Session: 36-02 Adjoint-based approaches - Part 2

Submission Number: 177910

Adjoint-Based Multidisciplinary Optimization of a Centrifugal Compressor With Aerodynamic and Structural Objectives 

The design of turbomachinery components inherently involves trade-offs between aerodynamic performance, structural strength, and vibration resistance. Achieving an optimal and reliable configuration therefore requires concurrent consideration of these coupled disciplines within a single optimization framework. Gradient-based multidisciplinary optimization (MDO) approaches have become particularly attractive for such applications, as they enable efficient exploration of high-dimensional design spaces while ensuring strong coupling between disciplines.

This paper presents gradient-based multidisciplinary optimizations of the SRV2 radial compressor, where the geometry is parameterized using 128 CAD variables. The optimization aims to maximize aerodynamic efficiency while enforcing structural integrity and maintaining the target mass flow rate at two operating points of the compressor map. Both the fluid and solid domains are discretized with structured meshes to ensure high fidelity and numerical robustness. The structural domain is modeled using hexahedral finite elements, which provide superior accuracy in stress and vibration predictions. The gradients of both aerodynamic and structural cost functions are obtained through adjoint solvers. This approach allows for the efficient computation of sensitivities with respect to all 128 design parameters, at a computational cost independent of the design-space dimension. The adjoint-based framework thus enables the execution of large-scale multidisciplinary optimizations at reasonable computational expense.

A first optimization considers only aerodynamic objectives and constraints. The optimal design achieves a gain at peak efficiency exceeding 4% relative to the baseline configuration. However, structural analysis of the optimal configuration reveals a deterioration in mechanical performance: von Mises stresses increase substantially, and the first natural frequency decreases significantly, indicating a higher susceptibility to vibration-induced failure. To address these issues, a second optimization incorporates structural constraints on von Mises stress and the first eigenfrequency, ensuring mechanical robustness while improving aerodynamic efficiency.

The two optimized configurations are compared in terms of geometry evolution, aerodynamic efficiency, and structural response. The results clearly demonstrate the necessity of multidisciplinary optimization in achieving compressor designs that are not only aerodynamically efficient across multiple operating conditions but also structurally reliable.

Presenting Author: Arnaud Châtel Von Karman Institute For Fluid Dynamics

Presenting Author Biography: Dr Arnaud Châtel works at the von Karma Institute as Research Engineer since November 2021 in the Turbomachinery and Propulsion department. He graduated with a master’s degree in mechanical engineering from the University of Mons (Belgium) in 2015. Following this, he completed the Research Master’s program at the von Karman Institute in June 2016.
From 2016 to 2021, Arnaud pursued his PhD jointly at the von Karman Institute and at the University of Mons. His PhD research focused on developing innovative optimization with applications to turbomachinery components. Currently, he continues his work in the Turbomachinery Departement, where he focuses on multidisciplinary shape optimization of turbomachinery components, aiming to improve their performance.

Authors:

Arnaud Châtel Von Karman Institute For Fluid Dynamics
Tom Verstraete von Karman Institute for Fluid Dynamics
Nao Taniguchi Research & Innovation Center, Mitsubishi Heavy Industries, Ltd.
Tadashi Kanzaka Research & Innovation Center, Mitsubishi Heavy Industries, Ltd.