In the paper some methods are presented, with the corresponding practical examples, related to MonteCarlo (MC) techniques for thermal model/test correlation purposes.
The MonteCarlo techniques applied to model correlation are intended to be used as an alternative to empirical ‘manual’ correlation techniques, gradients methods, matrix methods based on least square fit minimization.
First of all, Design Of Experiments (DoE) tools are used to determine the model response to uncertain parameters and the confidence level of such a response. A sensitivity map is built, allowing the design of the test to maximize the response of the system to the uncertain parameters.
Techniques derived from the extreme statistics are used to extrapolate data beyond test limits, with a sufficient confidence in the queue behaviour.
Finally the tool of the random walk is used for the model correlation, and an original ‘random waltz’ is introduced, being the combination of Latin hypercube sampling and pure random walk.