Co-design of dynamic aeroelastic systems using algorithmic differentiation

Lead Research Organisation: Imperial College London
Department Name: Dept of Aeronautics

Abstract

The aim of the research is to investigate strategies for optimum design of a dynamic system, together with its controller, process usually referred to as co-design, and apply that methodology to aeroservoelastic vehicle design. This can be split into the following categories:
- The methodology for devising a dynamic system and corresponding controller
- The methodology for a gradient-based optimisation
- Algorithmic differentiation to find the required sensitivities
The goal is to investigate general solutions that can be applied to a wide range of applications, although the focus here will be on very flexible solar-powered aircraft. To begin, a simple, purely structural problem will be investigated: simulation of a cantilever beam with varying cross-sectional area under an axial load. The external load, applied on the beam is a ramp load starting at zero and the maximum load is reached in time constant seconds which is a variable parameter. The aim will be to identify optimal area distribution for dynamic load minimisation for varying time constant.
In a second phase, the same simulations simulation will be carried out using 3-D solid elements into the open-source SU2 environment. This software is developed in a modular format and is actively developed; hence it provides a robust platform with diverse applications. For the next stage, an additional layer of complexity will be introduced in the form of a fluid. The coupled structural and fluid dynamics can be simulated and optimised for a cantilever wing as a static problem. Time history can then be introduced to the same problem under simple commanded manoeuvres to perform dynamic simulation and optimisation. Adding realistic constraints, the methodology will be further developed to consider the co-design for performance of a solar aircraft wing.

Studentship Projects

Project Reference Relationship Related To Start End Student Name
EP/N509486/1 01/10/2016 31/03/2022
1817454 Studentship EP/N509486/1 01/11/2016 30/04/2020 Charanya Ravi
 
Description Through the work funded by this award, a solver for unsteady FSI problems in low Mach flow has been built in SU2. Previously, compressible flow solver was the only option for unsteady problems which is not suitable for low Mach flows. Through the newly implemented solver, low speed flows including flow from stationary can be simulated accurately which opens a large avenue of unsteady FSI problems that can be explored. In addition to this, a framework has been created for coupled dynamic FSI adjoints in order to efficiently calculate gradients for optimization.
Exploitation Route The outcomes so far can be used to better guide shape optimization of aeroelastic structures for dynamic responses. Examples of problems that can be explored include gust-load alleviation, flutter, limit-cycle oscillations, etc. The generalized unsteady multizone discrete adjoint framework that has been created can be used by others for different types of multizone problems including conjugate heat transfer.
Sectors Aerospace, Defence and Marine,Environment

 
Description Collaboration in Multidisciplinary Optimization Methods with TU Kaiserslautern 
Organisation Technical University Kaiserslautern
Country Germany 
Sector Academic/University 
PI Contribution We have jointly developed a method to obtain design sensitivities in large-scale fully-nonlinear fluid-structure interactions problems, for instance, for the optimization of a wing internal construction to achieve an optimal shape while in flight. Imperial's team was in charge of the development and implementation of the coupled theory.
Collaborator Contribution The team at TU KL developed an implemented the algorithmic differentiation strategy and have supported the parallelization of the software for large-scale computations.
Impact We have one journal paper under review (not reported yet) and we are preparing a release of our solution under the open-source SU2 CFD software.
Start Year 2016
 
Title SU2: The open-source CFD software 
Description SU2 is one of the leading open-source fluid dynamic solvers for aerospace applications. Most of the fluid-structure interaction capabilities have been developed at Imperial, including adjoints for design, and the architecture for simultaneus aerodynamic shape and structural topology optimization. Our contributions started to appear in the official distribution from version 4 of the software and have grown to include also new parallelization strategies, preconditioners for unsteady low-Mach solutions and sparse linear algebra libraries. 
Type Of Technology Software 
Year Produced 2016 
Open Source License? Yes  
Impact SU2 has a very large user base, with over 100k downloads as of January 2017. Users include aircraft companies, start-up, and students from all over the world. Our contribution expands the range of available simulation and it is currently being tested at Boeing to support wind design. 
URL https://su2code.github.io/