SiWiRoRa - Simulated Wind-turbine Rotor-blade Radargrams

SiWiRoRa

SiWiRoRa stand for 'Simulated Wind-turbine Rotor-blade Radargrams'.

Overview

A novel kind of dataset comprised of 9504 grayscale images (quadratic, 224 px) representing simulated radargrams. SiWiRoRa enables analytical machine-learning experiments in the emerging area of radar-based computer-vision research on wind-turbine rotor blades. Here, radargrams are images showing distance on the horizontal x-axis (increasing from left to right) and time on the vertical y-axis (increasing from top to bottom) as well as reflected intensity given by the colorscale (increasing from black to white).

Details

Geometries have been modeled using Cyberbotics Webots (1188 different configurations) and stochastic elements have been added in post-processing (additional 8 independent repetitions per 1 configuration). Besides this aforementioned stochastic noise, the dataset has a full-factorial design consisting in... 

9 distinct levels of rotor speeds (clockwise rotation when viewed from the exterior onto the rotor and the radar behind it)

11 levels of the yaw angles (between radar direction and nacelle orientation)

6 variations of wind pressure (forcing the rotor closer to the radar)

2 different time offsets (corresponding to alternative triggers of the radar)

Geometries have been chosen so as to yield non-axisymmetrical radargrams (even apart from noise).

Acknowledgements

The present dataset complements and has been inspired by the field measurements published under

https://doi.org/10.5281/zenodo.8366654

(Hyperlink is provided in the References section).