Estimation of local heat transfer coefficient in hybrid electronic circuits - method comparison

The heat transfer coefficient value strongly depends on surface temperature rise and cooling fluid velocity. Thus, taking into account that usually large temperature gradients occur in air cooled electronic circuits, the local values of this coefficient might differ substantially throughout circuit outer surfaces. Consequently, the assumption of heat transfer coefficient uniformity for the entire circuit might lead to significant simulation errors. This problem was illustrated in this paper based on the example of a real hybrid circuit containing a transistor heat source and several temperature sensors. In the experiments, the heat source and sensor temperature values were measured in a wind tunnel at different power dissipation levels both for natural convection cooling and with variable cooling air velocity. Then, a two-layer thermal model of the circuit was created and solved using the Green’s function method. The local heat transfer coefficient values were estimated in the heat source and temperature sensor locations by minimizing the error between the measurement and simulation results. The estimation was carried out employing different numerical algorithms. Their effectiveness was evaluated and compared using the execution time and the number of iteration steps required to attain desired simulation accuracy as performance indicators. Moreover, the experiments rendered possible the fitting of empirical model parameters and hence allowing the computation of heat transfer coefficient values for different surface temperature rise values and cooling air velocities. The fitting was carried out using a modified bee colony optimisation algorithm.