Research Of Fast Multi-objective Optimization Method For Typical Heat Dissipation Structure And Its Software Development

To solve the multi-objective optimization problems in heat sink design,more and more heuristic algorithms,such as genetic algorithm,particle swarm optimization,simulated annealing algorithm and differential evolution algorithm are used to optimize the heat dissipation performance of heat sink.Particle swarm optimization(PSO)algorithm is favored by scholars because of its simple calculation,fast convergence,less control parameters and strong robustness.However,these heuristic algorithms often require a lot of computation to reach convergence.To reduce the calculation,the commonly used method is to construct the surrogate model with a small amount of sample data,and to obtain the approximate solutions of the problem by optimizing the surrogate model.Although multi-objective optimization based on surrogate model can improve the efficiency of heuristic algorithm,high-precision surrogate model is difficult to be ensured or even can not be constructed due to the complexity of the practical problems,which results in that the approximate multiobjective optimization can not obtain the Pareto-optimal solutions consistent with the real model.In order to improve the efficiency of multi-objective optimization of heat sink,solve the problem of the heuristic algorithm with large amount of calculation and low optimization efficiency,and ensure the optimization result is the result of the real model,in this work the multi-objective particle swarm optimization method is combined with the circular crowded sorting strategy and parameter transfer strategy.Firstly,the combined method optimizes the coarse model and obtains the Paretooptimal solutions by multi-objective PSO.Then,the circular crowded sorting strategy is used to select some representative solutions that are uniformly distributed,and the population of the fine model can obtain the information of optimization results of the coarse model through the parameter transfer strategy.Finally,multi-objective PSO is used to optimize the fine model and obtain the Pareto-optimal solutions of the fine model.The coarse model can usually enter the stage of local search and reach convergence after optimization.Since the fine model is similar to the coarse model,the fine model’s population can quickly enter the stage of local search and reach convergence after parameter transfer strategy.To verify the performance of the suggested method,two examples have been carried out by using the multi-objective particle swarm optimization with crowding distance selection.According to the results,the efficiency of optimization procedure is improved significantly,and the calculation accuracy can be ensured.For the multi-objective problem of pin-fin heat sink,the optimization efficiency is obviously improved by the method.To further improve the computational efficiency,the proposed multi-objective particle swarm optimization with crowding distance selection has been improved in this work.After the initialization of the fine model,the particles corresponding to the response of the interested region are selected as the new population of the fine model,so that the calculation of the fine model in uninterested region can be reduced.Thus,less population is needed for optimizing the fine model and the calculation efficiency can be improved.The improved method makes effective use of the information of the coarse model’s result and the fine model’s initialization.Two examples have been carried out by using the improved method.According to the results,the calculation of the fine model is obviously reduced,the number of Pareto-optimal solutions uniformly distributed in the interested region is increased,and the computational efficiency is greatly improved.The improved method is applied to the multi-objective optimization of circular heat sink,and the optimization efficiency of heat sink design is further improved.Finally,a matlab/GUI-based optimization toolbox graphical interface of multi-objective particle swarm optimization with crowding distance selection is designed,which provides a simple and convenient operating platform for end users.