Search for: All records

Abstract Active machine learning is widely used in computational studies where repeated numerical simulations can be conducted on high performance computers without human intervention. But translation of these active learning methods to physical systems has proven more difficult and the accelerated pace of discoveries aided by these methods remains as yet unrealized. Through the presentation of a general active learning framework and its application to large-scale boundary layer wind tunnel experiments, we demonstrate that the active learning framework used so successfully in computational studies is directly applicable to the investigation of physical experimental systems and the corresponding improvements in the rate of discovery can be transformative. We specifically show that, for our wind tunnel experiments, we are able to achieve in approximately 300 experiments a learning objective that would be impossible using traditional methods. 

Over the course of this project, the turbulence profiles from hundreds of different Terraformer roughness element configurations were collected, providing a very rich dataset of boundary layer flow as a function of upwind fetch. Vertically aligned turbulence measuring Cobra probes vertically separated by 160 mm are mounted on an automated articulating gantry programmed to move in three dimensions. In this manner, three vertical turbulence profiles in different lateral locations are measured during each experiment. The flow is driven by the eight vane axial fans and conditioned by the Irwin Spires through the inlet after which the boundary layer flow is developed over the Terraformer roughness element section before arriving at the measuring plane. The x direction is defined as the upstream from the vane axial fans, y direction is perpendicular to the x direction and represents the lateral locations of the measured vertical profiles. Finally, z represents the vertical height of the cobra probes from the floor of the UFBLWT.