the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 02 Nov 2023
Dynamic displacement measurement of a wind turbine tower using accelerometers: tilt error compensation and validation
Abstract. For vibration-based structural health monitoring (SHM) of wind turbine support structures, accelerometers are often used. Besides the structural acceleration, the measured quantity also contains the acceleration component due to gravity, which is known as tilt error. This tilt error must be quantified and taken into account, otherwise it can lead to incorrect evaluations, especially in the fatigue estimation or the dynamic displacement estimation using accelerometers. The standard solution is to explicitly measure the tilt angle, which requires an additional sensor for each measurement point and is not applicable for already recorded measurements without tilt information. Therefore, a novel tilt error compensation method is presented by using the static bending line. As a result the influence of the tilt error can be estimated in advance and no additional sensors for tilt measurement are needed. The compensation method is applied to accelerometer measurements of an onshore wind turbine tower and validated with contactless absolute distance measurements from a terrestrial laser scanning (TLS) system. The position and frequency-dependent tilt error of the investigated tower has a significant influence on the quasi static motion below 0.2 Hz with a minimum amplitude error of 9 %, whereas the normalised bending mode shapes around 0.3Hz are only slightly affected.
- Preprint
(2179 KB) - Metadata XML
- BibTeX
- EndNote
Status: closed
-
RC1: 'Comment on wes-2023-123', Anonymous Referee #1, 02 Feb 2024
Dear authors,
I've enjoyed reading your manuscript on your methods and experimental work on determining the displacement of a turbine tower from IEPE accelerometers and validating your findings using a laser (TLS) setup. The work seems to be conducted in a diligent manner and I have no reason to doubt the conclusions of the work.
The authors point out a classic error that is made when using accelerometer data for estimating displacements; the double integration will blow up any low frequency contributions in the spectrum. And in particular because a turbine is slightly tilted (bent) under the nominal load a low frequency contribution from gravity will appear, and exactly this contribution gets blown up by the double integration leading to erroneous estimates of the displacement. In this publication the authors show that it is possible to use a structural model of the turbine to predict this effect on various heights on the structure and introduce a filter to counter this. I like the elegance of this solution and it seems a viable strategy.
However, I do question whether this paper belongs in this journal. The challenges presented are not unique to wind turbines and the paper does not present any results that are unique to wind turbines (E.g. an example on how the method can be used to tackle a particular question in wind energy engineering) . Aside for the fact they were collected on a wind turbine no results are linked to the operation or particular dynamics of wind turbines, in fairness it might have been a regular tower.
Furthermore the authors motivate the need to resolve this issue to allow for a better estimation of the fatigue life and the displacement. They propose IEPE accelerometers to outperform DC capable sensors, such as MEMS, based on better noise characteristics. While this statement is historically true, high quality MEMS are available today that give IEPE strong competition. And I honestly question whether the differences in signal to noise ratio between a good quality MEMS and IEPE really make any significant difference in the final fatigue/displacement estimate. But if you could show me I'm wrong, then I'm happy to learn.
Meanwhile, while the achieved lower frequency bound of the IEPE is impressive, it is still not sufficient to actually capture the slow varying fatigue cycles (e.g. as caused by slow variations in windspeed) that play a significant part in the fatigue of (onshore) wind turbines and would be caught by e.g. MEMS. The authors acknowledge this and mention data-fusion would be required to obtain that goal.Similarly I look at the issue of the tilt error and wonder is this not just an issue of working with just 2 axis? Why could one not look at e.g. Tri-axial MEMS instead? While I didn't do the maths, I suspect that with a Tri-axial MEMS you would be able to compensate for the slow varying tilt error without the need of a model? (but I will accept that I might have overlooked something)
To summarize, there is nothing for me to question the actual work done nor the quality thereof or the quality of the measurement. And the paper was clear and easy to read. But as a whole for I didn't feel like it matched this journal and might be more appropriate in e.g. a journal that targets experimental methods. I'm however willing to review a revision of this manuscript if the authors can draw a stronger link with either the final goal of fatigue estimation (e.g. demonstrating an estimate of the quasi-static load from this measurements in comparison to a MEMS based estimate) or a stronger discussion on how wind energy specific phenomena that were initially obfuscated by the tilt error now are made visible.
Kind regards,
Citation: https://doi.org/10.5194/wes-2023-123-RC1 - AC1: 'Reply on RC1', Clemens Jonscher, 16 Feb 2024
-
RC2: 'Comment on wes-2023-123', Anonymous Referee #2, 12 Feb 2024
- Technically, the term RUL (Remaining Useful Life) is widespread and shall be used. (See line 16)
- Botz (2022) examined the quasi-static tower deflection using RTK. He does not specify standard deviation and quality flag of the solution. In general, RTK has long-term stable static precision. Care must be taken what data is precise enough. My personal experience from a case study is, that less than 20% of the measured data has a standard deviation < 0.02 m and a quality flag of 1 (fixed solution)
- Line 39: "by using of" revise for clarity
- Line 116: "The time integration of measurement data can also be performed in the time domain" revise for clarity
- Figure 3: axis description missing
Tilt error is indeed of great importance for dynamic displacement estimation in time domain. I strongly agree with the importance of measures for tilt error. As the tilt error strongly depends on the nacelle and rotor weight induced bending but also the thrust force induced bending, I'd love to see the SCADA data (wind and power output as a comparison to the plots. It would be important, especially for Figure 13.
I would recommend a flowchart for the data pipeline, as this is hard to understand from text only.
The paper is well-structured, with clear objectives, and it applies valid scientific methods that are well-documented for reproducibility. The results support the authors' interpretations and the discussion is relevant and well-backed. The conclusions are accurately derived from the results.
Citation: https://doi.org/10.5194/wes-2023-123-RC2 - AC1: 'Reply on RC1', Clemens Jonscher, 16 Feb 2024
Status: closed
-
RC1: 'Comment on wes-2023-123', Anonymous Referee #1, 02 Feb 2024
Dear authors,
I've enjoyed reading your manuscript on your methods and experimental work on determining the displacement of a turbine tower from IEPE accelerometers and validating your findings using a laser (TLS) setup. The work seems to be conducted in a diligent manner and I have no reason to doubt the conclusions of the work.
The authors point out a classic error that is made when using accelerometer data for estimating displacements; the double integration will blow up any low frequency contributions in the spectrum. And in particular because a turbine is slightly tilted (bent) under the nominal load a low frequency contribution from gravity will appear, and exactly this contribution gets blown up by the double integration leading to erroneous estimates of the displacement. In this publication the authors show that it is possible to use a structural model of the turbine to predict this effect on various heights on the structure and introduce a filter to counter this. I like the elegance of this solution and it seems a viable strategy.
However, I do question whether this paper belongs in this journal. The challenges presented are not unique to wind turbines and the paper does not present any results that are unique to wind turbines (E.g. an example on how the method can be used to tackle a particular question in wind energy engineering) . Aside for the fact they were collected on a wind turbine no results are linked to the operation or particular dynamics of wind turbines, in fairness it might have been a regular tower.
Furthermore the authors motivate the need to resolve this issue to allow for a better estimation of the fatigue life and the displacement. They propose IEPE accelerometers to outperform DC capable sensors, such as MEMS, based on better noise characteristics. While this statement is historically true, high quality MEMS are available today that give IEPE strong competition. And I honestly question whether the differences in signal to noise ratio between a good quality MEMS and IEPE really make any significant difference in the final fatigue/displacement estimate. But if you could show me I'm wrong, then I'm happy to learn.
Meanwhile, while the achieved lower frequency bound of the IEPE is impressive, it is still not sufficient to actually capture the slow varying fatigue cycles (e.g. as caused by slow variations in windspeed) that play a significant part in the fatigue of (onshore) wind turbines and would be caught by e.g. MEMS. The authors acknowledge this and mention data-fusion would be required to obtain that goal.Similarly I look at the issue of the tilt error and wonder is this not just an issue of working with just 2 axis? Why could one not look at e.g. Tri-axial MEMS instead? While I didn't do the maths, I suspect that with a Tri-axial MEMS you would be able to compensate for the slow varying tilt error without the need of a model? (but I will accept that I might have overlooked something)
To summarize, there is nothing for me to question the actual work done nor the quality thereof or the quality of the measurement. And the paper was clear and easy to read. But as a whole for I didn't feel like it matched this journal and might be more appropriate in e.g. a journal that targets experimental methods. I'm however willing to review a revision of this manuscript if the authors can draw a stronger link with either the final goal of fatigue estimation (e.g. demonstrating an estimate of the quasi-static load from this measurements in comparison to a MEMS based estimate) or a stronger discussion on how wind energy specific phenomena that were initially obfuscated by the tilt error now are made visible.
Kind regards,
Citation: https://doi.org/10.5194/wes-2023-123-RC1 - AC1: 'Reply on RC1', Clemens Jonscher, 16 Feb 2024
-
RC2: 'Comment on wes-2023-123', Anonymous Referee #2, 12 Feb 2024
- Technically, the term RUL (Remaining Useful Life) is widespread and shall be used. (See line 16)
- Botz (2022) examined the quasi-static tower deflection using RTK. He does not specify standard deviation and quality flag of the solution. In general, RTK has long-term stable static precision. Care must be taken what data is precise enough. My personal experience from a case study is, that less than 20% of the measured data has a standard deviation < 0.02 m and a quality flag of 1 (fixed solution)
- Line 39: "by using of" revise for clarity
- Line 116: "The time integration of measurement data can also be performed in the time domain" revise for clarity
- Figure 3: axis description missing
Tilt error is indeed of great importance for dynamic displacement estimation in time domain. I strongly agree with the importance of measures for tilt error. As the tilt error strongly depends on the nacelle and rotor weight induced bending but also the thrust force induced bending, I'd love to see the SCADA data (wind and power output as a comparison to the plots. It would be important, especially for Figure 13.
I would recommend a flowchart for the data pipeline, as this is hard to understand from text only.
The paper is well-structured, with clear objectives, and it applies valid scientific methods that are well-documented for reproducibility. The results support the authors' interpretations and the discussion is relevant and well-backed. The conclusions are accurately derived from the results.
Citation: https://doi.org/10.5194/wes-2023-123-RC2 - AC1: 'Reply on RC1', Clemens Jonscher, 16 Feb 2024
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 590 | 199 | 36 | 825 | 27 | 24 |
- HTML: 590
- PDF: 199
- XML: 36
- Total: 825
- BibTeX: 27
- EndNote: 24
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
