Final Report Summary - INTELWIND (Development of an intelligent condition monitoring system for application on critical rotating components of industrial-scale wind turbines)
Executive Summary:
IntelWind’s vision is to substantially decrease the number of failures in critical rotating components and minimise the need for corrective maintenance by developing and successfully implementing an intelligent condition monitoring system based on the integration of acoustic emission, vibration analysis, torque sensing and oil analysis techniques and modules. These developments are aimed to be incorporated on newly manufactured wind turbines but its installation in older wind turbines is possible and affordable.
Table1 shows; the greatest failures of rotating components occur in the gear box, generator and the main bearing, also shown is the failures modes with their occurrence rate and downtime. Though these failures are only a small percentage of the overall failures, the cost of parts, repair and replacement, along with downtime costs make such failure substantial with costs ranging from €200K to €400K. To mitigate these costs, the INTELWIND project has successfully developed and demonstrated in the laboratory and in the field, the ability of the sensors to reliably detect failures and with the use of the dedicated INTELWIND Intelligent software to analyse and interpret the data so as to identify failures thereby allowing planned maintenance and early intervention to prevent costly failure.
The IntelWind Condition Monitoring System (CMS) consists of a number of modules; namely Acoustic Emission, Vibration, Torque and Oil analysis (particulate and moisture). In addition reference parameters needed for data analysis such as the wind turbines operational status (e.g. wind speed and direction, power output) is acquired using other sensors (commonly already provided by the wind turbine logging system). Along with these hardware modules is a suite of software that has been integrated to form the IntelWind CMS. Embedded in the software is a number of analysis algorithms which evaluate from the acquired data such parameters as the Root Mean Square (RMS) value of the signal, the Energy (extracted from the integral of the squared voltage), the peak or maximum voltage of the signal, peak to peak value, the Crest factor (Peak/RMS) which gives an estimate of the wear e.g. in a bearing and Entropy, Kurtosis, Energy Spectral density and others. These values are compared with the ‘healthy’ KPIs (Key Performance Indices) and in the event the values are outside the acceptance range an alarm is activated. This gives IntelWind Intelligence for automated monitoring and early intervention.
The analysis and system performance confirmed that the robustness of the system during the operation under actual field conditions. Measurement data and analysis thereof show the suitable treatment of the multi-parametric problem faced during monitoring of wind turbine components, yet further time is needed for fuller validation.
Apart from the capabilities of the system to identify faults in the critical rotating components of the wind turbine, the INTELWIND system presents a step forward in condition monitoring for wind turbines: The Surface Acoustic Wave (SAW) sensor installed for measuring torque on the high speed shaft, in combination with the software capabilities allows for continuous measurement (and recording) of the load on drive train components. Coupling this with RomaxWind software for the estimation of the remaining life of the components, means a new era in preventive maintenance for wind turbine components has been achieved.
Project Context and Objectives:
The manufacturing revenue generated through wind turbine production reached almost $37 billion in 2011, following yearly growth of almost 2% for the four preceding years, reports MarketLine. In this same period production volumes showed a 22% yearly growth to reach close to 44 GW in 2011. The world wind turbine component market is forecast to record yearly growth in region of 6-11% between 2011 and 2015, according to research from TechNavio, bringing the industry revenue to $49 billion.
Energy demand is the key factor fuelling market growth. High-capacity wind turbines have been gaining a foothold consequently. Leading companies currently operating in the global wind turbine component industry include Xinjiang Goldwind Science & Technology, Vestas Wind Systems, GE Energy and Sinovel Wind Group.
The cumulative wind power installation in the EU-27 at the end of 2012 was 106GW. Wind energy covers 7% of the EU gross electricity consumption and it is expected that 15-17% of the EU electricity will be met by Wind energy by 2020, while a target of around 28% by 2030 is promoted by EWEA. Looking into 2020 and beyond (2050) with all decarbonisation scenarios Wind Energy and thus Wind Turbines will play a key role. In this very positive scenario of larger wind turbines (2.5 5 and 10MW) and increasing wind energy usage, it is essential that energy per Watt price be kept competitive and affordable. In this context; operation and maintenance costs (O&M) which constitute a sizeable share of the total annual costs of running a wind turbine, needs to be lowered.
For example, a gearbox replacement could cost €250k-€350k (note 1$=0.73€), while refurbishing it (where a condition monitoring system (CMS) is in place) will cost in the region of €150k-€200k and a single stage replacement will be €50k-€90k (source: www.gl-garradhassan.com). With all repairs and replacement there is considerable downtime (14 days typically); taking this into account the corrective gear box replacement with no CMS, the revenue loss is estimated to be €415k, with current CMS €405k and with a CMS that interfaces with the operating system the total loss can be substantially reduced to €280k and with more complex interfacing the damage can be minimal. For this reason, operation and maintenance costs are increasingly attracting the attention of manufacturers seeking to develop new designs requiring fewer regular service visits and less out-time. Therefore, one of the highest priorities for the European wind energy industry is currently the significant improvement of the reliability of wind turbines, involving a substantial reduction in current inspection and maintenance costs mainly associated with ‘unpredicted’ failures of critical rotating components.
Today, the health and performance of the wind turbine which consists of the nacelle, blades (rotor blades is a safety issue) and the tower (tower vibration is safety issue) are routinely monitored using systems that have come to be known as ‘Structural Health Monitoring (SHM)’ and ‘Condition Monitoring Systems (CMS). SHM monitoring refers to monitoring of the wind turbine tower, while CMS is associated with the rotating mechanical systems in the Nacelle. Monitoring of the wind turbines and Nacelle mechanical systems in particular has a number of benefits which lead to operational cost reduction and increased safety. The benefits are:
• Substantial reduction in maintenance cost, been able to address faults at an early stage
• Lower downtime (increased availability), since continuous monitoring allows for planned maintenance
• Planned maintenance (reducing unplanned downtime)
• Greater and better documentation of ‘defect events’ for statistical monitoring and future redress
• Better safety as catastrophic failure may be avoided
In this IntelWind project, the Condition Monitoring System monitored the hardware within the Nacelle, namely the drive train (gear box, bearings and shafts) and the generator. The purpose of the ‘continuous’ monitoring is to determine the occurrence of defects/failures and their propagation over time. The failures modes with their occurrence rate and downtime are shown in table 1 (RHS: source General Electric); while the failure modes are typical, the downtime is variable from plant to plant (one needs to bear in mind the lead times for components e.g. a new gear box might be as long as 6 months) and thus a mechanism for early identification of faults prior to complete breakdown and early preventive intervention (planned maintenance) is essential to maximise operational efficiency, lower costs and thus higher profits, not to mention human safety.
In the context of the IntelWind project, the main project objectives were:
1. Develop and test individual modules (AE, Vibration, Torque, Oil etc.)
2. Develop associated hardware and software for data acquisition and analysis (algorithms) for testing of individual modules
3. Design and implement the main control unit (MCU) to enable all measurement data (combined) to be acquired, analysed and stored
4. Develop signal processing and fault analysis techniques to evaluate acquired data from sensors
5. The development of the INTELWIND dedicated SCADA
6. The development of tools to optimise the maintenance planning based on monitored conditions of gears box, bearings and other rotating parts
7. Development of a data management system for collating, storing data and of maintenance, repair and other reports
8. Integration and validation of sensors and sensor modules
9. Validation and Field trials of the integrated CMS using a ‘live’ wind turbine i.e. under actual operational conditions
10. Imparting the knowledge gained on CMS development and use in real life situations to the SME through an awareness and training workshop
11. A plan for the dissemination and exploitation of the project foreground
Having developed the main modules, testing them individually and collectively on an experimental rig, the validation involved measuring operational parameters of the wind turbine and acquiring the measurement data of the INTELWIND modules and the analysis of the critical rotating parts in the nacelle. To reiterate, the modules installed included the Acoustic Emission module for monitoring the bearings on the wind turbine nacelle, the vibration module for monitoring the behaviour of the gearbox, the torque module for measuring the torque through the SAW sensor on the high speed shaft of the wind turbine and the oil module for monitoring the quality of the drive train lubrication system. Additional modules were installed in parallel for verification. These include a secondary (commercial) condition monitoring system based on Acoustic Emission sensors for the bearings, a secondary vibration measurement system used by CRES for gearbox validation testing during accredited testing for certification and a conventional strain gauge based system for measurement of the torque on the high speed shaft.
The validation took the form of measuring and acquiring a number of data (parameters) from the numerous sensors for analysis by the dedicated INTELWIND software. Integration of the modules also involved reference parameters for associating the measurements with the wind turbine operational status. The parameters measured were (a) Wind inflow conditions (measured through the meteorological mast), (b) Wind turbine operational parameters and (c) Mechanical loading on wind turbine components (see D5.1)
The analysis and system performance confirmed the robustness of the IntelWind system during the operation under actual field conditions. Measurement data and analysis thereof show the suitable treatment of the multi-parametric problem faced during monitoring of wind turbine components.
Apart from the capabilities of the system to identify faults in the critical rotating components of the wind turbine, the INTELWIND system presents a step forward in condition monitoring for wind turbines: The SAW sensor installed for measuring torque on the high speed shaft, in combination with the software capabilities allows for continuous measurement (and recording) of the load on drive train components. Coupling this with RomaxWind software for the estimation of the remaining life of the components, means a new era in preventive maintenance for wind turbine components has been achieved.
Project Results:
Please note, all tables and figures mentioned in the text are available in the attached PDF.
The condition monitoring system (CMS) developed in the two year duration of this project is shown schematically in figures 1-3. Figure 1 shows a block diagram of the data acquisition, analysis and communication structure. Figures 2 give an overview of the information collection path and figure 3 the remote logging and communication path ways. In this S&T section we discuss in some depth the main S&T results and achievements from the work packages (WP) WP2 to WP5. WP1 which is the ‘System specification and sample procurement’ is discussed briefly.
WP1: System specification and sample procurement
The main aim of the tasks in this work package was to define the system specification. Prior to defining the system specification it was essential to understand the pre-existing and most frequent faults that occur in wind turbines that are in the field and in the course of use over many months and years. Moreover it was important to establish the tools currently being used in Condition Monitoring Systems (CMS) and how well they identify and recognise the faults as they occur and perpetuate.
Table 1 shows the faults that arise and their severity; this information was provided by the End-user EDPR.
For the evaluation of components located in the nacelle (indicated above), so far the wind energy industry has employed condition monitoring techniques mainly based on vibration analysis, process parameter analysis and oil temperature techniques. In an overview of commercially available CMS systems (Crabtree 2010), of the CM systems that focus on the drive train components i.e. excluding CMS for blades, the majority are vibration based (14 out of 17), while the other 3 were based on oil quality monitoring. The oil monitoring is mostly manual with the oil being collected and sent off for analysis. Torque measurement for instance, is not used. This is the current state-of-the art.
Having established what, where and how the current CMS systems address the turbine monitoring, the IntelWind aims to provide an automated diagnosis and trending of main bearing, gearbox and generator employing the most advanced condition monitoring algorithms combining simultaneous inputs from:
• Vibration using signal features related to specific faults
• Oil analysis using on-line particle count and moisture sensors
• Acoustic Emission inputs acquired above 20kHz
• Running speed vs. time for Order Tracking optionally, removing the load effect
• Torque sensor input
• Additional parameters including power, wind speed, etc.
The specification established on the back of the above relates to the tools and techniques to be used for the automated CMS and there is a comprehensive description of these in the deliverable D1.1.
WP2: IntelWind Sensing System
In any typical wind turbine, the main components of the drivetrain are the main bearing, main shaft, gear box, brake, generator shaft and the generator itself (figure 4).In the context of Condition Monitoring (CM) the main components monitored are the main bearing(s), gearbox and the generator. In wind turbines the rotor shaft arrangement is central to any CM system. This is where the forces and moments due to the wind act upon and get transmitted from the rotor to the generator. The IntelWind CM system addresses this by developing the necessary hardware and software tools to monitor the onset of defects that would lead to failures, thereby by allowing early planned intervention. The internal configuration of a drivetrain will vary from manufacturer to manufacturer and model to model but the basic configuration of the drivetrain and the CM system is same and thus could be adapted to any turbine.
Within this WP, the sensors (Acoustic (AE), Torque and others) have been developed, including the associated electronics and data acquisition. These developments have been done bearing in mind the placement of the sensors in new wind turbines but can be retrofitted to existing wind turbines as testing/validation has shown.
AE Module
The consortium’s SMEs have taken an active role in defining several of the system specifications by stating their requirements; the AE module should focus on monitoring the three gearbox stages and the main bearing. The suggested number of sensors was three and their acquisition time 12 seconds. Apart from components of the ‘drive-train’ the SMEs were particularly interested in using AE to monitor the pitch bearings on pitch controlled wind turbines. Three AE sensors were requested, one per pitch bearing.
Why is the use of AE justified? The gear box of the wind turbine experiences a complex loading pattern due to the high variability of the wind and other transient events, resulting high wear and tear. In the past condition monitoring has relied upon Vibration, oil analysis (debris and quality) and temperature monitoring to determine wear and tear. While these techniques have been extensively used with some success, however they can only monitor gear tooth fatigue cracks etc. at a very late stage, AE gives us a means for early detection and thus plan for early intervention i.e. planned maintenance (Ref. Fused Acoustic Emission and Vibration Techniques for Health Monitoring of Wind Turbine Gearboxes and Bearings, D.J. Lekou et.al.). High frequency AE (100-900 kHz used in IntelWind) has a number of advantages over vibration; it is immune to the noise from the surrounding machinery components and the quick attenuation of the ultrasonic propagation wave which makes it sharper and more pronounced allows to accurately locate the fault origin. These attributes allow AE to detect pitting, cracking and other potential defects earlier than the classical methods of vibration and oil.
How is AE to be used? Here we use wideband type of high frequency AE transducers with enhanced sensitivity and data acquisition and software analysis tools developed to detect defects even in noisy environments. AE permits real-time acquisition of discreet AE signals (waveforms), also trending analysis with ability to record acoustic parameters with respect to time (time-driven data), as well as transient signals generated by active failure. The sensor location can be seen in Figure 4. Figure 5 shows AE sensors attached to the main bearing, high speed gearbox bearing and the low speed gear box bearing.
The two algorithms i.e. software for the verification of the installation (based on analogue triggering) and the monitoring software have been developed; these involve AE waveform acquisition and analysis i.e. parametric data acquisition, feature extraction, envelope analysis, trend analysis, file creation and saving, HTML file, etc.
The signal analysis methodology uses:
(a) Trend analysis where the ‘Root mean square (RMS)’, Crest Factor, peak value and entropy are extracted and compared to a baseline (healthy values)
(b) Envelope analysis (EA): EA can identify periodic impacts, deteriorating rolling bearings. If a defect exists a ‘burst’ of high frequencies are generated each time contact with the defect is made. Bearing defect frequencies relate to the geometry of the bearing and the speed. The impulses generated by a bearing defect excites the ‘natural’ frequency of the bearing housing etc., where by a signal generating from these impulses appear as periodic bursts of high frequency energy at intervals determined by the bearing defect (bearing defect frequency has modulated the bearing housing resonance. Enveloping this signal first requires it to be filtered via a band pass filter centred on the carrier frequency. Enveloping is now applied to extract the repetition rate relating to the particular bearing defect frequency, this is the FFT transformed. What is important is not the frequency but the energy (amplitude). Based on these and ‘Bearing condition value’, an intelligent decision is made to flag a defect or not.
Hitherto in current CM Systems that are in use, fault monitoring using AE is not widespread.
Vibration Module:
Two of the most common techniques employed in CM are vibration analysis and oil monitoring. In this section we discuss the vibration analysis which can be applied inline (in a continuous data acquisition) or offline; meaning periodic data acquisition.
The vibration sensors of the INTELWIND system are able to provide data concerning abnormal vibrations arising from damage or misalignment of rotating components while the oil analysis sensors can provide information about the level of bearing wear and presence of water in the lubricant. The combination of these sensors with the AE sensing module above will provide a considerable amount of useful information regarding the condition of the rotating components of the wind turbine and will permit the detection of any damage at the earliest possible stage of its development (progression).
Vibration analysis using accelerometers as the main detection device can be applied to the whole of the wind turbine drivetrain i.e. the main bearing, the gearbox and the generator (figure 4, green circles). Figures 6 and 7 show two the accelerometers attached to the Nacelle main bearing and gear box bearing respectively.
The Wind turbines are subjected to constantly varying loads, significantly influencing the reliability and thus continuous vibration monitoring is recommended. This also makes condition monitoring based on common vibration analysis techniques a challenge. It is thus necessary to decouple the negative effect of varying wind load on the acquired condition monitoring data in order to be able to accurately and reliably diagnose the actual condition of the drivetrain. The implementation of a diagnostic condition monitoring system whose operation is inert to varying wind loading would offer a step-change over state-of-the-art technology, contributing to the substantial improvement of the reliability and cost effectiveness of wind energy. The Intelwind combination of vibration analysis sensors with oil analysis sensors and subsequently with acoustic emission and torque sensors offer significant advantages over stand-alone traditional vibration analysis (Vibration plus oil) of the drivetrain of current CM Systems. Furthermore, with Intelwind CMS due to these additional data acquisition far less trending is necessary in order to reach a safe conclusion regarding the actual condition of the drivetrain.
The algorithms commonly used can be classified as time and frequency domain. For time domain the monitored parameters are typically peak, peak to peak, mean, root mean square (RMS), Crest Factor (ratio between amplitude and RMS within a defined time window) and Kurtosis amongst others. In the case of frequency domain, they are Fast Fourier Transform (FFT) Spectrum, order analysis, envelope analysis (amplitude demodulation) and side band (shoulder) analysis. Based on these analysis, detailed conclusions on the drivetrain can be taken.
Oil Monitoring:
Oil monitoring is applicable to the gearbox and it is possible to monitor the oil for dirt, wear debris (particles), incorrect oil and others mainly arising from degradation with time. By monitoring the gearbox box oil the lifetime of the gearbox which is the most costly item in the nacelle can be maximised. The monitoring is generally in most ‘in field’ older systems done as an offline oil sample analysis. In newer systems and in Intelwind what is offered is an inline real-time continuous monitoring. These sensors are of two types; oil condition sensor which measures the changes in the oil quality and particle counting sensor (figure 8) which measures total particle count as well as ferrous and non-ferrous counts. When the gears and bearings start to deteriorate the particle generation rate (and size) increases substantially and this is monitored.
Offline oil analysis is useful too, since elemental analysis can be conducted and other parameters (viscosity, acid number etc.) not monitored with the inline particle analysis can be extracted. Yet the IntelWind CMS is focussed on ‘in-line’ measurements.
Moisture and Temperature sensor:
Moisture can cause components to corrode, overheat and malfunction and failure, it is said that water increases the oxidation of the lubricant by more than ten times and bearings loose 75% of their potential service life before oil degradation is noticed, resulting in serious faults in the drivetrain. More severe the moisture ingression greater is the potential risk to the drivetrain, this necessitates moisture monitoring. Modern moisture sensors based on thin film capacitance sensing, can along with smart algorithms provide % relative humidity (RH) values and temperature from dissolved water within the lubricant (oil). Combining with Oil sensor (Particle and quality), this monitoring allows informed maintenance and planned intervention. In the Intelwind CMS system Kittiwake (FG-K16948-KW) moisture senses were used.
Torque Sensor:
Torque measurement have not been utilised in CM systems for drivetrain due to cost and primarily due to the difficulty of attachment (including data reading/transmission). Notwithstanding the cost if the difficulty of attachment and data reading/transmission can be overcome then torque measurements can yield information in drivetrain faults i.e. rotor faults such as mass imbalance. Rotor faults are known to cause torsional oscillation or a shift in the torque-speed ratio. Torque monitoring can thus be used to detect these oscillations and shifts and associate these with rotor faults. Also, shaft torque may be used as indicator for decoupling the phantom perturbations due to higher load. In parallel, for a system operating under highly variable loading conditions, as in the case of wind turbines, torque can be used to directly measure the load sustained by the system and thus, improve remaining life prediction, which is what IntelWind project strived for.
Intelwind has demonstrated the use of Surface Acoustic Wave (SAW) Torque sensors with novel design for the attachment of the sensors to the high speed shaft. The RF rotary coupler that connects the passive sensing elements to a wireless reader was developed along with the reader itself. Figure 9 shows the RF rotary coupler with sensing elements attached to the test shaft used for sensor calibration.
Table 2 highlights the techniques used in IntelWind, the component being monitored and their advantages and disadvantages. The results from these sensors will be highlighted under WP4 and WP5
The above is a summary of the highlights of the hardware i.e. the sensors (Acoustic (AE), Vibration, Torque and others) that have been developed. Next we highlight data fusion and the main control unit.
Main Control Unit/Data Fusion:
The aim of the Main Control Unit (MCU) is to combine the data from the AE, torque, vibration and other sensors. The MCU analyses them using customised algorithms based on existing mathematical models, logs the data and transmit the processed data to the SCADA.
As schematically shown in figure 1, the AE data was acquired using an AdLink PCI 9816H card (equipped with four high linearity 16-bit A/D converters and coverings the AE sensor bandwidth of 100-900 kHz). The other data i.e. Torque, vibration etc. will be acquired using NI cDaq-9191 card which can send data to a host PC over Ethernet or IEEE 802.11 Wi-Fi. Following the acquisition and analysis the data was stored in the Industrial PC (SYS-4U4320-7A01) located in the Nacelle. In order to control the data acquisition, signal processing and decision making process software was developed NI LabView, additional Matlab scripts were embedded into the LabView software.
In summary the main highlights of the MCU are:
• The Main Control Unit (MCU) is responsible for gathering together all available measurements and forwarding them in a specified format, to the Decision Making Unit.
• The format of the output is a simple HTML file that contains data in tabular format, time stamped and tagged with respect to the measurement location.
• This HTML file is saved locally and updated each time a measurement is done
• The MCU accesses the HTML file periodically and pulls the data for further processing in the Decision Making Unit
• The MCU then decides if the raw data files of the measurement will be kept or deleted, depending on the specific thresholds provided by the user for each KPI. The MCU deletes it, if no KPI threshold is breached
This concludes the WP 2 work technical highlights.
WP3: Software and Remote Supervisory Unit
The technical objective of this work package was the development of the:
• IntelWind system software and signal processing for automated operation, paying particular attention to the SCADA for remote logging and further evaluation of the faults detected
• The Intelligent Decision support tools to assist with the maintenance planning
• Overall data management.
Below we highlight the main developments of the IntelWind system software and signal processing for AE, Vibration, Torque and Oil.
Acoustic Emission Software and Signal Processing:
The software developed (some of the coding is shown in figures 10-15: figure 10 gives a snap shot of the AE software allowed us to validate the sensor installation and perform the main task of monitoring. Sensor installation was confirmed by a FFT transform of the acquired signal; the distinctive fundamental and the harmonics (which are related to the material) enabled the sensor installation to be validated, while the monitoring of the component was carried out by trending features extracted from the AE signal (from Hub and Nacelle modules) and Envelope analysis (Nacelle only).
The Trend Analysis involves establishing a reference or base level i.e. component in a healthy state (note: this reference varies with the wind dependent load) and comparing this to the data acquired later in time. Changes or deviations can then be detected. Analysis of the signal waveform and extraction of the ‘Root mean square (RMS)’, Crest Factor, peak value and entropy can then be compared to the baseline (the relevant formulae are given in the deliverable D3.1). The variation i.e. standard deviation (figure 11) is then used to signal the presence of a defect; the alert warning level is pre-set to a factor X times the standard deviation of the signal and the alarm level ‘Y’ times the standard deviation.
Envelope Analysis of the AE signals; defects in the inner race, outer racer and roller bearings produce peaks in AE signals at different frequencies depending on the defected component. Every time the defect (e.g. outer race defect) hits any other element of the bearing (e.g. rolling element), it causes an increase in the AE signal. This signal is filtered via a band-pass filter (coding shown in figure 12) and after demodulation (figure 13) FFT is applied to that signal, followed by searching for the peaks (figure 14) in the FFT spectrum. Along with the other parameters this now provides information on the component of the bearing where the defect is.
Next we discuss the vibration signal analysis software.
Vibration Signal Analysis Algorithms
The vibration signal processing is shown as a block diagram in figure 15.
The time domain signals once acquired was analysed to yield; signal peak value, peak-peak value (figure 16), RMS (figure 17), Crest Factor (figure 18), Kurtosis value (figure 19)and Threshold Level Crossing. Time domain signal will then be transformed into frequency domain for more detail analysis and increased diagnostic capability. The following algorithms have been used (a) Power Spectrum (b) Power Spectrum of the signal Envelope and (c) Power Cepstrum (figure 20)
Torque Analysis Algorithms/Software
The same analysis and software as for vibration was applied for the Torque analysis
Oil Signal Analysis Algorithms
The oil analysis is based on the Kittiwake Debris and Moisture sensors;
• Kittiwake (FG-K16121-KW) Detecting both ferrous and non-ferrous metallic particles
• Kittiwake moisture sensor (FG-K16946-KW) which indicates the water levels in lubricant
These sensors are ‘off the shelf’ sensors from Kittiwake and come with the associated software for particulate and moisture analysis and in IntelWind this software is used for condition monitoring.
Data flow of individual modules to central SCADA and Data Management
The INTELWIND SCADA unit is installed independently from the wind turbine and is capable of monitoring all the modules within a wind turbine and additional modules on a wind farm.
On a given wind turbine, the condition monitoring modules on the various components, i.e. AE-hub, AE-nacelle, VM, OM and SAW, provides the content of the HTML pages, which are then fed into the Central SCADA.
Upon accessing the various HTML files (see figure 21 for HTML file creation code), the file is transferred to the central computer via File Transfer Protocol (FTP). Each HTML file has a data and time stamp, power output, rotating speed and other parameters discussed above e.g. RMS, Crest factor etc. The software of the Central SCADA system then:
• Stores the downloaded files
• Synchronises the data from the individual modules
• Rejects a file based on smart decision
• Checks whether a file has not been updated and reports this as an error log
• Performs additional data manipulation if necessary
• Checks with predefined criteria on possible alert values regarding the condition of the monitored wind turbine components
• Saves the synchronised data
• Refreshes the relevant HTML file of the Central SCADA
• Sends (if applicable) alert signals to user
The above sections present the highlights of the main S&T work in WP3. For a fuller and more comprehensive discussion the reader is referred to the deliverables D 3.1 and 3.2.
WP4: System Integration and evaluation under laboratory conditions
This section reports on the highlights of the integration of the hardware/software modules to form Intelwind CM System.
The hardware and software developed in WP2 and 3 were integrated and validated during WP4. Prior to the integration of sensors etc. with the wind turbine (WT), the sensor modules were validated through laboratory testing and were shown to perform as per the system specification (D1.1). Initial tests were carried out on the ROMAX rig. A second round of tests was done later at ROMAX which picked up ’inserted’ defects. From other tests performed the INTELWIND system’s ability to find difficult to analyse faults (developed naturally) in industrial WT gearboxes including planetary defects and bearing problems were demonstrated. Additional tests were performed for the torque sensor at Transense. Further testing and system validation was carried on the CRES wind turbine as part of task 4.2.
The different techniques in particular AE and Torque were validated through extensive laboratory trials. The AE system (Four Vallen VS900RIC AE) was tested on a test rig (figure 22) provided by Romax in order to prove the reliability and the capabilities of AE module for fault detection.
In this regard the AE system (figure 23) was validated by its ability to detect two different types of common rotating machinery faults; misalignment (misalignment can produce shortcomings such as premature bearing failure, increase in energy consumption, excessive seal lubricant leakage, coupling failure, etc.) and punctual defects (spalling or pitting, mechanical spalling occurs at high stress contact points). These are widely known as bearing failure modes. In this work, the detectability of angular misalignment using INTELWIND AE module with variable operating conditions was tested. The displacement of the shaft detected with proximity sensors was compared with the Acoustic Emission Technique and the Acoustic Emission has shown its capability to detect the shaft displacement using the envelope analysis technique.
In figures 24 and 25 we show the FFT spectrum of the proximity signal in the horizontal and vertical directions. It is evident from this the 2X peak in the vertical direction has a greater amplitude than the 2X in the horizontal (radial) direction. The 1X/2X ratio for the vertical and horizontal directions is 16.6 and 28.57 respectively. This is indicative of an angular misalignment in the shaft.
The complimentary AE testing consisted 9 different conditions where speed, axial and radial load were varied (deliverable D4.1). To detect the frequency of the modulation ‘envelope analysis’ was performed. It was evident that the peak value of 1X and 2X increases with increasing speed. However, there was an inverse relation of 1X peak value with load. In contrast, 2X increased with increasing load and the speed had little impact in 1X/2X ratio. There was also a clear trend of increasing 1X/2X ratio with increasing load.
Figures 25 and 26 are the FFT spectrum of the vertical proximity sensor signal under condition 6 and the enveloped AE signal i.e. a comparison. It is evident from these figures that the peak at 0.18X (5Hz) is detectable in both graphs; a manifestation of a misalignment. The modulation produced in the AE envelope at 1X is evident. However, the 2X modulation is not so clear but it shows an increase with increasing load and speed (see deliverable D4.1 for graphs). It is also worth noting that although the proximity sensor signal measurements shows that the shaft displacements is constant with changing load and speed, the signature produced in AE is affected by load and rotating speed.
In the case of spalling, a defect was artificially produced in the bearing for test purpose. The outer race defected area generates high frequency AE transients (bursts) produced by the asperity contact between the outer race defect when it came in contact with the rolling element. Figure 27, the FFT of the enveloped AE signal shows the 1X peak, attributed to misalignment, the 2nd peak at 6.61X this was from calculations assigned BPFO (Ball passing frequency of the outer race) and the 3rd peak is the harmonic (2x 6.61X).
The source of AE transients is due to the artificially created defect i.e. the material protrusions above the surface roughness of the outer race. Envelope analysis was shown to be a reliable technique for defect detection in the outer race of the bearing, though the time domain signal did not show a significant evidence of the artificially seeded defect. Furthermore, the signal to noise ratio was seen to increase considerably with FFT envelope analysis technique.
In conclusion, in respect of AE, envelop analysis provides a reliable means of defect detection and allows the maintenance process to be planned. Full details and discussions are in the deliverable D4.1.
Another important module in the integration is the Torque module. Two types of detachable plate transducers were proposed and designed. The first one used eight pins as an interface between the plate and the shaft, named Pin Design. The second one used a hard steel wire as an interface between the plate and the shaft, referred to as Circular Design. These two models were manufactured after thorough analysis by modelling. They were manufactured in Stainless Steel 316, which had the mechanical properties to withstand stresses under laboratory conditions without yielding but possibly not for real use in wind turbines due to the higher torques expected in real shafts. For this reason a third pair of plates with the best possible design and a different grade of Stainless Steel, PH17-4 SS, which had better mechanical properties was built. This is referred to as ‘Hybrid Design’ (please see deliverable D4.1). Two SAW sensing elements, HFSAW and LFSAW, were installed on each pair of the plates to turn them into torque transducers (figure 28 RHS).
One of the novelties of the torque work was the design and development of the RF rotor coupler (sensor antenna) by the consortium member Transense. Fig. 28 (LHS) shows two halves of the rotor couple made of the 5mm thick foam (Rochacell® WF) as a dielectric substrate and a copper foil microstrip. The same type of the copper foil was also attached to the back side of the foam as a ground plane. The two halves of the rotor couple were installed on the test shaft (Figure 28 RHS).
Tests carried out using the ‘hybrid design’ with and without wires (D4.1) at room temperature and a static torque of -1600Nm to +1600Nm corresponding to a ±2500Nm applied to the real coupling, showed that the installation without wires was best in terms of hysteresis and direct contact between the shoulders and the shaft also gave considerably smaller variation of the offset and the torque sensitivity with temperature, thereby increasing the accuracy of the temperature compensation of the torque readings.
The maximum torque measurement error achieved was 0.34% full range, indicating the plate transducer design developed in the course of the project is suitable for installation on the wind turbine coupling shafts for defect detection.
Full details of the torque transducers are given in D4.1.
Following laboratory validation, the vibration module hardware and software was integrated and tested on its efficiency and accuracy in detecting and assessing damage on sample components under controlled conditions. The tests were performed on:
• an actual industrial wind turbine with defective components and
• on laboratory test rigs with simulated faults
This was done in order to evaluate diagnostic ability of integrated system and compare predictability of Acoustic Emission and Vibration techniques under controlled condition. In order to allow this analysis, software was developed based on the Feldman PCM (this allowed AE, vibration and torque data input).Using this newly developed software all signals were analysed and results combined. Signals acquired were automatically processed in order to establish Key Performance Indexes (KPI’s) directly relating values with the machine component condition for sensor location. KPI’s evaluated for AE and vibration signals were transferred to the decision making Unit for trending and comparison with alarm levels in an HTML format. KPI’s produced will then be able to be plotted Vs time in order to identify deterioration trends for both AE and Vibration, an example is shown in figure 29, with a data set in figure 30.
In conclusion of WP4, the different modules of the INTELWIND sensing system have been validated through laboratory tests. Their ability to perform according to the system specifications (D1.1) has been proven. From tests performed under System Integration the ability of the INTELWIND system to find difficult to analyse faults developed naturally in industrial wind turbine gearboxes including planetary defects and bearing problems has been demonstrated.
WP5: System Validation (Field Testing of Integrated System)
The task and the highlight of this work was the validation of the IntelWind hardware/software CM System under actual working conditions. The evaluation was conducted on a full-scale NEG MICON 750kW wind turbine. The intelligent INTELWIND condition monitoring system; the hardware and software was installed on the turbine with a view to assess the critical components of the machine and to establish the reliability and stability of the CM System, through the experience attained from more than 3 months of operation under actual conditions of the INTELWIND system housed on the NEG MICON 750kW turbine.
In this WP5, the INTELWIND, intelligent condition monitoring system was implemented based on the integration of acoustic emission, vibration analysis, and torque sensing and oil analysis techniques, specifically for critical rotating components in wind turbines. A layout of the installation and location of the sensors are given in figurers 2 and 4.
The system was already tested under laboratory conditions (see above and D4.1) and proved that it is able to detect faults such as misalignment and bearing outer race defect. The long term operation of the system under conditions of a typical wind turbine (temperature, vibrations, humidity, etc.) was validated through these field tests.
The hardware modules of the integrated INTELWIND condition monitoring system (please see deliverable D2.1) that was installed on the wind turbine and integrated for field testing were:
• Acoustic Emission module (AE): Deliverable D2.2
• Vibration monitoring module (VM): Deliverable D2.3
• Oil analysis module (OM): Deliverable D2.3
• Torque sensor module (Surface Acoustic Waves (SAW)): Deliverable D2.3
The signal analysis of these modules is described in INTELWIND deliverable D3.1 and the technical characteristics of the INTELWIND central wind turbine system are provided in deliverable D3.2. The evaluation trials of the system under laboratory conditions have been extensively described in deliverable D4.1 and highlights above.
The wind speed and wind direction measurements used in the frame of the INTELWIND project were from the meteorological mast situated near the wind turbine (the mast itself is outside the project).
Figures 31-35 show the sensor installations on the generator, nacelle’s yaw, main bearing, high speed gearbox, low speed gear box etc. (these are a selection of images, for a fuller account see D5.1) For ease of assimilation a few figures presented earlier in the text are represented here.
The validation involved measuring operational parameters of the wind turbine and acquiring the measurement data of the INTELWIND modules and the analysis of the critical rotating parts in the nacelle. For a fuller description and condition monitoring results, please see deliverable D5.1.
The analysis and system performance confirmed the robustness of the CM system during the operation under actual field conditions. Measurement data and analysis thereof show the suitable treatment of the multi-parametric problem faced during monitoring of wind turbine components.
Apart from the capabilities of the system to identify faults in the critical rotating components of the wind turbine, the INTELWIND system presents a step forward in condition monitoring for wind turbines: The SAW sensor installed for measuring torque on the high speed shaft, in combination with the software capabilities allows for continuous measurement (and recording) of the load on drive train components. Coupling this with RomaxWind software for the estimation of the remaining life of the components, means a new era in preventive maintenance for wind turbine components is in the near horizon.
Potential Impact:
4.2 Use and dissemination of foreground
The main objective of the dissemination strategy is to enrich the project beneficiaries (SMEs) with the know-how gained from the project through reports and workshops and also educate and inform the wider audience in particular those engaged in the wind turbine industry via conferences and publications.
The IntelWind dissemination strategy aims to maximize the economic benefit of it to the SMEs as quickly as possible, following the end of the project. In this regard the RTD’s of the project held an Awareness Workshop in Greece in September 2013 to disseminate and train the SMEs (within this project) on the foreground knowledge acquired and also train the SMEs in the installation and software operation of the IntelWind Condition Monitoring System via a formal one day presentation course.
There are several agreed routes amongst the partners for the general dissemination activities of the project results, these include:
• Inclusion of project results on the Beneficiaries’ websites.
• Production of dissemination material in the form of papers, presentations, posters/exhibition material and flyers.
• Presentations and publications in industrial and academic conferences and journals.
• Establishment of the project website that is accessible to all, with both a confidential and publically accessible section.
• Development of a dissemination diary to identify all open external meetings where IntelWind should be show-cased.
• Identifying the knowledge that can be exploited from the research and results of this project
• Development of appropriate training material for in-house use in Beneficiary organizations, as well as for use in academic or continuing professional development courses.
• Educate the members of the consortium in technology transfer and protection of intellectual property rights; the absolute necessity to maintain strict confidentiality in order to make patenting of innovations possible
• Patent innovations, oversee training programmes, workshop and dissemination of non-patentable results and innovations; patentable results may be disseminated after patents have been obtained
• Provision of project brochures directed both towards potential users and to other relevant European organizations
• Have ‘Memorandum of Understanding (DOW and D6.4)’; specifies consortium partners’ intentions regarding ownership and commercialisation of IP resulting from joint research: Partners have agreed that the intellectual property generated during the project will be owned by the SMEs.
The above has been successfully implemented, the below bullet points are still to be fully implemented. These activities are shared within the consortium.
• Co-ordination with other relevant EU and national projects where identified as been relevant to IntelWind.
• Co-ordination with European, national and regional trade associations and technology networks to make their members, particularly SMEs, aware of the developments of the IntelWind project.
• Inclusion of the results in the industrial Beneficiaries’ marketing strategies, particularly outside Europe.
• Provision of data into relevant European and international standards (if applicable).
• Provision of project brochures directed both towards potential users and to other relevant European organizations.
• Implement a Content Management System (CMS) which would hold a list of industrialists and others interested in exploiting the results of the project
• Continue to monitor the industries requirements and explore how we can capitalise on the insights gained from the project outcomes
• Actively engaging with the EU Project Officer to promote and network using the good-offices of the EU
• Implement with interested parties, letters of interest (LI), system transfer agreements (STA) and material transfer agreements (MTA), which include the transfer of technical know-how
• commercialisation of IP resulting from joint research: Partners have agreed that the intellectual property generated during the project will be owned by the SMEs and SMEs will exploit where possible
• Collaborative research and data sharing: Access to data to researchers outside the consortium under strict guidelines of data sharing policy with approved researchers and institutes
At this juncture (TRL6, see below), a number of dissemination activities in relation to the foreground have already taken place and more are planned. The following activities have been completed ( see Table 3).
Further dissemination activities will be undertaken in 2014 -2015 onwards, these are:
• 4th Annual Wind Power Romania and Easter Markets, 21-22 January, JW Marrriott, Bucharest, Romania. This conference gives us an opportunity to promote IntelWind, explore the Eastern European market and network and solicit companies to engage with us on testing/validation and opening up their markets to our products both individual and the CM System as a whole
• Emerging Markets workshop: This special workshop will highlight the opportunities for wind energy in the eastern European and emerging markets. Sharing experience and discussing about future trends and business opportunities will be the main focus of this program. This will be on the 5th and 6th February 2014 and again an opportunity to promote IntelWind and network.
• EWEA 2014, 10-13 March, Barcelona, Spain. This is one of Europe’s Premier Wind Energy Events. It is envisaged we will take part in the exhibition. This will also allow for valuable networking, who you know matters!! It is expected that our end-user EDPR (from Spain) will be one of the participants
• Offshore 2015 (10-12 March 2015): In this timeline we expect IntelWind to be fully validated and ready for commercial exploitation, with this in mind we will present the IntelWind CM system at the conference
It is expected that the participation at these conferences will be funded by the attending SMEs and RTDs themselves.
In respect of exploitation in this post project period before commercialisation, the consortium envisages a number of developments that will be of keen interest to the wind turbine community. These are:
• The newly developed Torque module as part the CM system or a standalone module
• The fault recognition using the combined AE and Vibration data, not individually, this allows enhanced predictability
• Further testing of IntelWind System in a modified configuration at CRES and EDPR, this has been agreed in principal with CRES and EDPR, others will be enlisted.
Having evaluated the exploitable results particularly with reference to the well established TECHNOLOGY READINESS LEVEL (TRL) criteria of 9 readiness levels;
TRL 1: Basic principles observed and reported
TRL 2: Technology concept and/or application formulated
TRL 3: Analytical and experimental critical function and/or characteristic proof-of concept
TRL 4: Component/subsystem validation in laboratory environment
TRL 5: System/subsystem/component validation in relevant environment
TRL 6: System/subsystem model or prototyping demonstration in a relevant end-to-end environment (ground or space)
TRL 7: System prototyping demonstration in an operational environment (ground or space)
TRL 8: Actual system completed and "mission qualified" through test and demonstration in an operational environment (ground or space)
TRL 9: Actual system "mission proven" through successful mission operations (ground or space)
Based on the above, the IntelWind Condition Monitoring System is deemed to be at Technology Readiness Level 3-5, where some modules are at a higher stage in its development. It is expected to bring all modules to TRL 5 with further development and laboratory testing to ensure the starting base for field/environmental testing is the same for all modules. Following this, the next steps will be further validation in the field covering many different turbines of varying age, power etc. This would be level 6 and then moving from a laboratory system to a commercial system near ready to market (TRL levels 7 to 9).
In this regard, following the project review meeting and as suggested by the PO and the reviewer, it has been agreed with:
EPPR, our end-user to carry out prolonged (length of time TBD) testing and validation of the IntelWind CM system on the wind farms. This would be helpful in that the validation will be on different turbines to that used at CRES
CRES, too has indicated their willingness to participate in further validation of the IntelWind system. The previous validation done at CRES was on a 12 year old turbine, on this post review testing it is planned to do this on a different turbine as well as the previous turbine, in this matter data from the pre review meeting can be compared with post review meeting data
These actions will enable the IntelWind CM system to be at a TRL level of 7, 8 or 9 i.e. ready for commercialising (ready to market).
In readiness to market the competition i.e. systems currently available and their advantages and disadvantages have been reviewed (Deliverable D6.4) and 3 of the systems that we believe will be our main competition has been carefully evaluated, in terms of what is on offer, its applicability to a wide range of wind turbines and cost. Following this, the IntelWind modules on its own i.e. separately and the IntelWind CM system as a whole will be ready for marketing. Marketing entails an exploitation plan. This is discussed next.
The Exploitation plan is as stated in the PUDF (D6.4). With the project having come to an end, extra financial resources are needed to make the product ready for marketing and in this regard a number of technical actions are envisaged pre exploitation and will be the responsibility of the stated partner in Table 4.
The existing electronics, software and sensors have already been tested on operational capabilities in the main project timeline. In this post project period (table 4) the modules and system will go through a modification and/or redesign. Following this, further adjustments will be made to make the system sellable and Alpha Testing (Installation) will be performed on wind turbines at Romax and CRES, with a view to make the modules and CMS into a practical installable system. Easy Installation and Configuration will be paramount.
In the 2nd Phase (Beta Testing), the installation on commercial wind turbines (2-3 systems or more) will be undertaken. As we write both EDPR and CRES has agreed to evaluate/validate the redesigned/modified IntelWind CM system. During this phase, fault detection will be configured and alarms will be adjusted. With the installation on some critical turbines, we will be able to improve/adjust alarming and fault recognition to ensure that the CMS system is truly a ‘intelligent conditioning and monitoring system’. This brings the CMS to a saleable/marketable state.
In respect of market potential, figure 43, is an example of revenue cycle for a European onshore operator/utility. Note the x-axis shows quarters from Q4 FY2014/15 up to Q4 FY2018/19.
It is clear from figure 43, that the market potential is there at the correct price for a proven and validated system.
With Alpha and Beta tests complete (modules and systems at TRL level 7-9 i.e. at a mature state), the SME Romax who is working with various utility companies in Europe will use their existing connections to offer the Intelwind system to these companies (in particular E.ON Scottish Power, Infigen and AES). Our strong SMEs are well placed to exploit the wind turbine operators in UK, Belgium, Turkey and Spain, also Greece with the assistance of CRES (the Greek national entity for the promotion of renewable energy). Further it is hoped to target countries outside the consortium such as France and Romania.
Another route to commercialization that will be considered is the formation of the ‘Intelwind Ltd’. InnoTecUK will be able facilitate the formation of the company and the marketing of the technology with the SMEs who have direct contacts with the wind industry. The exact consortium agreement with regard to the Intelwind Ltd is yet to be agreed and will be done closer to system readiness to market.
Fuller details of the SME plans can be found in the deliverable D6.4.
The exploitation of the project will be measured in a variety of ways (stated above) and various partners will be involved, in particular the SMEs. The plan is to ensure consortium SMEs will benefit from the outcome of this IntelWind project in the form of financial enhancement to the SME, greater employment in the SME countries (in general in the EU) in respect of R&D (system development), service or maintenance staff in monitoring turbines, and secondary benefits to the supply chain. To the turbine operator and owner the benefit is from the expected savings in the cost of replacement of expensive equipment e.g. gearbox by using the IntelWind Intelligent Condition Monitoring system. Table 5 is an illustration of the minimum savings that could be realised by a CMS system, the IntelWind Intelligent CMS is expected to give a greater return.
Finally please note that a Patent: Torque sensor arrangements, Filing No. GB1221050.6 Filing date 22nd November 2012, Expected grant date 2016-2017, has been filed and if granted as expected will give the SME Transense and the consortium financial benefit as the technology is applicable not only to the Wind Industry but also to other industries that use motors, rotors and shafts.
List of Websites:
http://intelwind-project.com
