Optimising the shape of aircraft to improve fuel efficiency and reduce the noise they make is a task normally carried out by highly-skilled professionals using high fidelity computational fluid dynamics tools. Swedish start-up Airinnova has been looking to change this, however, using resources provided by the PRACE SHAPE programme to fully automate parts of the optimisation process.
Founded in 2015 by a team of experts from KTH, Airinnova is a start-up company that provides and develops advanced computational technology solutions for cutting-edge aircraft design, computational aerodynamics, and multi-disciplinary optimisation. It is currently actively involved in European research projects, engineering training and software development for aerospace applications.
High-fidelity computational fluid dynamics (CFD) analysis is a major tool for modern aircraft design and optimisation but requires special skills in mesh generation and executing the analysis code, constraining its use to only the highly skilled. As such, Airinnova has been pursuing ways of performing CFD analysis automatically. This is being done by developing an automated procedure to carry out Reynolds-averaged Navier-Stokes based CFD analyses based on watertight aircraft geometries defined by computer-aided design (CAD) files or the emerging CPACS standard.
Visualised turbulance wake, with top images showing data flow of the CFD automation prototype that Airinnova provides
Automated CFD analysis has the potential be applied to many fields within the aviation design industry, including aerodynamic shape optimisation and the compilation of aero-data for flight simulations. Mengmeng Zhang of Airinnova has now led two PRACE SHAPE projects focused on the first of these.
Shape optimisation is an important task in aircraft design. The ultimate goal of the process is to design lighter, more fuel efficient and quieter aeroplanes by reducing drag, especially at high speeds. In order to find solutions that are not only close to the baseline shape, but also to reveal many more candidate shapes, methods involving computationally expensive fluid dynamics analysis have to be applied.
High fidelity CFD simulations, such as RANS solvers with turbulence models, take up a huge amount of computing resources. The computational time (that is, how long it takes to run the simulations) is a limiting factor that determines whether it is possible to carry out simulations on the available computational resources in real time. In order to make it is possible to actually perform such computationally intensive work, the simulations must be parallelised, which reduces total run-time and makes the simulations feasible when run on highperformance computing resources.
In recent years, the American Institute of Aeronautics and Astronautics (AIAA) Aerodynamic Design Optimisation Discussion Group (ADODG) has created a basic set of benchmark test cases for exercising aerodynamic optimisation methods on in order to highlight and compare the specific contributions of different approaches. Airinnova’s project used the Case IV wing of this Common Research Model (CRM) to test their automatic approach that goes from a design parametrisation all the way through to a CFD calculation.
The work from the first of the two SHAPE projects was done using the PRACE research infrastructure Tier-O resource MareNostrum at BSC, Spain, and the Tier-1 resource Beskow at KTH-PDC, Sweden. It has provided a significant step towards making it possible to perform all the core processes that require special skills – from creating the mesh to executing the high-fidelity CFD analysis code – automatically. It is possible that this will help a wider field of engineering researchers and developers to design aircraft in a much simpler and more efficient way. The benefits to Airinnova are also great; with this technology, it will be possible for well-trained Master’s students to carry out aerodynamic design using high-fidelity CFD.
RANS mesh of the A320-alike aircraft with engine nacelles modelled, generated by Pointwise via automatic meshing scripts
Following on from the first PRACE SHAPE project, Zhang and her colleagues have continued collaborating with the PDC Centre for High Performance Computing at the KTH Royal Institute of Technology (KTH-PDC) to investigate the performance analysis of the open source CFD code SU2, and to further develop the automationprocess for the field of aerodynamic optimisation and design. SU2 is a computational analysis and design package that has been developed to solve multi-physics analysis and optimisation tasks using unstructured mesh topologies. Within this second PRACE SHAPE project, the CFD computations were sped up using HPC resources, and the SU2 code was tested on the previously-mentioned CRM as well as a regional jet-liner (A320-alike).
This second project has led to the development of an automation process to run high-fidelity CFD analysis starting from a given geometry format. This has demonstrated that the automation of CFD analysis for aircraft design is both viable and valuable. The ultimate technical goal of this ongoing research is to develop a flexible and smooth integration of the whole automation process into an MDO (multi-disciplinary optimisation) design environment. This could lead to a revolution in aircraft shape optimisation that will hopefully lead to future aircraft which use less fuel and make less noise.
Project title: High level optimization in aerodynamic design
The resources awarded were:
This project was awarded 100 000 core hours on MareNostrum hosted by BSC, Spain and 150 000 core hours on Beskow hosted by KTH PDC, Sweden.
Research field: Computational fluid dynamics
- For more information see http://www.enccs.se/