Thermal-electric coupling systems analysis of electronic cabinets with consideration of numerical noises

One of the difficulties in multidisciplinary design optimization for electronic cabinets is that the design functions are often not smooth or even discontinuous due to numerical noises. In order to solve this problem, a global optimization strategy for multidisciplinary systems was developed. Firstly, the system analysis problem was transformed into an optimization problem according to the idea of SAND (simultaneous analysis and design). Kriging models were introduced as surrogate models for the output variables of each subsystem in order to filter the numerical noises. The initial Kriging models were built by using sparse sample points. The location of next samples was determined by an “infill sampling criterion” which was derived by the maximum likelihood estimation. When the simulation tools exhibit large numerical noises, this method can greatly reduce the number of subsystem simulations needed in system analysis compared with other methods, such as the fixed point iteration method. Lastly, a typical thermal-electric coupling problem was taken as an example for demonstration of the effectiveness of the proposed method.