Instantaneous Shaft Speed Estimation for Wind Turbine Using a Multi-order Probabilistic Approach in a Continuous Streaming Data Processing Framework

Accurate speed estimation is essential for dynamic analysis of rotating machinery; particularly for condition monitoring and operational modal analysis of applications with significant load and speed variation. The wind turbine is such an application. However, a sufficiently accurate tachometer, to keep track of the instantaneous speed, is not always available in field conditions. This opens the opportunity for virtual tachometer methods, where speed information is extracted from other sources than a physical encoder. We propose the use of a virtual tachometer using a multi-order probabilistic approach that allows accurate speed estimation based on an accelerometer signal placed on the system, e.g. gearbox, housing. The main benefit of this technique is that there is no need to associate a priori one harmonic with one mechanical component. Instead, only a basic knowledge about the kinematics of the system is required. The validity of this approach has been shown in literature on signals of a typical length of maximum 10 minutes. In this paper we adapt the technique such that it can be used for estimating the high-speed shaft speed of a wind turbine gearbox continuously over long periods of time. We illustrate the validity by means of experimental data of an offshore wind turbine.