Content uploaded by José A.F.O. Correia
Author content
All content in this area was uploaded by José A.F.O. Correia on Apr 18, 2017
Content may be subject to copyright.
THE INTERNATIONAL CONFERENCE ON
WIND ENERGY HARVESTING 2017
20-21 April 2017
Coimbra, Portugal
Paper XXXXXXX
1 WINERCOST 2017
FATIGUE ANALYSIS OF WIND TURBINE BLADES
Ricardo E. Teixeira1
INEGI, University of Porto
Porto, Portugal
Pedro Moreira2
INEGI, University of Porto
Porto, Portugal
Paulo J. Tavares3
INEGI, University of Porto
Porto, Portugal
José A.F.O. Correia4
INEGI, University of Porto
Porto, Portugal
Miguel Calvente5
University of Oviedo
Gijón, Spain
Abílio De Jesus6
INEGI, University of Porto
Porto, Portugal
Alfonso Fernández-Canteli7
University of Oviedo
Gijón, Spain
ABSTRACT
This paper intends to show the methodology used to perform a fatigue analysis in wind turbine
blades. Experimental results, obtained according to the standard IEC 61400-13, were used to develop
a Matlab software solution. The software uses rainflow algorithm to count the cycles for each strain
range, and calculate damage using Miner’s law.
NOMENCLATURE
DLR
=
German Aerospace Center
INTRODUCTION
Renewable energy market is growing, leading to a need of sound evaluation of the structural integrity of the wind
turbines blades. With this purpose good measurement campaigns should be performed in accordance with standard
IEC 61400-13 "Wind turbines - Part 13: Measurement of mechanical loads", that describes the measurement of
structural loads on wind turbines. [1]
Regarding wind turbines fatigue assessment, there are two main type of loads to consider:
Flapwise direction due to aerodynamic loads (shear, drag, lift);
Lead-lag directional due to inertial loads (gravity, blade dynamics).
These loads usually occur in orthogonal bending directions. Flapwise is the direction where highest loads occur.
These loads vary strongly in amplitude and mean value. Regarding lead-lag loads, they are mainly originated by
the own weight of the blade. Also the load direction changes twice during a single evolution and lead-lag loads are
more regular than fapwise loads. [2] These two type of loads are cyclic so they are the leading cause to fatigue in
wind turbine blades.
A Matlab code for the fatigue life analysis of wind turbine blades was developed. This software was developed
taking into account real data from sensitized blades and guidelines given by the standard IEC 61400-13. The
software is based on a rainflow algorithm to analyze the several loading cycles, and performs damage using
Miner’s law.
METHODOLOGY
1
Development Engineer, INEGI, University of Porto, Portugal, reteixeira@inegi.up.pt
2
Senior Researcher, INEGI, University of Porto, Portugal, pmoreira@inegi.up.pt
3
Senior Researcher, INEGI, University of Porto, Portugal, ptavares@inegi.up.pt
4
Assistant Researcher, INEGI, University of Porto, Portugal, jacorreia@inegi.up.pt
5
Researcher, Gijón Faculty of Engineering, University of Oviedo, Spain, munizmiguel.uo@uniovi.es
6
Assistant Professor, INEGI/Faculty of Engineering, University of Porto, Portugal, ajesus@fe.up.pt
7
Emeritus Professor, Gijón Faculty of Engineering, University of Oviedo, Spain, afc@uniovi.es
THE INTERNATIONAL CONFERENCE ON
WIND ENERGY HARVESTING 2017
20-21 April 2017
Coimbra, Portugal
Paper XXXXXXX
2 WINERCOST 2017
Fatigue is a local damage process caused by the application of cyclic loading even if the load is lower that
the yield stress. Usually, fatigue loads can be divided in two different groups: constant and variable amplitude. In
real load cases it is uncommon to have constant cyclic loads. The case of variable amplitude loads is named load
spectrum or load history.
During its service life, a turbine blade is subjected to cyclic loads that can have constant or variable
amplitude. For cases of constant amplitude loading the determination of the amplitude and the number of cycles
is easy to be performed. Although, if the amplitude of the load changes with time, further processing is needed.
Fatigue cycles can be counted using time histories of the loading parameter such as force, torque, stress, strain,
etc. Cycle counting methods are used to summarize irregular and long load histories giving the number of cycles
for different amplitudes. There are some counting methods of one parameter such as level-crossing, range-mean
and range-crossing, however these methods are not able to deal with local stress-strain hysteresis behavior, and
this has a lot of influence in fatigue failure. Two-parameter cycle counting methods like rainflow are widely used
nowadays.
After the cycle counting there is the need to obtain the damage induced by the cycles. For this purpose
Miner linear damage and accumulation rule can be used. [4]According to this law, a component stressed at has
a life of cycles and the damage after is given by
. This fraction of cycles creates a damage of
the total damage . Fatigue failures occurs when the total damage introduced by multi-level fatigue loading
reaches the critical damage,
Mathematically, the Miners rule is given by,
With the purpose of perform this analysis a matlab application was done. The main interface of the
application is shown in Figure 1.
Figure 1 - Main interface of the application.
This application counts the number of cycles in each variable range (strain, stress or torque). To run the
application the "load files" button should be pressed and a typical explorer window will open to allow the selection
of the files to analyze (At this stage of the application only .mat extension is supported). As soon as files are loaded
a message box appears confirming that the files were loaded successfully (Figure 2).
There is a parameter called delta that is the minimum range of cycles that matters for the analysis. Cycles
with lower range will be neglected. The units for this parameter in this stage of the application are micro-strain
because the analyzed data was in micro-strain. It is also possible to select the number of range intervals that we
pretend being that the standard suggests at least one hundred in order to have a good resolution. The speed bin is
only important if the option "Excel with statistical information" is selected because the name of the excel file will
be "Statistical_information (Speed Bin)".
THE INTERNATIONAL CONFERENCE ON
WIND ENERGY HARVESTING 2017
20-21 April 2017
Coimbra, Portugal
Paper XXXXXXX
3 WINERCOST 2017
Figure 2 - Loading successful message box.
The software allows the analysis of more than one 10 minutes files and the outputs are:
plot of generic and filtered signals (Figure 3 and Figure 4) respectively;
plot of generic and filtered signal with peaks and valleys (Figure 5 and Figure 6 respectively);
2D and 3D histogram for generic and filtered signal (Figure 7, Figure 9, Figure 8 and Figure 10
respectively);
Excel file with statistical information.
Each output can be selected checking the respective check box.
2D histograms are represented in terms of strain range and 3D histograms are represented in terms of strain range
and mean value. The rainflow matrices shows the number of cycles for each strain range, as well as the average
value of the amplitude and mean of each strain range. The excel file that is generated shows the statistical data for
each bin, this is the maximum and minimum strain value for each file, as well as the minimum of the minimums
and the maximum of the maximums. It also prints the average amplitude and mean value for each file as well as
the average for all files. It also calculates the damage equivalent load and standard deviation.
Figure 3 - Generic Signal.
Figure 4 - Filtered signal.
THE INTERNATIONAL CONFERENCE ON
WIND ENERGY HARVESTING 2017
20-21 April 2017
Coimbra, Portugal
Paper XXXXXXX
4 WINERCOST 2017
Figure 5 - Generic signal with peaks and valleys.
Figure 6 - Filtered signal with peaks and valleys
Figure 7 - 2D histogram of generic signal.
Figure 8 - 2D histogram of filtered signal.
Figure 9- 3D histogram of generic signal.
Figure 10 - 3D histogram of filtered signal.
RESULTS
The example shown in previous figures represents the analysis performed to one file corresponding to lead-
lag direction.
Concerning the plots, it is possible to see that a great amount of cycles in the original signal correspond to
noise, and for this reason the signal filtering is underlying as well as the definition of delta value.
In this particular case study the definition of the delta value was based in the curve. The curve used
was the same given by German Aerospace Center (DLR).[5] In this document only curves for and
THE INTERNATIONAL CONFERENCE ON
WIND ENERGY HARVESTING 2017
20-21 April 2017
Coimbra, Portugal
Paper XXXXXXX
5 WINERCOST 2017
are given. For the case shown in Figure 3, is the best approximation. The curve used is shown in figure
6. To obtain the value of delta a large number of cycles was considered, in this case, so it was considered
that the strain values corresponding to this number of cycles did not contribute to fatigue. This approach leaded to
a delta value of.
Comparing Figure 7 with Figure 8 (or Figure 9 with Figure 10) it is possible to notice that a big amount of
cycles with small strain range (approximately 4950 cycles) were neglected form generic signal to filtered signal.
This improves the results and allows a bigger resolution in the results.
Figura 11- curve.
CONCLUSIONS
As future works some tasks are defined,
Perform fatigue tests to material specimens at different R values. This will allow to perform more
accurate analysis as well as a better definition of delta value;
Apply a probabilistic approach to obtain the fatigue curves of the material;
Apply Miner's law to obtain the damage induced in the blades.
ACKNOWLEDGEMENTS
The authors of this work would like to express their gratitude to the SciTech-Science and Technology for
Competitive and Sustainable Industries, R&D project NORTE-01-0145-FEDER-000022 co-financed by
Programme Operational Regional do Norte ("NORTE2020") through Fundo Europeu de Desenvolvimento
Regional (FEDER) and the Portuguese Science Foundation (FCT) through the post-doctoral grant
SFRH/BPD/107825/2015 the for their collaboration, financial and technical support during this research works.
REFERENCES
[1] I. E. Commission, "WIND TURBINES – Part 13: Measurement of mechanical loads," ed, 2014.
[2] R. P. L. Nijssen, "Fatigue life prediction and strength degradation of wind turbine rotor blade composites,"
Knowledge Centre WMC and DPCS group of Aerospace Engineering, 2011.
[3] REN21, "Renewables 2016 Global Status Report," 2016.
[4] P. P. Milella, Fatigue and corrosion in metals: Springer Science & Business Media, 2012.
[5] G. A. R. Establishment, "Fatigue of materials and components for wind turbine rotor blades," 1996.