Fairing shape optimization in engine bypass duct based on random forest and curvature-based deformation

This paper proposes a novel drag reduction optimization strategy based on curvature-based free form deformation (CFFD) and random forest (RF) with small-scale training datasets. The strategy includes both deterministic and robust optimization. The objective is to determine a new fairing shape that minimizes total pressure loss in an engine bypass duct. An improved CFFD method is proposed to parameterize shapes and generate new designs. RF serves as a surrogate model, with infill samples generated by minimizing the predictor infill criterion to enhance its performance. The novel strategy demonstrates strong performance in both deterministic and robust optimization. Compared to the base shape, the total pressure loss in the engine bypass duct with the new fairing shape is reduced by 27% and 25%, respectively. The RF model exhibits better predictive performance in the case of small-scale datasets relative to traditional Kriging model. The mean absolute prediction error is reduced by 77%. Extensive testing also reveals that the novel optimization strategy demonstrates broad applicability and strong potential for practical use. The predicted mean relative errors of shape pressure drag, shape wall friction drag, and duct wall friction drag are 8.96%, 9.54%, and 0.07%, respectively.