A discrete adjoint-based multi-component coupled optimization framework for incompressible flows in turbomachinery

This paper presents a discrete adjoint-based optimization framework for the incompressible flow in turbomachinery. The entire computational framework is developed within the open-source DAFoam platform, which enables steady flow simulations of rotor-stator systems by coupling the rotating and stationary domains through a mixing-plane boundary condition based on the multiple reference frame approach. The framework employs an adjoint method for efficient gradient evaluation in high-dimensional design spaces, significantly facilitating the rapid identification of optimal solutions in multi-parameter optimization. The proposed methodology is rigorously validated against two canonical incompressible turbomachinery flow configurations: (i) the GAMM Francis turbine, and (ii) the high Reynolds number axial-flow pump. The computational results demonstrate high-fidelity gradient evaluation, as verified through finite-difference methods, and achieve significant improvements in hydraulic efficiency (2.86% and 2.43%, respectively) while maintaining all design constraints.