Multi-objective optimization design of scramjet nozzle based on grey wolf optimization algorithm and kernel extreme learning machine surrogate model

作者：Shuhong Tong, Mingming Guo, Ye Tian, Jialing Le, Dongqing Zhang, Hua Zhang · 发表于：Physics of Fluids · 年份：2024 · DOI：10.1063/5.0188627 · 被引用次数：31 · 研究领域：Machine Learning and ELM、Fault Detection and Control Systems、Nuclear reactor physics and engineering

This study delves into the parametric design of the scramjet nozzles, utilizing the Catmull–Rom curve, to meet the high-performance design requirements. It establishes a high-dimensional, multi-objective optimization design method based on a surrogate model for the nozzle. In addition, this research proposes a surrogate model for nozzle performance to enhance the accuracy of the traditional surrogate model and prevent the multi-objective optimization design method from optimizing in an incorrect direction. This model incorporates the grey wolf optimization (GWO) algorithm and kernel extreme learning machine (KELM). Various machine learning algorithms are compared and analyzed, demonstrating that the performance parameters predicted by the GWO-KELM model are the most accurate, and the generalization of GWO-KELM is verified. Utilizing the particle swarm optimization algorithm assisted by GWO-KELM, the multi-objective optimization of the nozzle is further investigated. This study obtains the optimal Pareto front, analyzes the distribution of design variables in the Pareto solution set, and reveals the impact of geometric parameters on nozzle performance. Comparing the representative nozzle from the Pareto front with the truncated maximum thrust nozzle, it is found that the thrust, lift, and outlet Mach number increase by 3.3%, 12.2%, and 0.5%, respectively, while the outlet height decreases by 5.3%. This research contributes to overcoming the limitations of traditional design me...