Automatic wind tunnel-based optimisation of an automotive underbody diffuser

Abstract

A wind tunnel-based morphing system was devised and utilised for aerodynamic data collection and real-time optimisation of an Ahmed body equipped with a diffuser. Three degrees of freedom were controlled, i.e. ride height, rake angle of the underfloor, and angle of the diffuser plane. Their impact on performance was investigated with a fixed ground. Real-time optimisation was carried out with the aim of determining the most suitable optimisation method for this problem. Tests were carried out using simulated annealing, particle swarm optimisation, pattern search and two genetic algorithms. The results showed that the algorithms demonstrated significantly different performance. However, they were all able to converge on a solution in spite of the hysteresis, which is a characteristic of diffuser flows, and the noise inherent in the system. Pattern search provided the most efficient convergence to the global maximum, despite several discrete aerodynamic changes within the search space, such as sudden flow separation or vortex breakdown. However, it was found to be sensitive to the initial position and noise in the data. The genetic algorithms were found to provide the most reliable convergence, although they were hindered by their inability to make small adjustments in the final stage of convergence. Population sorting was demonstrated as a way to improve the performance of population-based algorithms.
Several new trends in diffuser performance were also observed, most notably that rake, even at small angles, not only generates downforce, but also significantly decreases the critical ride height and energises the diffuser, allowing it to work at higher angles. Up to 1000 different configurations per hour could be tested, making the system attractive for multi-dimensional aerodynamic optimisation, which would be very costly using computational fluid dynamics or conventional wind tunnel testing.

More information

Identifiers

ORCID for Pawel Kekus: orcid.org/0000-0001-7165-7446

Catalogue record

Export record

Altmetrics