Abstract
The present paper aims to optimize one of the highly loaded, compact, axial compressors investigated by NASA, using modern RANS-based optimization techniques. The NASA rotor 37 blade is used as a starting point and has been approximated by a CAD-based parametrization. Predictions by the RANS solver are compared with experimental and numerical data to validate the models. A gradient-based aero-optimization is performed to improve the performance on a single operating point while maintaining the same mass flow. The gradients are obtained through an adjoint approach where the cost is nearly independent of the number design variables. Particular attention is given to the quality of the computed sensitivities to reach a better convergence. A total of 53 design variables from the blade, hub, and shroud geometry are used to define the design space of the compressor. A 5.5% relative improvement in total-to-total efficiency is achieved by a Sequential Quadratic Programming optimization algorithm. While the efficiency is improved at a single operating point, the full operating map shows a narrow efficiency peak with a wider operating range toward the stall condition. Nonetheless, the overall performance of the optimized design is significantly improved compared to the baseline performance, with a limited computational effort.
