An intrinsic part of the development process of any product is optimisation of its features. This may involve polishing the product’s performance, reducing environmental emissions, minimising the amount of material needed in order to reduce cost and maximise profit, or trying to achieve any other goal by modifying the product’s design. Optimisation is so ubiquitous that countless methods of performing it have been developed. However, the majority of them is still based on trial-and-error.
Nowadays optimisation is largely computational. The increasing capabilities of computers were followed by the development of more complex and sophisticated optimisation algorithms, which use clever mathematics to decide which shapes of a given body should be tested to increase the likelihood of discovering a better, more efficient or cheaper design. The testing is done either by computational simulations or by physical testing of prototypes.
In the field of aerodynamics performance of designs can be established either by Computational Fluid Dynamics (high-fidelity simulations of air flow) or by building physical models of a series of different shapes and testing them in wind tunnels. Both of these processes are tremendously expensive and time-consuming when tens, or even hundreds of configurations need to be verified.
An attempt to make optimisation faster and more cost-efficient is the concept of automatization of wind tunnel testing. The idea is to create an aerodynamic body with the ability to modify its shape (morph), that could be controlled by an advanced algorithm running on a PC. This way the process of optimisation could be dramatically speeded up, reducing the cost for designers and enabling them to explore the capabilities of their products deeper than before without exceeding their budgets.
This idea was first tested in 2014 during an undergraduate project at the University of Southampton. An trail object, a simple model of a car diffuser, was designed specifically so that its 3 degrees of freedom (pitch angle, ground separation and diffuser plate angle) could be controlled by linear actuators. The signals to these were sent by a Genetic Algorithm that computed which settings of the diffuser should be examined. The testing of the setup, despite its simplicity and imperfection, provided very promising results – it was possible to test up to 10 configurations a minute and the final shape was reasonably close to what was suggested by previous experimental optimisation.
Despite the outcome of the project there are a number of obstacles to be overcome; above all the complexity of the shape to be optimised is severely limited by the use of linear actuators. In aerodynamics the majority of shapes are not simple geometrical shapes, like the car diffuser, but smooth and round surfaces with complex curvatures. The prospects of automatic morphing of these surfaces will be investigated as part of the PhD project. Another major part of the research will be considering the feasibility of a commercial application of the system.
While the practicality of automated wind tunnel testing is yet unknown, the potential is immense, the system promising dramatically reduced turnaround times in a design process, as well as reduced costs and improved performance.
About the Author
Pawel Kekus is a graduate of Aeronautics & Astronautics from the University of Southampton. Following the success and recognition of his undergraduate work, he decided to continue the research on automatization of wind tunnel testing for a PhD, while remaining in the same faculty. His motivation is driven mainly by his ambition to promote innovation and improve the every-day life of people around the world by contributing within his field of science. Working at Southampton’s modern Boldrewood Innovation Campus, he will be able to make use of cutting-edge facilities for his work, including the first anechoic wind tunnel in the UK.