Session: 36-01 Adjoint-based approaches - Part 1

Submission Number: 174437

Adjoint Stress Optimization of a Radial Compressor 

According to the Eulerian Turbomachinery equation, radial compressor impellers with axial inlet flow do specific work either through blade speed or circumferential flow velocity at outlet. From an aerodynamic perspective small flow velocities are desirable as pressure losses scale with velocity squared. Hence, to maximize efficiency, we strive for high blade speeds at outlet which comes at the cost of increased fatigue. Switching from aluminum to other materials, e.g. titanium, is possible yet costly.

Thus, in the paper, we will present a study that aims at mitigating this by reducing stress through an optimization of the rear side of the impeller. One key challenge in this is to maintain manufactureability throughout the shape update: We have to maintain rotational symmetry. This is not the natural habitat of adjoint-based optimization workflows that tend to yield chaotic sensitivities and shape updates, respectively. Nonetheless, this class of optimizers comes at the benefit of being extremely quick and exploiting a high level of design flexibility at the same time.

Hence, we intend to present how the adjoint sensitivities w.r.t. stress need to be transformed and smoothened in order to deliver a succesful, stable and meaningful morphing that allows to maintain the current manufacturing and machining process of the radial compressor at hand.

On top, we will study an additional question using the same algorithmic setup: The axial length of the impeller body plays a crucial role in terms of stress, weight and center of gravity. We will investigate how different axial lengths affect these key indicators by starting individual optimizations for different starting geometries (with different lengths) to develop an understanding of the trade-off.

Finally, we ask and answer how second order tetrahedrals, that come with higher computational cost than first order elements, affect the optimization result.

Presenting Author: Nicolas Lachenmaier Rolls-Royce Solutions GmbH

Presenting Author Biography: 06/2014-today: CFD engineer at Rolls-Royce Solutions GmbH focusing on turbocharger development and optimization

01/2023-today: Lecturer for courses in Aerodynamics, CFD and Turbocharging of Combustion Engines (DHBW Ravensburg / Karlsruhe Institute for Technology)

07/2016-07/2022: PhD on “Multidisciplinary gradient-based optimization of radial turbines”
TU Darmstadt, Institute of Gas Turbines and Aerospace Propulsion

10/2011-03/2014: Master of Science with Honors
Computational Engineering
Friedrich-Alexander-Universität Erlangen-Nürnberg

04/2010-11/2015: Bachelor of Science
Mathematics
FernUniversität Hagen

09/2008-09/2011: Bachelor of Engineering
Business Administration and Engineering
DHBW Ravensburg Campus Friedrichshafen

Authors:

Nicolas Lachenmaier Rolls-Royce Solutions GmbH
Elena Kolb Rolls-Royce Solutions GmbH