Analyzing vibrations is of great interest in the development and maintenance of wind turbines. For this reason, built-in sensors located within the tower, nacelle, and blades of modern plants are used to acquire vibrational data. The number of sensors that can be installed, however, is limited. Furthermore, the measurable frequencies and spatial resolution are predetermined in the initial sensor configuration during manufacture, and in many cases it is not possible to apply additional, or new, sensors to the plants if the original sensors fail during operation.
A possible solution—at least for temporary measurements—is to stick acceleration sensors onto the components that are being surveyed. For practical reasons, however, the number of these sensors that can be used is also limited. In many cases, mounting the sensors requires considerable effort (e.g., the use of cranes or rope access) and results in downtime of the power plant. In addition, for certain applications, the attached sensors may disturb the measurement, for example, when aerodynamic sources of vibrations or noise should be located on the blades.
To overcome these limitations, we have developed a method for measuring the vibrations of a wind turbine that is based on a laser Doppler vibrometer (LDV).1, 2 Our technique enables vibration measurements to be made at a distance and completely independently of the object itself. Commercial LDVs are available for various applications, but generally only for small ranges and stationary targets. We therefore developed our own LDV, with a 1.5μm laser, which meets laser safety requirements and is appropriate for measurement distances of 250m or more. The main issue we have addressed with our technique is the examination of rotor blades while the wind turbine is under operation. It is thus necessary for the laser spot to follow each desired measurement point on the rotating blades for a particular duration (a few seconds). For this purpose, we mounted the optics of our LDV on a pan-tilt unit (PTU) that is controlled by a camera-based tracking system (see Figure 1).
Figure 1. Optics of the laser vibrometer mounted on a pan-tilt unit. A small stationary camera for tracking can also be seen.
We considered a number of different optical bands (ranging from visual to thermal IR wavelengths) for supplying our image exploitation algorithm with an easy-to-analyze video stream. We eventually decided to use a stationary short-wave IR camera. Although this camera does not provide the best contrast, it is still sufficient to isolate the rotor from the background. This camera also enables us to register the laser spot of the LDV.
We have also developed real-time software that can be used to identify the three rotor blade tips within the camera images (see Figure 2). On the basis of these 2D positions, we can calculate all relevant parameters for a dynamic 3D model of the moving rotor, i.e., the center of rotation, radius, orientation of the rotor plane relative to the camera's position, rotational frequency, and phase of the wings. These parameters are continuously updated by processing the video stream. Our 3D model allows us to predict the trajectory of a desired measurement point for a brief period in the future. In this way, we can compensate for the latencies that are inherent in the data processing and in the pan-tilt head's control system. We can thus actuate the PTU to drive the laser spot of the LDV along that predicted trajectory. Distortions caused by the camera optics, parallax errors, as well as inertia and torque effects of the electro-mechanics, however, can cause deviations of the laser spot from the calculated point. We therefore use a hotspot detector to locate the laser spot in every image. By comparing the real and desired positions, the algorithm learns how to compensate for the deviations. With the use of this feedback loop, we have achieved our accuracy aim, i.e., 0.4mrad for targeted precision.
Figure 2. Image from the video stream obtained from the system camera. The image shows the wind turbine scale model, with the positions of the detected blade tips and laser spot marked in green and red, respectively.
In our work we have also optimized our long-range LDV for measuring vibrations of objects that exhibit macroscopic movements. In particular, we use an automatic frequency control to compensate for the macro Doppler shifts that arise. In addition, we use a polarization diversity technique to improve the signal-to-noise ratios of our measurements (as they are degraded by dynamic speckle effects).
In summary, we have developed a new system—based on a laser Doppler vibrometer—for measuring wind turbine vibrations. With our constructed system we have demonstrated the functionality of the concept and the required accuracy of the technology. We have also developed tracking algorithms and a hardware configuration (using a scale model of a wind turbine) that meet real-life angle and dynamics requirements. The next step will be to adjust the demonstrator setup for the recording of vibration data from a real wind turbine. This will require modification of both the laser and the camera.
This work is supported by the Federal Ministry for Economic Affairs and Energy, on the basis of a decision by the German Bundestag.
Ilja Kaufmann, Clemens Scherer-Klöckling, Peter Lutzmann, Norbert Scherer-Negenborn, Reinhard Ebert
Fraunhofer Institute of Optronics
1. R. Ebert, P. Lutzmann, C. Scherer, N. Scherer-Negenborn, B. Göhler, F. van Putten, Laser vibrometry for wind turbines inspection, Proc. Adv. Solid State Lasers
, 2014. doi:10.1364/ASSL.2014.ATh3A.3
2. R. Ebert, Laser vibrometry for wind turbines inspection. Keynote presentation at SPIE Smart Structures/NDE 2016.