## Study of wind turbine wake aerodynamics through the application of motion tracking techniques

##### Master thesis

##### Permanent lenke

http://hdl.handle.net/11250/2401201##### Utgivelsesdato

2016##### Metadata

Vis full innførsel##### Samlinger

##### Sammendrag

In this master thesis an optical tracking system used to evaluate the 3D motion of an object in space has been studied. The aim of the study was to reconstruct the path of a rotating plank, and evaluate the accuracy of the reconstruction. The tracking system involved the use of cameras making prospective modelling and camera calibration necessary. Accuracy of reconstructing world positions from 2D image coordinates has been established through both numerical and experimental investigation. A simple tracking experiment was performed on a rotating plank, to determine deviations from its proscribed trajectory. The tracking method was to be applied in an experimental facility in the Aerodynamic laboratory at NTNU, but was omitted due to construction delay during the period of this thesis work.
Cameras were calibrated using functions in MATLAB Computer Vision Toolbox. Reconstruction errors were determined through back projection of calibration points. The calibration parameters were thereafter used in a numerical analysis to estimate optimal camera positions about the frame object. Monte Carlo simulations were performed with varying angles between setups with two, three and four cameras, respectively. Last, a point tracking experiment was conducted by rotating a plank marked with a traceable point. Two cameras on a linear traverse were used, varying the angle between them and the angle of the object position.
Mean reconstruction errors from the calibration experiment resulted in 0.05 mm in $x$-direction, 0.04 mm in $y$-direction, and 0.1 mm in $z$-direction. From the numerical analysis, the angle of $45^{\circ}$ in vertical direction between each camera and the horizontal plane resulted to be the optimal positioning for all combinations. This implies a total angle of $90^{\circ}$ between two cameras. Furthermore, the setup with four cameras resulted in most accurate reconstruction with a maximum error of only 0.13 mm. In the point tracking experiments, the radii of the rotating paths were estimated within 99.4\% of actual radius for non angled plank; 99.1\% for $20^{\circ}$ and $35^{\circ}$ angled plank; and 98.4\% for $45^{\circ}$ angled plank. Fluctuations of depth was registered in all tracking sequences. Vibration of plank was visually registered, but the magnitude of the vibration was difficult to estimate.