A recent project by the CFMS data science and advanced simulation teams has achieved a 15% improvement in the performance of an existing aerofoil through innovative digital engineering methods.
Computational fluid dynamics (CFD) is a well known and important part of a multi-disciplinary design and optimisation process, which involves a number of iterative steps which are computationally expensive, memory intensive and time consuming. These limitations prohibit rapid and comprehensive design space exploration, which restricts the evaluation of alternative designs and is one of the major pain points in the high value design industry.
Recent advances in artificial intelligence (AI), particularly from improved graphical processing units (GPUs) and highly effective model architectures, are now creating exciting opportunities to augment CFD simulated data with real world wind tunnel test data - a ‘best of both worlds’ outcome. Such opportunities have the potential to address the current challenges around CFD analysis in multi-disciplinary design space exploration scenarios.
By implementing the initial CFD analysis of the aerofoil with inputs from the Design Optioneering phase, the CFMS teams built a design space using the relevant flow characteristics of the mission profile. Two-dimensional panel method and boundary layer solver (XFOIL) analyses, which are computationally less expensive, were used to explore the design space, creating data that enabled the AI to optimise the aerofoil against the mission profile.
Data from the CFD analysis taught an AI model to capture the relationship between the design space and the corresponding performance space in a surrogate model. Then, by passing the model through an optimisation loop to identify the optimum geometry configuration that yields the best possible performance characteristics.
Once the AI model is trained, the simulated performance of different aerofoil configurations is usually available in a fraction of a second which is substantially faster than any other method. With appropriate tools and methods, this can also work on more complex 3D designs.
The 15% improvement in aerodynamic performance shows the benefits of this approach and offers major benefits towards reducing aerospace emissions in future, as well as offering benefits to other industry sectors.
To find out more, contact CFMS.
Machine Learning can be used to mine data from large scale simulations and identify the critical design parameters for optimisation. This enables the designer to make use of an automated process of geometry assessment with increased levels of confidence.