Aerodynamic optimization method for compressors based on adaptive design space reconfiguration

Conventional parameterized optimization approaches of compressors exhibit strong subjectivity in setting control parameter ranges, which may lead to irrational design space structures and compromise optimization efficiency. To address this challenge, this paper proposes an adaptive design space reconfiguration (ADSR) strategy. The ADSR starts with a small design variable range to simplify the search space. During optimization, it adaptively adjusts boundaries, empowering the algorithm to achieve real-time design space adjustment and exploration. By integrating directly manipulated free-form deformation (DFFD) parameterization with a Bayesian optimization algorithm, an adaptive global aerodynamic optimization platform for compressors is developed. Taking the transonic compressor rotor ROTOR37 as the test case, this study investigates the influence of DFFD basis-function selection and initial design-space size on the performance of the ADSR. Results demonstrate that, compared to conventional optimization methods, the ADSR achieves a maximum adiabatic efficiency improvement of 1.78% within the same iteration number, while avoiding surrogate model accuracy degradation caused by a large initial design space. As the initial design space decreases, the effect of the ADSR strategy gradually increases, and the B-spline-based DFFD parameterization achieves better optimization outcomes compared to the Bernstein-based DFFD parameterization. With an initial variable range of [−0.5 mm, 0.5 mm], the B-spline-based DFFD method achieved optimal adaptive optimization performance, demonstrating a 2.56% improvement in design-point adiabatic efficiency and a 2.9% increase in surge margin compared to the original compressor. This study confirms that the ADSR strategy effectively resolves the conflict between deterministic design space definition and uncertain optimal solution distribution in traditional parametric optimization, significantly enhancing optimization efficiency.