Artificial intelligence and flow control: toward the automation of adaptive vehicle aerodynamics

Research Project, 2018

Motivation and scientific contribution

The power consumption of road vehicles is largely determined by aerodynamics. Thus, a better aero-package can significantly extend the mileage of future electric cars, and contribute to the fossil fuel independence we are all craving. The classic optimization of the aero-performance of vehicles is based on computational fluid dynamics (CFD) and wind tunnel experiments (WTE). Normally, a flow condition is given, a set of configurations is tested and the most suitable solution is selected. This kind of process gives often a sub-optimal configuration, where neither all parameters are taken into account, nor the configuration selected is optimal for different flow conditions. What if we can develop an adaptive vehicle aerodynamics? What if we have a live-monitoring system adjustable to any given flow condition? Here is where artificial intelligence (AI) can drastically change the classic design process. By coupling AI and WTE it is then possible to create an adaptive, closed-loop control that can independently optimize the aerodynamics of vehicles for any external flow condition.

Background and goal

We have developed and tested an AI-based solution that governs an array of flow control actuators designed for vehicle aerodynamic drag reduction. The object was to develop an AI-flow-control able to suppress the flow separation bubble that naturally occurs at the front rounded corner of a simplified vehicle (a bluff body). Many components were already at our disposal. A wind tunnel model that represents a simplified truck cabin was used as a demonstrator. The model was already proved to be a suitable solution for flow control applications [1]. An array of piezo-electric micro-actuators was already tested in a previous AoA-funded collaboration with Valenciennes University. The micro-blowers showed very interesting results. In particular, an open-loop control was tested, and WT experiments showed the potential of the device as well as its limits when no AI is applied. Some control configurations have a stronger effect than others, considering the reduction of the separated flow region. Nevertheless, how and which variables to tune toward the optimal condition is not an easy tasks to solve. An AI-based closed-loop control was therefore developed. The AI-flow-control was found to succesfully suppress the flow separation bubble that naturally occurs.

References

1. G. Minelli, E. A. Hartono, V. Chernoray, L. Hjelm, and S. Krajnović, “Aerodynamic flow control for a generic truck cabin using synthetic jets,” Journal of Wind Engineering and Industrial Aerodynamics, vol. 168, pp. 81–90, 2017.