Using adjoint CFD to quantify the impact of manufacturing variations on a heavy duty turbine vane

TL;DR

The study develops and validates an adjoint CFD workflow to quantify how manufacturing variations affect heavy-duty turbine vane aerodynamics. It compares hand-derived and algorithmic-differentiated adjoint solvers on a subsonic 1.5-stage turbine test case, finding the algorithmic-differentiated version more accurate. The workflow is then applied to 102 optical white-light scans of a turbine vane using mesh morphing to model variations, with results showing linear behavior for MV impacts and close agreement with finite differences for mass flow (though some discrepancies for pressure loss due to morphing artifacts). Efficiency gains are substantial, achieving roughly a 28-fold reduction in computational cost compared with evaluating many full RANS cases, enabling rapid MV analysis for design optimization.

Abstract

We consider the evaluation of manufacturing variations to the aerodynamic performace of turbine vanes using the adjoint method. The empirical data is based on 102 white light scans from casted parts. We compare expensive calculations by the finite disfference method with cheap adjoint calculations and we find high correlations.

Figures (16)

• Figure 1: Adjoint-based Toolchain for Impact Evaluation
• Figure 2: Turbine and mesh data
• Figure 3: 2D schematic not in scale
• Figure 5: Stagger Angle
• Figure 6: Maximum Blade Thickness