Multi-objective optimization design of cooling fins for rotary engine

To design a cooling fin structure suitable for small air-cooled rotary engines, a method combining deep neural networks (DNN) and non-dominated sorting genetic algorithm II (NSGA-II) was employed for multi-objective optimization of fin parameters. Simulation experiments were conducted to obtain a sample dataset of various structural-performance parameters, which was then used to train the DNN prediction model. The optimization objectives were to minimize the average engine temperature, temperature distribution non-uniformity, and total engine weight. Multi-objective optimization was performed, resulting in optimized performance and corresponding structural parameters. Subsequently, the TOPSIS decision-making method, combined with different weight factor matrices, was applied to select five optimal compromise solutions from the Pareto front, and the most suitable solution was chosen based on design requirements. Finally, CFD simulations were conducted to validate the optimization results. Compared with the original engine without fins, the optimized solution (TOPSIS-A) achieves a reduction in maximum temperature of 108 K (19.6%) and a reduction in average temperature of 89.1 K (18.4%), along with a 45.9% improvement in temperature uniformity, at the cost of a mass increase of 394.7 g (21.9%). These results further demonstrate the effectiveness of using multi-objective optimization methods in guiding the design of cooling fins.