An inverse aerodynamic design framework for compressor blades based on generative model

The aerodynamic design of compressors is critical to the overall performance of aeroengines. Traditional compressor design methods are typically categorized into forward and inverse design. Forward design relies on predefined geometric parameters and iteratively adjusts them to achieve target performance, while inverse design begins with the desired performance specifications and directly derives the blade geometry. However, these existing methods exhibit significant limitations, including heavy reliance on expert knowledge, high computational complexity, and limited robustness. This paper proposes a generative inverse design framework based on a diffusion-driven gradient optimization network to mitigate the drawbacks of traditional methods. By integrating the strong global exploration capability of diffusion models with the efficient local adjustment ability of gradient-based optimization methods, the framework overcomes the limitations of single neural network models in complex design tasks. It takes key aerodynamic performance of the compressor, such as mass flow rate, isentropic efficiency, and total pressure ratio, as inputs and outputs compressor blade design parameters that meet these objectives. The framework is validated through the design of three-dimensional transonic axial compressor rotor blades. Results indicate that the framework can rapidly generate multiple viable geometries that meet the specified design requirements, with a maximum relative error of less than 1.5% compared to the target performance metrics. This significantly improves design efficiency and reduces dependence on the expert knowledge of designers, offering a promising direction for the intelligent design of turbomachinery.