The measurement of the rotational speed of rotating machinery is typically performed based on mechanical adherence; for example, in encoders. Nevertheless, it can be of interest in various types of applications to develop contactless vision-based methodologies to measure the speed of rotating machinery. In particular, contactless rotor speed measurement methods have several potential applications for wind turbine technology, in the context of non-intrusive condition monitoring approaches. The present study is devoted exactly to this problem: a ground level video-tachometer measurement technique and an image analysis algorithm for wind turbine rotor speed estimation are proposed. The methodology is based on the comparison between a reference frame and each frame of the video through the covariance matrix: a covariance time series is thus obtained, from which the rotational speed is estimated by passing to the frequency domain through the spectrogram. This procedure guarantees the robustness of the rotational speed estimation, despite the intrinsic non-stationarity of the system and the possible signal disturbances. The method is tested and discussed based on two experimental environments with different characteristics: the former is a small wind turbine model (with a 0.45 m rotor diameter) in the wind tunnel facility of the University of Perugia, whose critical aspect is the high rotational speed (up to the order of 1500 RPM). The latter test case is a wind turbine with a 44 m rotor diameter which is part of an industrial wind farm: in this case, the critical point regards the fact that measurements are acquired in uncontrolled conditions. It is shown that the method is robust enough to overcome the critical aspects of both test cases and to provide reliable rotational speed estimates.
Enhancing the performance of a rotational speed measurement system through data fusion
