CMSWind – Advanced Condition Monitoring System for the Assessment of Wind Turbine Rotating Parts
Case Study
By Antonio Romero
TWI played a key role in an EU project that has developed an advanced condition monitoring system (CMS) and methods of continuously monitoring rotating parts in wind turbines.
Figure 1 Bandirma Wind Energy Power Plant (WEPP)
Field trials have now been completed in CMSWind, a collaborative research project looking into the feasibility of using vibration analysis (VA), motor current signature analysis (MCSA) and acoustic emission (AE) in combination to monitor the condition of in-service wind turbines. The developed prototype was installed in a Vestas V90 3MW wind turbine in Bandirma Wind Energy Power Plant in Turkey, where it was validated.
Proposed solution
The CMSWind project produced an advanced system for condition monitoring of wind turbine rotating machinery components (Figure 2).
Figure 2 CMSWind condition monitoring system
Utilising three novel techniques, specifically designed for wind turbines and their components, the system improves wind turbine machinery reliability by up to 50%. This estimation is based on the fact that the system allows unnecessary maintenance and time out of service for wind turbines to be reduced or even eliminated, leading to improved reliability and operation. Vibration analysis (Figure 3 a), AE (Figure 3 b) and MCSA (Figure 3 c) techniques are respectively used to monitor the condition of the gearbox, rotary components and generator (which together account for 53% of wind turbine downtime). In this way, the CMSWind system monitors continuously the health of all the important rotating machinery within the turbine.
Figure 3 Transducers used: a) Accelerometer installed on the gearbox; b) AE sensor installed on the high-speed shaft; c) Current clamp to monitor the generator.
Field trials and results
The system was installed in the wind turbine for six months. Gathered data was stored in bins according to two criteria: power output and the rotor speed at the moment of acquisition. Figure 4 shows how the data is distributed in each case. Using the power criterion the data looks more equally distributed, with a larger number of files stored at the highest and lowest power outputs. If rotor speed is used, the CMSWind software indicates that most of the files were gathered at high rotor speeds.
Figure 4 Data distribution: a) Using power output (kW); b) Using rotor speed (rpm).
Results from the VA module using the second criterion are shown in Figure 5. In order to initially assess the status of the wind turbine, a study of the root mean square (RMS) evolution over time was carried out for each single bin. The sharper changes are caused to power output variations as depicted in Figure 5b. The signature does not show big deviations in any of the 30 bins analysed. When the CMSWind system had finished processing the data, it concluded that the health of the machinery had not been affected within the six months that the system was operational. That matched the information provided by the operator, who did not identify any kind of malfunction or failure during the same period of time.
Figure 5 Evolution of the RMS and power output in bin 29: a) RMS of the acceleration; b) Power output.
Once the machinery was confirmed to be healthy, a baseline representative of this healthy status was developed using the data gathered (Figure 6).
Figure 6 Baseline based on the RMS of the acceleration
During the baseline generation process the actual data was first sorted into the referred bin directory based on rotor speed. A specified number of data is collected within the bins for averaging. If the bin is complete, the RMS values of all the waveforms within that bin are calculated and averaged.
Alert system
The alert strategy designed for the VA process is shown in Figure 7. Parameters extracted from the new measurements are compared with the baseline limits and a warning is created if the parameters go beyond set limits. When a certain number of warnings is exceeded, an alert is generated indicating potential damage to the wind turbine. The algorithm is universal and can be applied to other types of wind turbine.
Figure 7 Alert strategy for vibration analysis
The CMSWind system is able to detect deviations from the healthy state and provide the probability that they represent a defect by using baselines similar to the one shown in Figure 6. The incorporated alert strategy effectively informs machine operators of potential damage to the wind turbine machinery as soon as it occurs.
For more information please visit www.cmswind.eu or contact us.
Antonio Romero
MSc electrical and electronics engineer
Antonio Romero is an MSc electrical and electronics engineer from Castilla La Mancha University (Spain). His master thesis topic was ‘Design and manufacturing of conditioning circuits using piezoelectric crystals’.
Antonio came to TWI in June 2013 as a placement student within the condition and structural monitoring section. In January 2014 he started a PhD granted by TWI and Brunel University London, researching ‘Condition monitoring in rotating machinery using vibration analysis’. In particular, he is trying to improve current methodologies for fault detection in gearboxes using vibration analysis.
Antonio has worked in three different collaborative projects, all of them linked to rotating machinery fault diagnosis: VA-RCM (health assessment of rotating machinery of train doors), REMO (condition monitoring of tidal turbines’ rotating parts) and CMSWind (condition monitoring of wind turbines’ rotating machinery). He was closely involved in the development of the condition monitoring systems and processing algorithms for fault detection. Apart from vibration analysis he uses acoustic emission for the same purpose.
