An approach to thermal placement in power electronics using neural networks

The author deals with the optimum thermal placement of heat sources in hybrid power circuits. The combination of the original heuristic method and the Hopfield neural net is presented. Because of the introduced hierarchical approach, a significant reduction of the computational effort is achieved. In this case, before starting the optimization procedure the number of heat sources is reduced by grouping some of them into new imaginary sources. It causes a decrease of the number of processing elements and synapses in the neural net that results in finding the solution in a relatively short time.<<ETX>>