Condition based maintenance is increasingly used to curb wind turbine operation and maintenance (O&M) costs, especially at offshore sites where visits are prohibitively expensive. Replacing corrective maintenance with condition-based maintenance cuts downtime and damage, while improving the planning of activities and the use of resources and equipment, which reduces costs.
Many systems are already on the market to monitor the health of components. Approaches such as SCADA data, drivetrain monitoring, visual inspections and oil samples are all common practice in wind energy and have proven their added value.
But all these techniques only start providing useful information once components have really started to degrade. As this degradation is strongly related to the loads acting on the components, ECN has been seeking a solution to monitor mechanical loads.
Condition based maintenance is increasingly used to curb wind turbine O&M costs, especially offshore. Replacing corrective maintenance with condition-based maintenance cuts downtime while improving the planning of activities and the use of resources and equipment (Siemens)
Since most of the loads are introduced via the rotor blades, ECN’s development have been focusing on a low-cost method to monitor the blade root bending moments and process the data to help turbine operators decide on issues such as maintenance action to prevent failures, when to postpone or prioritise visits, or whether to extend turbine life.
Features of an FOBM system
The system is based on optical fibres with Bragg gratings to measure the strain in the blade roots. Compared with the copper strain gauges commonly used for measuring blade root bending, optical fibres should theoretically drift less over time and last longer, while providing reproducible data.
Many attempts to demonstrate the advantages of optical fibres for measuring strain in rotor blades have failed. Not only was the technology not mature enough, but the influence of temperature was underestimated, and installing sensors and measurement equipment left turbines at a standstill for too long. Final results looked far from professional and once measurements were finally obtained, it was unclear how to use them – they could only be added to the large amount of data already available to the operators. ECN now claims its FOBM system offers the benefits of optical fibre technology without its drawbacks. The system consists of:
• A patented sensor assembly that is easy to install and replace, requires no calibration, and provides reliable, accurate and reproducible strain data over a very long period;
• A commercially available interrogator to read the fibre optic sensors;
• A measurement computer that derives loads data from strain data and combines the blade loads data with turbine power line communication (PLC) data;
• Wireless-LAN to enable communication between the rotor and the turbine base;
• Software for data processing that filters and cleans up the time series, categorises the data per design load case, and provides key figures, statistics, and graphs to the operator for O&M optimisation.
The entire system can be installed by regular maintenance technicians in less than a day and requires no further skills in fibre optics. To measure blade bending moments in flapwise and edgewise direction, four sensors must be installed in each blade root. All four sensors are connected to individual strings that are connected to the interrogator. The measured data can be obtained remotely.
Innovative algorithms for data cleaning and filtering
In general, operators want to know if their turbine operates within its design envelope, so that measures can be taken if necessary. Next, operators want to know if some turbines are more heavily loaded than others, so O&M activities can be prioritised along with analyses of condition monitoring data and inspection results.
ECN has implemented the data analysis as follows. From the beginning, the software starts building up a capture matrix with statistically sufficient data that can be used as a reference data set.
The structure of the FOBM system (ECN)
All time series relevant for the capture matrix (which displays results according to load conditions) are stored, for purposes including traceability. The time series are also used to derive fingerprints, envelopes and criteria for rejection. Once the measurement campaign is running, the software detects the load cases (operational modes) present in the 10 minute time series, and splits the time series into single mode files – such as normal operation, start-up, shutdown or emergency shutdown – and filters out erroneous data. Next the software determines statistical data of the single mode files – such as minimum, maximum, mean, standard deviation and equivalent loads – updates the load spectra plots, and analyses the frequencies.
Time series themselves are only stored if they are used for fingerprint information or when the processing is not finished successfully. The software provides monthly reports, giving information about the captured data, deviations from long-term statistics, and a comparison with fingerprint data. Furthermore, the monthly reports contain information about extreme loading conditions and possible errors in the measurement system.
In the longer term, the operator can ask for reports with information about the cumulative loading of the blades and rotor (equivalent loads, fatigue spectra) to monitor the consumed lifetime. This information can be used for comparison with design data or for comparison with data of other turbines.
Each month an overview of the performance of the measurement system is generated. This report provides details of measurements over the preceding month and over the system’s lifetime.
The data processing performs checks on the data quality, as data might be rejected due to measurement faults or unidentified operational modes. These time series are quarantined for further analysis as required and load results can be adjusted for lost data.
Equivalent loads are calculated for all modes. For stable operational modes such as power production, equivalent loads are plotted as a function of the wind speed. For transient modes, values are plotted for each file. Values should be within limits set for the capture matrix and the plots should enable extreme values to be easily identified.
To inform users about encountered loads, these are plotted for the relevant month as well as for the equipment’s lifetime. For the monthly figures, a reference based upon the capture matrix is used. For the lifetime, an external reference should be used.
This external reference can be chosen depending on the application. For prioritising maintenance, other turbines of the same farm could be used for this reference. When the user is interested in calculating the equipment’s theoretical lifetime, a design spectrum outlining the equipment’s intended parameters of use is the most obvious choice.
Monthly statistics from an FOBM installation (ECN)
Frequency plots (average power density):
Under certain conditions, time series are analysed to examine frequencies, which can reveal changes in the blade’s structural properties apparent through shifts in frequency or amplitudes.
Optical Sensor Assembly
The FOBM system requires four sensors per blade to determine the blade load bending moments in both edgewise and flapwise directions. ECN has developed a special sensor assembly for strain measurement in wind turbine blades as a plug-and-play device. The sensor consists of a fibre with a Bragg grating mounted between two studs via a carrier. The studs are mounted on the inner side of the blade root at a distance of 10 cm.
The carrier ensures that the fibre follows the displacements of the studs so the strain in the blade root is measured over enough length to avoid local effects of the blade material. The carrier protects the fibre from sharp bending and also accommodates a second Bragg grating for temperature compensation.
To install a sensor, the technicians glue the studs on the blade with the help of a dedicated mounting tool. After a short curing time, the carrier can be mounted on the studs and the patch cables are plugged in. The carrier with the fibre and connectors is assembled in the factory under well-defined conditions and is already calibrated.
During operation, the sensor provides strain data under both tension and compression with high accuracy (5 με) over a long period of time. Re-calibration is not necessary, because the sensor itself is very stable. If, for whatever reason, the sensor should fail it can be easily replaced. Since all assemblies are calibrated in the factory, on-site calibration is not necessary.
This patented sensor assembly has proven to work reliably, to provide strain data with high accuracy, and to offer easy installation and replacement. The fibre itself has a very high ultimate strain and can easily last a turbine’s lifetime. The most critical element is the connection between the blades and the studs. This is chosen for application in existing turbines, although the sensor’s design also suits other fastening methods that match blade manufacturing processes.
Apart from the sensor and the software for data analysis, the system consists of a commercially available interrogator to read out the fibre optic sensors, a measurement computer that derives loads data from strain data and combines the blade load data with turbine PLC data and wireless-LAN to enable communication between the rotor and the turbine base.
Interrogators that meet the requirements of wind turbine applications are available on the market, although robustness, price level and performance still require additional development. Several types of gratings can be used, but Draw Tower Grating in combination with a WDM interrogator with three channels is preferred in this application. This approach ensures a robust and stable measurement system, while the disadvantages of wavelength gratings is minor, due a well controlled production process of complete measurement strings. In this case four different sensor assemblies are sufficient to cover the whole measurement range for one blade, while the configuration for each blade can be kept identical.
The measurement PC is a standard device with a Windows operating system. Data is imported from the turbine PLC as well as from the blade load measurement system. The measurement software combines these data, and processes the data which results in single mode files. Based on these files, a database is filled with statistical data, load spectra frequency plots and a capture matrix with fingerprint information. The amount of data to be stored is limited, because the original time series are thrown away after successful processing. Only time series which are used as fingerprint information (capture matrix) are stored in the system.
User interface and remote access:
The user can access the system to retrieve the information required. This can be realised via standard communication provisions. During the development phase, the reporting module runs on the remote computer. For commercial implementation other configurations will be more appropriate.
L.W.M.M. Rademakers and T.W. Verbruggen work at ECN.