For the thermal management of the electric products, the compact modeling method is commonly applied to the numerical analysis with the simplification of the each component used in the products, on the view point of the best thermal design. In the generation of the compact modeling, the method that the model parameters are optimized toward to the fitness between the temperature value of numerical analysis result and the actual hardware testing data, is conventionally used, with some monitored points which are given in advance. In this parameter optimization, the fitness function is to be 1 on the no (0) temperature difference between them at all monitored locations, and on the other hand, that is to be 0 if the temperature difference between them is infinite. However, it was found that this conventional method brought the fitness function shape of one sharp mountain, as the analysis result, and the method, that represented the fitness function with the quadratic polynomial, caused the important problem on the view point of the analysis quality if the Response Surface Method is used of the parameter optimization. Therefore in this research to resolve this problem, we suggest the method that the response surfaces, which are based on the heat conduction law, are composed for each monitored location, and the fitness function is given with these surfaces. This method is applied into some compact modeling and the benefit is verified. The meanwhile of the temperature difference between the numerical analysis result and the solution field is decreased half, and the divergence makes one-tenth decrease.
Reliability-Based Stability Analysis of a Baltic Cliff by the Combined Response Surface Method