Final Report Summary - MONITUR (Reduction in Maintenance Costs of Wind Turbine Renewable Electricity Generation through Online Condition Monitoring)
Executive Summary:
Reduction in Maintenance Costs of Wind Turbine Renewable Electricity Generation through Online Condition Monitoring:
To achieve the EC binding targets of 20% electricity generation from renewables by 2020 will require between 600 and 833 TWh to be produced by wind power which will require between 119,000 to 165,000 new wind turbines. According to the European Wind Energy Technology Platform “O&M strategies are based on periodic inspection; fault statistics from existing stock are either unavailable or insufficient; maintenance needs to be minimised through preventative strategies and tools like conditional monitoring; performance of offshore farms is severely impacted by poor reliability leading to major production losses.”
Wind turbines are not achieving a design life of 20 years with failure rates of >1/turbine/year.
Replacing a €3k bearing can turn into a major €165k cost involving cranes, crews etc. For onshore turbines operation and maintenance costs around 10-15% of income over 20 years, rising for offshore to 20-25%.
PROJECT OBJECTIVES:
Scientific:
1. Develop dependencies between vibration and stress in components of wind turbines and develop novel modification of Palmgren-Miner rule to improve accuracy of damage prognosis
2. Develop novel non-stationary digital signal processing and novel anomaly detection techniques for diagnosis of non-stationary resonances of rotating units of wind turbines
3. Validate approaches for on-line diagnosis; prognosis and root-cause analysis via controlled experiments
Technical:
1. Develop differential damage diagnosis technology
2. Develop differential fatigue damage prognosis technology
3. Develop differential damage root-cause analysis technology
4. Develop of open system architecture software platform to integrate diagnosis, prognosis and root-cause analysis technologies
5. Develop hardware platform and communications methodology
6. Integrated open system, architecture vibration fatigue diagnosis, prognosis and root cause analysis system
DESCRIPTION OF THE MONITUR TECHNOLOGY
Many of the traditional non-destructive testing (NDT) techniques are unsuitable for wind turbines due to their size and location. Other techniques that are presently in development for wind turbines involve embedding strain sensors and/or piezoelectric excitation in the blade structure. This approach is interesting but there are fundamental issues - installation must be done during manufacture, service or repair of the sensors is virtually impossible and the sensor performance may degrade over time. The Monitur technology employs mounting one tri-axial accelerometer at the bearing pedestal of a low speed shaft of a wind turbine; this accelerometer will capture vibration in three directions (e.g. vertical, horizontal and axial). Vibration data from an accelerometer and shaft speed data from a speed sensor are continuously processed for diagnostics. This concept of monitoring fatigue failures in wind turbines is based on 3 novel technologies:
Fatigue failures diagnosis technology: The implemented radically new fatigue diagnosis technology for rotating parts of wind turbines relies on the measurements of wind turbine vibrations and turbine shaft speed and a sequence of novel signal processing techniques.
Damage root cause technology for wind turbines: Root causes for high vibration amplitudes that exceed the threshold values are defined. Possible root causes are faulty design, installation errors and off-design turbine operation.
Damage prognosis technology: The method of calculating wind turbine blade stresses due to the measured bearing housing vibrations is successfully implemented for the prognosis technology for the operated wind turbine.
The software package will provide the wind turbine operators with real time information on the condition of blades enabling them to schedule effective maintenance and to shut down equipment to avoid destructive failures.
THE PROJECT WEB PORTAL is available to the Project Consortium and the Home Page is also available to the General Public at: http://www.monitur.eu
Project Context and Objectives:
Project Context and the Main Objectives:
The main objective of the Monitur project is to develop radically novel technologies to implement on-line vibration damage diagnosis, prognosis and root-cause analysis that can be applied retrofit or new, to a wide range of rotating turbomachinery. In order to overcome the barrier of not being able to excite vibration modes of rotating shafts, hub and blades during wind turbine operation we will develop a novel approach to exploit the narrowband vibration excitation from the wind turbine shaft as it passes through critical speeds of these components for excitation of component resonances. To overcome the uncertainty of the resonant frequencies of diagnosed rotating components during operation we will employ our novel proposition to exploit the transient operation of wind turbines (start-up; shut-down; acceleration and deceleration) for damage diagnosis. The key innovation will therefore be also to develop signal processing algorithms to interpret non-stationary resonance vibrations from rotating machinery. This will enable the development of online vibration fatigue diagnosis, prognosis and root cause analysis:
Diagnosis – novel detection of resonant frequencies of rotating components during transient regimes will allow for differential fatigue damage diagnosis (i.e. localisation of fatigue damage to a particular stage and rotating component). This will be achieved by detection/diagnosis of non-linearity which will be based on non-linear higher order spectra (i.e. chirp-Fourier triple correlation and chirp-Fourier fourth correlation); diagnosis will be performed at early fatigue stage at low false alarms and missed detections as opposed to late stage (2-5 fold improvements); thus eliminating failures due to fast damage propagation and missed detections and unnecessary un-scheduled maintenance due to false alarms.
Prognosis – develop novel dependencies between stresses in rotating parts and vibration signals from rotating shafts. This will be done by a combination of experimental measurements on selected instrumented rotating machines and by finite element analysis. This will enable on-going real-time prognosis of future fatigue damage and predictive maintenance
Root-cause analysis – develop a technology to estimate vibration of rotating units at multiple modes excited by shaft vibrations; estimate vibration of multiple shaft harmonics. This will enable identification of root cause of fatigue damage in order to intelligently manage proactive maintenance. Root-causes that we can diagnose include faulty design; material defects; manufacturing errors; assembly and installation errors; maintenance errors (imbalance, eccentricity and alignment of shaft).
Project Results:
Work progress and achievements during the period RP2:
Work Package WP1 – Development and Validation of Novel Signal Processing Techniques and Novel Stress-Vibration Dependencies
Start date: Month M1
Key Objectives:
▪ Develop and validate novel signal processing techniques to diagnose nonlinearities due to damage
▪ Develop and validate by experiments novel dependencies between rotating component vibrations and stresses
Tasks:
▪ T1.1 Simulate vibrations from damaged and un-damaged wind turbine components
▪ T1.2 Perform shaker or hammer tests and in-field experiments with damaged and undamaged wind turbine components
▪ T1.3 Validate the higher order spectra techniques for non-stationary wind turbine vibrations
▪ T1.4 Validate the new signal processing technique for online technology adaptation
▪ T1.5 Validate the novel amplitude phase extraction from the non-stationary higher order spectra
▪ T1.6 Validate novel anomaly detection technique for damage diagnosis
▪ T1.7 Validate relationship between vibration and stress in rotating components of wind turbines
S & T Results in Work Package WP1: (see also Deliverable Reports D1.1 D1.2 D1.3 and D1.4)
The results achieved in Work Package WP1 include conducting Shaker tests using the swept sine and impact excitations to investigate the diagnosis effectiveness of wind turbine blades under undamaged or fatigue damaged conditions. A special clamp was designed for rigid mounting of turbine blades to the shaker armature, which assured vibrations to be passed from the shaker to the blade without any interferences or nonlinearities. For shaker experiments, a laser vibratometer was employed for measurements of blade velocity and an accelerometer on the shaker was employed for providing a feedback loop for driving a shaker.
For experimental validation of the implemented fatigue damage diagnosis, 300 realisations of vibration were acquired for the each of the first 5 modes. As a result, a total of 3000 realisations were acquired for the first 5 modes (600 realizations per mode) and 2 states: 300 realisations from fault-free blades and 300 realisations from damaged blade.
For in-field experiments in WP1, a wind turbine model no. LE-600 manufactured in the UK by ‘Leading Edge Turbines’ was purchased and assembled. This has a 1.54m diameter and is capable of producing outputs up to 1000W. This turbine is widely used in Europe in urban and semi-urban areas. It has a downwind arrangement which means that the rotor is at the back of the turbine. Within the in-field area, the LE-600 wind turbine will be positioned on a 9m high guyed tower and secured by guyed wires.
In summary, the validation of the implemented signal processing techniques using simulation data and the swept sine excitation is successful; highly effective simulation diagnosis results (i.e. the estimates of the total probabilities of correct diagnosis are in range of (99-100)%) are matched with experimental results from shaker tests and in-laboratory tests with the operated wind turbine for 32% damage size. The simulation results also show the potential of the implemented fatigue diagnosis for early damage diagnosis (i.e. for effective diagnosis of 5% relative damage size) and effective diagnosis in variation of a wind speed. These successful simulation results should be further confirmed by in-field experiments with operated wind turbine.
These novel results, obtained for the first time in worldwide terms for wind turbine, can be viewed as a preliminary successful experimental validation of the fatigue damage diagnosis technology for operated wind turbine in-laboratory conditions. These results also show the potential of the implemented fatigue diagnosis for effective diagnosis using two vibration modes of rotating blades of wind turbines. The successful experimental results from tests with the operated wind turbine in laboratory conditions should be further confirmed by in-field experiments with operated wind turbine (WP4).
A FE-model is developed for rotating and non-rotating blades of wind turbines. The developed FE-model is used to estimate the dynamic behaviour of wind turbine blades. The modal analysis is performed to extract the resonant frequencies and mode shapes of blades. An experimental test plan is prepared based on the FE analysis, for the optimal placement of strain gauges along the blade.
The developed FE stress-vibration model is successfully validated with the experimental test results and achieved low errors <3.4% and <4.2% in strain between the FEA and experiment test strains for the sine sweep and the sine dwell respectively for multiple blade vibration modes. The developed stress vibration methodologies could be applied for all rotating units of wind turbines (i.e. blades, hub and shaft).
Deliverable Reports D1.1 D1.2 D1.3 and D1.4 have been completed and submitted to the EC REA Participant Portal.
Work Package WP2 – Development of vibration damage diagnosis, prognosis and root cause analysis technologies
Start date: Month M7
Key Objectives:
Develop novel damage diagnosis, prognosis and root cause analysis technologies for wind turbines
Tasks:
▪ T2.1 Develop fatigue diagnosis technology
▪ T2.2 Develop damage root cause technology
▪ T2.3 Develop damage prognosis technology
S&T Results in Work Package WP2: (see also Deliverable Reports D2.1 D2.2 D2.3 and D2.4)
The first section of the Deliverable D2.4 details the signal analysis part (diagnosis, prognosis and root cause analysis) of the Monitur method that will be used in this project. A short summary of the previous research work with relevant faults i.e. wind turbine blade cracks due to fatigue is given. The Monitur signal analysis method is discussed step by step with the help of a very simplified one degree of freedom model of a wind turbine blade using MathCad program.
The second part of the deliverable presents the mathematical approach in more detail and also more with a commercial insight emphasising the novelty of the chosen approach. The development of the methodology is also given a historical perspective i.e. how the different steps have been developed. This deliverable does not go to the finest detailed level of the different phases of the development of the Monitur method as these will, following the description of work, is covered in Deliverables D2.1 D2.2 and D2.3 in Reporting Period RP2 - see below:
The Deliverable Report D2.1 explains the Monitur methodology in signal processing. The main aim is to support ISRI in their programming work in WP3 and to make the mathematical approach understandable to the SME partners.
There are two developed algorithms as follows:
• FFT method with frequency summation technique proposed by VTT.
• Higher-order spectral analysis based on the Chirp Fourier algorithm proposed by Cranfield University.
Of the above algorithms the approach suggested and tested by Cranfield University has been already reported in deliverable D2.4 Interim report on diagnosis, root cause and prognosis technology.
The Cranfield approach is in this report presented by VTT in compact format and as VTT has understood it as a flowchart with the necessary explanations. The algorithm proposed by VTT is suggested as a very simple optional solution for imbalance and misalignment. It has not been tested with the project test data. When compared to the more complicated solution in addition to the simplicity one potential advantage is that the analysis is only done when the acting forces are high since e.g. imbalance is a function of the running speed raised to the power of two.
It should be noted that a number of commercial solutions exist for the diagnosis of such fault types as imbalance, bent shaft and misalignment and that the commercially less competed area is the fatigue fault diagnosis of the wind turbine blades for which the Chirp Fourier based Higher Order Spectrum analysis method developed by Cranfield University is suggested. In the case of a blade
fault the methodology has been tested with a small solid material blade with an artificially (extremely high excitation in vibration shaker) created fault.
In the third part of D2.4 possible risks in the chosen approach are discussed i.e. what could go wrong and why and then most importantly what could be done in order to guarantee the success of the chosen methodology.
The Deliverable Report D2.2 explains in an easily understandable manner, with the aid of a flowchart, the diagnosis approach used in this section of the Monitur project. The focus of this report is in what is commonly known as vibration diagnosis. This covers the diagnosis of the most common vibration related faults such as imbalance, misalignment and bent shaft. It should be noted that blade fault is covered in D2.3. The previously reported D2.4 (Interim report on diagnosis, root cause and prognosis technology) methodology covers a number of fault types, of these imbalance and misalignment together with the blade fault diagnosis have been tested in laboratory. This analysis of this report brings up further questions to be addressed during the exploitation phase of the project. The first important question this brings up for exploitation is can this new technology effectively replace the more straight forward approaches used today for the other faults apart from blade failure?
Secondly the resulting effects of many of these traditional faults are interlinked. Proving all these potential combinations will need further engineering effort before the effectiveness of the current systems can be matched. Thirdly an additional challenge is that quite often a number of these faults can be present simultaneously and so defining the primary faults is again more difficult.
D2.3 goes through the prognosis analysis that is the third part of the Monitur methodology. This deliverable report goes into detail in explaining the Monitur methodology of prognosis analysis in a programmer-friendly manner in order to support ISRI’s work in WP3 and to give the SMEs a better understanding of the methodology. The methodology developed by Cranfield University has previously been reported in Deliverable D2.4 (Interim report on diagnosis, root cause and prognosis technology). As explained in the deliverable D2.1 the methodology for blade fatigue fault has been tested with small wind turbine blades which typically have a solid structure. Consequently, the
capability of diagnosing large blade fatigue failure is still to be proved on constructed type blades. The prediction of the remaining lifetime of wind turbine blades is from a scientific point of view a more open question and there is no proof yet that the methodology works. Evidence from further testing with real data from wind turbines will lead to a decision on the capabilities of the suggested methodology to meet the project specification.
3.3 Work Package WP3 – Development of Software and Hardware Platforms
Start date: Month M5
Key Objectives:
Develop open system architecture software platform to integrate diagnosis, prognosis and root cause analysis technologies..
Develop hardware platform and communications methodology/system.
Integrate open system architecture vibration diagnosis, prognosis and root cause analysis system.
Tasks:
▪ T3.1 Integrate diagnosis, prognosis and root-cause analysis technologies within open system architecture software platform
▪ T3.2 Select, design and integrate hardware components
▪ T3.3 Develop wired and wireless communication system
▪ T3.4 Integrate software and hardware into system
S&T Results in Work Package WP3: (see also Deliverable Reports D3.1 D3.2 and D3.3)
The proposed open system architecture software platform consists of six layers: Data acquisition (DA), data manipulation (DM), state detection (SD), health assessment (HA), prognostic assessment (PA), and root cause analysis (RCA) layers. The final results are related to information on number of detections and false alarms, detailed information on diagnostic decisions made, and list of the root causes of the diagnosis damage of the wind turbine and the quantitative estimation of the root causes.
Turbine vibrations will be used for damage diagnosis, health assessment, prognosis and root cause analysis in Reporting Period RP2. For non-stationary operation, measurements of shaft rotation frequency are required, in order to accurately extract the relevant spectral components. For vibration measurements a tri-axial IEPE accelerometer was selected with appropriate sensitivity (100V/g), range (50g), frequency bandwidth (0.7Hz-5kHz) hermetic sealing, etc. The Independent measurement of the shaft rotation speed is achieved through measurements of the induced output voltage from the wind turbine. The local data analysis is composed of a laptop and wireless router and a set of signal processing algorithms. The router serves as a communication interface between the wireless DAQ system and the local laptop.
Acquisition of measurements is performed using two independent DAQ systems are used for acquisition of vibration and shaft speed measurements. The system is also able to operate in wired mode, since it contains Ethernet module besides the WiFi module. A battery power source was implemented. As the wireless chassis NI-cDAQ 9191 requires 9-30V, a 6 cell Lithium Polymer (LiPo) battery is considered providing 22.2V voltage and 111Wh power capacity, which offers between 16 and 17 hours of uninterrupted data acquisition. The DAQ equipment is enclosed within the weatherproof enclosure of IP54 rating. The enclosure provides a minimum protection against dust and water for outdoor environment.
Initial integration was performed by configuration of the wireless router and the data acquisition system using the provided software. In addition, a program for data acquisition and storage, which is able to communicate with the wireless DAQ systems, was developed in Labview and installed on the laptop.
Deliverable Report D3.3 was due at Month M9 and has been completed and submitted to the EC REA Participant Portal.
Deliverable D3.1 reported that the DLL files produced in the project have successfully been integrated with the Monitur user interface software that has been developed in LabVIEW. The user interface software with all its functionalities described has been developed and tested on a laptop and confirms with the specification detailed in deliverable report D3.3.
Deliverable D3.2 reported that the following are the main components that have been used for the communication and prototype hardware of the Monitur system.
• A Tri-axial accelerometer
• National Instruments Wireless DAQ chassis (NI-9191)
• National Instruments NI-9215 Voltage Acquisition Module
• 15” Touch Panel HMI Screen (TPC2215)
• Wireless Ethernet Router
Two panels have been created to house the Monitur hardware and communication system (as detailed above) for prototype field testing. The first panel is installed on the wind turbine allowing data acquisition from the accelerometer and the data is wirelessly transmitted to the second panel that houses the router and HMI for receiving and processing the data.
3.4 Work Package WP4 – In-Field Trials, Demonstration and Optimisation
Start date: Month M16
Key Objective:
Demonstrate, validate and optimise the developed system via in-field experiments.
Tasks:
▪ T4.1 Integrate diagnosis, prognosis and root-cause analysis technologies within open system architecture software platform
▪ T4.2 Select, design and integrate hardware components
▪ T4.3 Develop wired and wireless communication system
S&T Results in Work Package WP4: (see also Deliverable Reports D4.1 and D4.2)
The finalised prototype Monitur hardware and software were first tested at the ‘Research and Development Centre of Vibro-Acoustics and Fatigue’ at Cranfield University. It was decided that this would represent the best course of action to ensure that the latest software version running with the finalised format of the DLL Files (Dynamic Link Library Files) and their integration with the LabVIEW platform were all working correctly before starting the in-field experiments.
The Monitur user interface software has been developed to allow users to view data being acquired in real time, show the status of the turbines, view and sort alarms and warning events that have occurred and view data from data files generated from the software including processing results of the file.
The software captures data from 4 channels (3 accelerometer channels and 1 channel for rotation speed), creates data acquisition files which are then passed to the integrated DLL files to perform analysis. Results are produced in a file which is then read by the software, interpreted and updates the user with the current status of the wind turbine including flagging warnings and errors as necessary. All data and alarms generated are logged for future reference to the data.
The software also allows the user the view historic data files generated by the Monitur technology together with viewing results for the file. The user is also able to view, change and save parameters in the Monitur configuration files through the user interface software.
The DLL files produced in the project have successfully been integrated with the Monitur user interface software that has been developed in LabVIEW. The user interface software with all its functionalities described has been developed and tested on a laptop and confirms with the specification detailed in Deliverable Report D3.3.
The first in-field tests were performed at the Cranfield University site at Silsoe, Bedfordshire (UK) on a 0.8/1.0kW 5-blade industrial wind turbine and the Monitur Data Acquisition Panel was mounted on the tower of the wind turbine by a pre-designed and pre-fitted special bracket.
The Monitur HMI Panel was connected to the Mains 240V AC power supply, switched on and the Monitur system prototype starting capturing real time, in-field data from the wind turbine. Testing of the complete system continued over a number of days with speeds varying from zero to 30kph.
Potential Impact:
POTENTIAL IMPACTS AND USE:
Results of a study carried out by TU Delft, Durham University and incorporating the results of the EPSRC SuperGen project have concluded that 1 failure/year/turbine is common and reliability decreases with the size of the turbine. When components in a wind turbine fail, the ripple effect can be staggering. Operators must deal with crane mobilisation expenses (as high as €100,000 per incident), lost energy production, excessive costs per kW-hr, and untimely delays in obtaining replacement parts in a burgeoning industry where the demand for necessary components routinely outstrips supply.
The costs from unplanned shutdowns and maintenance fixes can further be compounded by accessibility issues, particularly when the nacelle of a wind turbine is 100 m off the ground or situated offshore miles out at sea. Worker safety too is always a cause for concern. It is estimated that costs of operation, maintenance and parts for a wind turbine with a 20 year life is 10-15% of the total income9. For offshore turbines the cost is estimated to be 20-25% of the total income. The inexact nature of the science of calculating O&M costs begins with the variety of ingredients that make up the whole in addition to scheduled and unscheduled maintenance. An analysis four years ago from Windstats, shows that the costs of O&M ranged from €15-26/MWh. The latest data from the International Energy Agency, reporting from 12 different countries, gives a similar range of €7-26/MWh.
World condition monitoring equipment and services earned revenue of €1,196.6 million in 2005 and are expected to reach €1,951 million in 2012. Most of the participants in the condition monitoring equipment and services market are predominantly focused towards developing and supplying products. Services tend to be the key differentiator in this market and given the fact that most of the equipment vendors are product focused, end-users resort to the idea of outsourcing their condition monitoring programs to service driven organizations.
The condition monitoring equipment and services market can be separated into 4 areas:
• Vibration Monitoring Equipment
• Lubricating Oil Analysis Equipment
• Thermography Equipment
• CM Consulting & Services
The industry is dominated by major players such as General Electric inspection technologies and SKF (who are also a bearing manufacturer). This illustrates that the manufacturers of rotating machinery and components often have proprietary condition monitoring systems. This is especially true in the gas turbine and steam turbine industry. We see the wind power industry as an opportunity for the European SME community to make significant inroads because there are well defined problems that could be solved by our technology developments.
List of Websites:
THE PROJECT WEB PORTAL is available to the Project Consortium and the Home Page is also available to the General Public at: http://www.monitur.eu
Name, title and organisation of the scientific representative of the project's coordinator :
Hasan Terzioglu, Managing Director - Mikrosay Yazihm ve Elektronik Enerji Sanayi Tic.A.S.
Tel: +00 90 216 459 8660
Fax: +00 90 216 459 8370
E-mail: hasan.terzioglu@mikrosay.com.tr
Reduction in Maintenance Costs of Wind Turbine Renewable Electricity Generation through Online Condition Monitoring:
To achieve the EC binding targets of 20% electricity generation from renewables by 2020 will require between 600 and 833 TWh to be produced by wind power which will require between 119,000 to 165,000 new wind turbines. According to the European Wind Energy Technology Platform “O&M strategies are based on periodic inspection; fault statistics from existing stock are either unavailable or insufficient; maintenance needs to be minimised through preventative strategies and tools like conditional monitoring; performance of offshore farms is severely impacted by poor reliability leading to major production losses.”
Wind turbines are not achieving a design life of 20 years with failure rates of >1/turbine/year.
Replacing a €3k bearing can turn into a major €165k cost involving cranes, crews etc. For onshore turbines operation and maintenance costs around 10-15% of income over 20 years, rising for offshore to 20-25%.
PROJECT OBJECTIVES:
Scientific:
1. Develop dependencies between vibration and stress in components of wind turbines and develop novel modification of Palmgren-Miner rule to improve accuracy of damage prognosis
2. Develop novel non-stationary digital signal processing and novel anomaly detection techniques for diagnosis of non-stationary resonances of rotating units of wind turbines
3. Validate approaches for on-line diagnosis; prognosis and root-cause analysis via controlled experiments
Technical:
1. Develop differential damage diagnosis technology
2. Develop differential fatigue damage prognosis technology
3. Develop differential damage root-cause analysis technology
4. Develop of open system architecture software platform to integrate diagnosis, prognosis and root-cause analysis technologies
5. Develop hardware platform and communications methodology
6. Integrated open system, architecture vibration fatigue diagnosis, prognosis and root cause analysis system
DESCRIPTION OF THE MONITUR TECHNOLOGY
Many of the traditional non-destructive testing (NDT) techniques are unsuitable for wind turbines due to their size and location. Other techniques that are presently in development for wind turbines involve embedding strain sensors and/or piezoelectric excitation in the blade structure. This approach is interesting but there are fundamental issues - installation must be done during manufacture, service or repair of the sensors is virtually impossible and the sensor performance may degrade over time. The Monitur technology employs mounting one tri-axial accelerometer at the bearing pedestal of a low speed shaft of a wind turbine; this accelerometer will capture vibration in three directions (e.g. vertical, horizontal and axial). Vibration data from an accelerometer and shaft speed data from a speed sensor are continuously processed for diagnostics. This concept of monitoring fatigue failures in wind turbines is based on 3 novel technologies:
Fatigue failures diagnosis technology: The implemented radically new fatigue diagnosis technology for rotating parts of wind turbines relies on the measurements of wind turbine vibrations and turbine shaft speed and a sequence of novel signal processing techniques.
Damage root cause technology for wind turbines: Root causes for high vibration amplitudes that exceed the threshold values are defined. Possible root causes are faulty design, installation errors and off-design turbine operation.
Damage prognosis technology: The method of calculating wind turbine blade stresses due to the measured bearing housing vibrations is successfully implemented for the prognosis technology for the operated wind turbine.
The software package will provide the wind turbine operators with real time information on the condition of blades enabling them to schedule effective maintenance and to shut down equipment to avoid destructive failures.
THE PROJECT WEB PORTAL is available to the Project Consortium and the Home Page is also available to the General Public at: http://www.monitur.eu
Project Context and Objectives:
Project Context and the Main Objectives:
The main objective of the Monitur project is to develop radically novel technologies to implement on-line vibration damage diagnosis, prognosis and root-cause analysis that can be applied retrofit or new, to a wide range of rotating turbomachinery. In order to overcome the barrier of not being able to excite vibration modes of rotating shafts, hub and blades during wind turbine operation we will develop a novel approach to exploit the narrowband vibration excitation from the wind turbine shaft as it passes through critical speeds of these components for excitation of component resonances. To overcome the uncertainty of the resonant frequencies of diagnosed rotating components during operation we will employ our novel proposition to exploit the transient operation of wind turbines (start-up; shut-down; acceleration and deceleration) for damage diagnosis. The key innovation will therefore be also to develop signal processing algorithms to interpret non-stationary resonance vibrations from rotating machinery. This will enable the development of online vibration fatigue diagnosis, prognosis and root cause analysis:
Diagnosis – novel detection of resonant frequencies of rotating components during transient regimes will allow for differential fatigue damage diagnosis (i.e. localisation of fatigue damage to a particular stage and rotating component). This will be achieved by detection/diagnosis of non-linearity which will be based on non-linear higher order spectra (i.e. chirp-Fourier triple correlation and chirp-Fourier fourth correlation); diagnosis will be performed at early fatigue stage at low false alarms and missed detections as opposed to late stage (2-5 fold improvements); thus eliminating failures due to fast damage propagation and missed detections and unnecessary un-scheduled maintenance due to false alarms.
Prognosis – develop novel dependencies between stresses in rotating parts and vibration signals from rotating shafts. This will be done by a combination of experimental measurements on selected instrumented rotating machines and by finite element analysis. This will enable on-going real-time prognosis of future fatigue damage and predictive maintenance
Root-cause analysis – develop a technology to estimate vibration of rotating units at multiple modes excited by shaft vibrations; estimate vibration of multiple shaft harmonics. This will enable identification of root cause of fatigue damage in order to intelligently manage proactive maintenance. Root-causes that we can diagnose include faulty design; material defects; manufacturing errors; assembly and installation errors; maintenance errors (imbalance, eccentricity and alignment of shaft).
Project Results:
Work progress and achievements during the period RP2:
Work Package WP1 – Development and Validation of Novel Signal Processing Techniques and Novel Stress-Vibration Dependencies
Start date: Month M1
Key Objectives:
▪ Develop and validate novel signal processing techniques to diagnose nonlinearities due to damage
▪ Develop and validate by experiments novel dependencies between rotating component vibrations and stresses
Tasks:
▪ T1.1 Simulate vibrations from damaged and un-damaged wind turbine components
▪ T1.2 Perform shaker or hammer tests and in-field experiments with damaged and undamaged wind turbine components
▪ T1.3 Validate the higher order spectra techniques for non-stationary wind turbine vibrations
▪ T1.4 Validate the new signal processing technique for online technology adaptation
▪ T1.5 Validate the novel amplitude phase extraction from the non-stationary higher order spectra
▪ T1.6 Validate novel anomaly detection technique for damage diagnosis
▪ T1.7 Validate relationship between vibration and stress in rotating components of wind turbines
S & T Results in Work Package WP1: (see also Deliverable Reports D1.1 D1.2 D1.3 and D1.4)
The results achieved in Work Package WP1 include conducting Shaker tests using the swept sine and impact excitations to investigate the diagnosis effectiveness of wind turbine blades under undamaged or fatigue damaged conditions. A special clamp was designed for rigid mounting of turbine blades to the shaker armature, which assured vibrations to be passed from the shaker to the blade without any interferences or nonlinearities. For shaker experiments, a laser vibratometer was employed for measurements of blade velocity and an accelerometer on the shaker was employed for providing a feedback loop for driving a shaker.
For experimental validation of the implemented fatigue damage diagnosis, 300 realisations of vibration were acquired for the each of the first 5 modes. As a result, a total of 3000 realisations were acquired for the first 5 modes (600 realizations per mode) and 2 states: 300 realisations from fault-free blades and 300 realisations from damaged blade.
For in-field experiments in WP1, a wind turbine model no. LE-600 manufactured in the UK by ‘Leading Edge Turbines’ was purchased and assembled. This has a 1.54m diameter and is capable of producing outputs up to 1000W. This turbine is widely used in Europe in urban and semi-urban areas. It has a downwind arrangement which means that the rotor is at the back of the turbine. Within the in-field area, the LE-600 wind turbine will be positioned on a 9m high guyed tower and secured by guyed wires.
In summary, the validation of the implemented signal processing techniques using simulation data and the swept sine excitation is successful; highly effective simulation diagnosis results (i.e. the estimates of the total probabilities of correct diagnosis are in range of (99-100)%) are matched with experimental results from shaker tests and in-laboratory tests with the operated wind turbine for 32% damage size. The simulation results also show the potential of the implemented fatigue diagnosis for early damage diagnosis (i.e. for effective diagnosis of 5% relative damage size) and effective diagnosis in variation of a wind speed. These successful simulation results should be further confirmed by in-field experiments with operated wind turbine.
These novel results, obtained for the first time in worldwide terms for wind turbine, can be viewed as a preliminary successful experimental validation of the fatigue damage diagnosis technology for operated wind turbine in-laboratory conditions. These results also show the potential of the implemented fatigue diagnosis for effective diagnosis using two vibration modes of rotating blades of wind turbines. The successful experimental results from tests with the operated wind turbine in laboratory conditions should be further confirmed by in-field experiments with operated wind turbine (WP4).
A FE-model is developed for rotating and non-rotating blades of wind turbines. The developed FE-model is used to estimate the dynamic behaviour of wind turbine blades. The modal analysis is performed to extract the resonant frequencies and mode shapes of blades. An experimental test plan is prepared based on the FE analysis, for the optimal placement of strain gauges along the blade.
The developed FE stress-vibration model is successfully validated with the experimental test results and achieved low errors <3.4% and <4.2% in strain between the FEA and experiment test strains for the sine sweep and the sine dwell respectively for multiple blade vibration modes. The developed stress vibration methodologies could be applied for all rotating units of wind turbines (i.e. blades, hub and shaft).
Deliverable Reports D1.1 D1.2 D1.3 and D1.4 have been completed and submitted to the EC REA Participant Portal.
Work Package WP2 – Development of vibration damage diagnosis, prognosis and root cause analysis technologies
Start date: Month M7
Key Objectives:
Develop novel damage diagnosis, prognosis and root cause analysis technologies for wind turbines
Tasks:
▪ T2.1 Develop fatigue diagnosis technology
▪ T2.2 Develop damage root cause technology
▪ T2.3 Develop damage prognosis technology
S&T Results in Work Package WP2: (see also Deliverable Reports D2.1 D2.2 D2.3 and D2.4)
The first section of the Deliverable D2.4 details the signal analysis part (diagnosis, prognosis and root cause analysis) of the Monitur method that will be used in this project. A short summary of the previous research work with relevant faults i.e. wind turbine blade cracks due to fatigue is given. The Monitur signal analysis method is discussed step by step with the help of a very simplified one degree of freedom model of a wind turbine blade using MathCad program.
The second part of the deliverable presents the mathematical approach in more detail and also more with a commercial insight emphasising the novelty of the chosen approach. The development of the methodology is also given a historical perspective i.e. how the different steps have been developed. This deliverable does not go to the finest detailed level of the different phases of the development of the Monitur method as these will, following the description of work, is covered in Deliverables D2.1 D2.2 and D2.3 in Reporting Period RP2 - see below:
The Deliverable Report D2.1 explains the Monitur methodology in signal processing. The main aim is to support ISRI in their programming work in WP3 and to make the mathematical approach understandable to the SME partners.
There are two developed algorithms as follows:
• FFT method with frequency summation technique proposed by VTT.
• Higher-order spectral analysis based on the Chirp Fourier algorithm proposed by Cranfield University.
Of the above algorithms the approach suggested and tested by Cranfield University has been already reported in deliverable D2.4 Interim report on diagnosis, root cause and prognosis technology.
The Cranfield approach is in this report presented by VTT in compact format and as VTT has understood it as a flowchart with the necessary explanations. The algorithm proposed by VTT is suggested as a very simple optional solution for imbalance and misalignment. It has not been tested with the project test data. When compared to the more complicated solution in addition to the simplicity one potential advantage is that the analysis is only done when the acting forces are high since e.g. imbalance is a function of the running speed raised to the power of two.
It should be noted that a number of commercial solutions exist for the diagnosis of such fault types as imbalance, bent shaft and misalignment and that the commercially less competed area is the fatigue fault diagnosis of the wind turbine blades for which the Chirp Fourier based Higher Order Spectrum analysis method developed by Cranfield University is suggested. In the case of a blade
fault the methodology has been tested with a small solid material blade with an artificially (extremely high excitation in vibration shaker) created fault.
In the third part of D2.4 possible risks in the chosen approach are discussed i.e. what could go wrong and why and then most importantly what could be done in order to guarantee the success of the chosen methodology.
The Deliverable Report D2.2 explains in an easily understandable manner, with the aid of a flowchart, the diagnosis approach used in this section of the Monitur project. The focus of this report is in what is commonly known as vibration diagnosis. This covers the diagnosis of the most common vibration related faults such as imbalance, misalignment and bent shaft. It should be noted that blade fault is covered in D2.3. The previously reported D2.4 (Interim report on diagnosis, root cause and prognosis technology) methodology covers a number of fault types, of these imbalance and misalignment together with the blade fault diagnosis have been tested in laboratory. This analysis of this report brings up further questions to be addressed during the exploitation phase of the project. The first important question this brings up for exploitation is can this new technology effectively replace the more straight forward approaches used today for the other faults apart from blade failure?
Secondly the resulting effects of many of these traditional faults are interlinked. Proving all these potential combinations will need further engineering effort before the effectiveness of the current systems can be matched. Thirdly an additional challenge is that quite often a number of these faults can be present simultaneously and so defining the primary faults is again more difficult.
D2.3 goes through the prognosis analysis that is the third part of the Monitur methodology. This deliverable report goes into detail in explaining the Monitur methodology of prognosis analysis in a programmer-friendly manner in order to support ISRI’s work in WP3 and to give the SMEs a better understanding of the methodology. The methodology developed by Cranfield University has previously been reported in Deliverable D2.4 (Interim report on diagnosis, root cause and prognosis technology). As explained in the deliverable D2.1 the methodology for blade fatigue fault has been tested with small wind turbine blades which typically have a solid structure. Consequently, the
capability of diagnosing large blade fatigue failure is still to be proved on constructed type blades. The prediction of the remaining lifetime of wind turbine blades is from a scientific point of view a more open question and there is no proof yet that the methodology works. Evidence from further testing with real data from wind turbines will lead to a decision on the capabilities of the suggested methodology to meet the project specification.
3.3 Work Package WP3 – Development of Software and Hardware Platforms
Start date: Month M5
Key Objectives:
Develop open system architecture software platform to integrate diagnosis, prognosis and root cause analysis technologies..
Develop hardware platform and communications methodology/system.
Integrate open system architecture vibration diagnosis, prognosis and root cause analysis system.
Tasks:
▪ T3.1 Integrate diagnosis, prognosis and root-cause analysis technologies within open system architecture software platform
▪ T3.2 Select, design and integrate hardware components
▪ T3.3 Develop wired and wireless communication system
▪ T3.4 Integrate software and hardware into system
S&T Results in Work Package WP3: (see also Deliverable Reports D3.1 D3.2 and D3.3)
The proposed open system architecture software platform consists of six layers: Data acquisition (DA), data manipulation (DM), state detection (SD), health assessment (HA), prognostic assessment (PA), and root cause analysis (RCA) layers. The final results are related to information on number of detections and false alarms, detailed information on diagnostic decisions made, and list of the root causes of the diagnosis damage of the wind turbine and the quantitative estimation of the root causes.
Turbine vibrations will be used for damage diagnosis, health assessment, prognosis and root cause analysis in Reporting Period RP2. For non-stationary operation, measurements of shaft rotation frequency are required, in order to accurately extract the relevant spectral components. For vibration measurements a tri-axial IEPE accelerometer was selected with appropriate sensitivity (100V/g), range (50g), frequency bandwidth (0.7Hz-5kHz) hermetic sealing, etc. The Independent measurement of the shaft rotation speed is achieved through measurements of the induced output voltage from the wind turbine. The local data analysis is composed of a laptop and wireless router and a set of signal processing algorithms. The router serves as a communication interface between the wireless DAQ system and the local laptop.
Acquisition of measurements is performed using two independent DAQ systems are used for acquisition of vibration and shaft speed measurements. The system is also able to operate in wired mode, since it contains Ethernet module besides the WiFi module. A battery power source was implemented. As the wireless chassis NI-cDAQ 9191 requires 9-30V, a 6 cell Lithium Polymer (LiPo) battery is considered providing 22.2V voltage and 111Wh power capacity, which offers between 16 and 17 hours of uninterrupted data acquisition. The DAQ equipment is enclosed within the weatherproof enclosure of IP54 rating. The enclosure provides a minimum protection against dust and water for outdoor environment.
Initial integration was performed by configuration of the wireless router and the data acquisition system using the provided software. In addition, a program for data acquisition and storage, which is able to communicate with the wireless DAQ systems, was developed in Labview and installed on the laptop.
Deliverable Report D3.3 was due at Month M9 and has been completed and submitted to the EC REA Participant Portal.
Deliverable D3.1 reported that the DLL files produced in the project have successfully been integrated with the Monitur user interface software that has been developed in LabVIEW. The user interface software with all its functionalities described has been developed and tested on a laptop and confirms with the specification detailed in deliverable report D3.3.
Deliverable D3.2 reported that the following are the main components that have been used for the communication and prototype hardware of the Monitur system.
• A Tri-axial accelerometer
• National Instruments Wireless DAQ chassis (NI-9191)
• National Instruments NI-9215 Voltage Acquisition Module
• 15” Touch Panel HMI Screen (TPC2215)
• Wireless Ethernet Router
Two panels have been created to house the Monitur hardware and communication system (as detailed above) for prototype field testing. The first panel is installed on the wind turbine allowing data acquisition from the accelerometer and the data is wirelessly transmitted to the second panel that houses the router and HMI for receiving and processing the data.
3.4 Work Package WP4 – In-Field Trials, Demonstration and Optimisation
Start date: Month M16
Key Objective:
Demonstrate, validate and optimise the developed system via in-field experiments.
Tasks:
▪ T4.1 Integrate diagnosis, prognosis and root-cause analysis technologies within open system architecture software platform
▪ T4.2 Select, design and integrate hardware components
▪ T4.3 Develop wired and wireless communication system
S&T Results in Work Package WP4: (see also Deliverable Reports D4.1 and D4.2)
The finalised prototype Monitur hardware and software were first tested at the ‘Research and Development Centre of Vibro-Acoustics and Fatigue’ at Cranfield University. It was decided that this would represent the best course of action to ensure that the latest software version running with the finalised format of the DLL Files (Dynamic Link Library Files) and their integration with the LabVIEW platform were all working correctly before starting the in-field experiments.
The Monitur user interface software has been developed to allow users to view data being acquired in real time, show the status of the turbines, view and sort alarms and warning events that have occurred and view data from data files generated from the software including processing results of the file.
The software captures data from 4 channels (3 accelerometer channels and 1 channel for rotation speed), creates data acquisition files which are then passed to the integrated DLL files to perform analysis. Results are produced in a file which is then read by the software, interpreted and updates the user with the current status of the wind turbine including flagging warnings and errors as necessary. All data and alarms generated are logged for future reference to the data.
The software also allows the user the view historic data files generated by the Monitur technology together with viewing results for the file. The user is also able to view, change and save parameters in the Monitur configuration files through the user interface software.
The DLL files produced in the project have successfully been integrated with the Monitur user interface software that has been developed in LabVIEW. The user interface software with all its functionalities described has been developed and tested on a laptop and confirms with the specification detailed in Deliverable Report D3.3.
The first in-field tests were performed at the Cranfield University site at Silsoe, Bedfordshire (UK) on a 0.8/1.0kW 5-blade industrial wind turbine and the Monitur Data Acquisition Panel was mounted on the tower of the wind turbine by a pre-designed and pre-fitted special bracket.
The Monitur HMI Panel was connected to the Mains 240V AC power supply, switched on and the Monitur system prototype starting capturing real time, in-field data from the wind turbine. Testing of the complete system continued over a number of days with speeds varying from zero to 30kph.
Potential Impact:
POTENTIAL IMPACTS AND USE:
Results of a study carried out by TU Delft, Durham University and incorporating the results of the EPSRC SuperGen project have concluded that 1 failure/year/turbine is common and reliability decreases with the size of the turbine. When components in a wind turbine fail, the ripple effect can be staggering. Operators must deal with crane mobilisation expenses (as high as €100,000 per incident), lost energy production, excessive costs per kW-hr, and untimely delays in obtaining replacement parts in a burgeoning industry where the demand for necessary components routinely outstrips supply.
The costs from unplanned shutdowns and maintenance fixes can further be compounded by accessibility issues, particularly when the nacelle of a wind turbine is 100 m off the ground or situated offshore miles out at sea. Worker safety too is always a cause for concern. It is estimated that costs of operation, maintenance and parts for a wind turbine with a 20 year life is 10-15% of the total income9. For offshore turbines the cost is estimated to be 20-25% of the total income. The inexact nature of the science of calculating O&M costs begins with the variety of ingredients that make up the whole in addition to scheduled and unscheduled maintenance. An analysis four years ago from Windstats, shows that the costs of O&M ranged from €15-26/MWh. The latest data from the International Energy Agency, reporting from 12 different countries, gives a similar range of €7-26/MWh.
World condition monitoring equipment and services earned revenue of €1,196.6 million in 2005 and are expected to reach €1,951 million in 2012. Most of the participants in the condition monitoring equipment and services market are predominantly focused towards developing and supplying products. Services tend to be the key differentiator in this market and given the fact that most of the equipment vendors are product focused, end-users resort to the idea of outsourcing their condition monitoring programs to service driven organizations.
The condition monitoring equipment and services market can be separated into 4 areas:
• Vibration Monitoring Equipment
• Lubricating Oil Analysis Equipment
• Thermography Equipment
• CM Consulting & Services
The industry is dominated by major players such as General Electric inspection technologies and SKF (who are also a bearing manufacturer). This illustrates that the manufacturers of rotating machinery and components often have proprietary condition monitoring systems. This is especially true in the gas turbine and steam turbine industry. We see the wind power industry as an opportunity for the European SME community to make significant inroads because there are well defined problems that could be solved by our technology developments.
List of Websites:
THE PROJECT WEB PORTAL is available to the Project Consortium and the Home Page is also available to the General Public at: http://www.monitur.eu
Name, title and organisation of the scientific representative of the project's coordinator :
Hasan Terzioglu, Managing Director - Mikrosay Yazihm ve Elektronik Enerji Sanayi Tic.A.S.
Tel: +00 90 216 459 8660
Fax: +00 90 216 459 8370
E-mail: hasan.terzioglu@mikrosay.com.tr