paper describes the application of an extended Maximum Likelihood principle to better quantify the amount of information yielded by the experiment and to reduce the effect of the uncertainty. The method is applied to an example of determining the contact resistance between a die and a substrate to demonstrate its use.