3.1. Determining the condition of lubricated components through monitoring of oil condition.
Lead Partner: DU/Ørsted.
This task will build upon an earlier collaboration between Ørsted Energy and CC Jensen to develop new methodologies for diagnosing the condition of components, by monitoring the condition of the oil lubricating them. The task will aim to develop new, improved, diagnostic tools which uses SCADA and other online monitoring data. An overall strategy for this type of online monitoring will be developed including the specification of appropriate alarm limits. The work will involve the design, manufacture and commissioning of a new test facility at Durham University which will be used to measure high fidelity experimental data under tightly controlled test conditions. This data will be used in conjunction with the turbine field data to validate the new diagnostic methodologies.
A new diagnostic methodology for monitoring the condition of oil lubricated mechanical wind turbine components.
3.2. Condition monitoring of wind turbine bearings through fusion of vibration-based health monitoring and acoustic emission (AE) features.
Lead Partner: UoS (DRG)/SGRE, Co-Partner: UoH.
This task will be concerned with optimised feature extraction from low and high-frequency data. As the different sampling rates reflect different physical processes, a multi-rate feature extraction can potentially yield far greater diagnostic and prognostic information. New methods and machine learning algorithms of operational evaluation specific to multi-rate systems will be developed. Data fusion technology promises the generation of optimal diagnostic features for SHM and CM; however, there are many current gaps in the technology and a lot of significant issues to be addressed giving space to novel blue sky technologies for the first time. A particular issue relates to the highly disparate timescales between vibration-based
CM/SHM and active/passive ultrasonic CM/SHM. AE data and general ultrasonic systems produce very large amounts of data continuously and this introduces a major issue of big data that has to be addressed practically and with computational efficiency.
The focus of these challenging PhDs will be on the design of a prototype CM system that will take advantage of data fusion technology for monitoring of the wind turbine bearing components, (i.e. main bearing and blade bearing) that will need a number of new algorithms and signal processing tools that simultaneously address the problem of feature fusion across disparate timescales and big data
Design of prototype AE-based measurement system; proof of concept of CM/SHM system for monitoring of the turbine bearings; optimised automated feature extraction for CM/SHM; algorithms/ software for feature extraction, feasibility study report.
3.3. Pitch and rotor system level fault monitoring.
Lead Partner: UoH.
This task is concerned with on-line fault monitoring of the rotor system for offshore wind turbines, including (a) individual pitch system monitoring and (b) blade root bending moment sensor monitoring. For (a) the pitch system monitoring includes the detection, isolation and estimation of faults in each of the three pitch actuators and corresponding pitch sensors. Pitch actuator fault examples include actuator stuck faults, hydraulic leakage, pump wear and also high air content in the hydraulic fluid, as well as actuator bias. Examples of pitch sensor faults are: sensor bias, total sensor failure, stuck and erroneous measurements. Excessive bending at the rotor hub leads to the development of both bearing and rotor damage, and eventual fatigue. Bending moment measurements are receiving increased attention because of their importance in indicating the rotor health and also for mitigating the unbalanced loading of the rotor. Hence, for (b) faults in the blade root bending moment measurements are estimated to provide some measure of the degree of rotor load imbalance and developing fatigue. Information about all of the faults in (a) and (b) are important “system level” indicators of rotor system health and performance which are valuable for predictive maintenance. In all of this work attention is to be paid to the on-line detection and isolation or estimation of “incipient faults” which are normally hard to detect, so that fault estimation can be provided at an early stage before the onset of system failure or shut down. Incipient faults can be monitored using robust model-based methods that are well-known in the Control Systems community. The robustness involves a suitable discrimination between actual fault effects and variations caused by modelling uncertainty and disturbance.
The proposed monitoring methods are based on a combination of Fault Estimation (FE) and Fault Detection and Isolation (FDI) (well-known control system monitoring). The verification and validation of the proposed fault monitoring methods should be carried out in a suitable high-fidelity wind turbine benchmark, using statistical Monte Carlo campaigns that replicate a suitable range of wind turbine operating conditions, with and without faults. If possible the proposed fault monitoring designs would be verified and validated on a platform wind turbine system, based on standard control system operation.
Descriptions of robust design and Monte Carlo evaluation strategies for pitch actuator fault detection and estimation will be the subject of as external publications. The work will consider various pitch actuation and pitch sensing faults. Pitch blade bending moment sensor fault monitoring will be developed and evaluated to provide a method of indicating the extent of blade bending, as intended information for predictive-maintenance and also for use in on-line unbalanced load mitigation control. The work on this will be a further topic for external publication and dissemination to the project community.
3.4. The original themes for 3.3 and 3.4 were discontinued and replaced by a single new project on fault monitoring and wind turbine rotor and blade switch systems (see 3.3).
3.5. Generator and converter fault analyses, including open- and short-circuits, and health and condition monitoring.
Lead Partner: UoS (EMD)/SGRE, Co-Partner: DU.
To develop robust approaches identifying fault conditions and their prevention/isolation. Faults within the generator and/or the converter can cause catastrophic failures including irreversible demagnetisation of permanent magnets, overheating and converter damage, therefore to improve reliability, it is essential to design machines to withstand faults and develop converters and condition monitoring techniques to identify and isolate faults. The analyses will include open-circuit, short circuit, inter-turn short-circuit faults, as well as potential irreversible demagnetisation of permanent magnets, with critical assessment of the benefits of modular and balanced designs compared with the conventional non-modular designs. New fault modeling techniques considering the materials, losses, thermal, transient electromagnetic characteristics will be developed. Novel condition monitoring techniques will be developed by considering the manufacturing and operating tolerances (rotor static and dynamic eccentricities). Health and condition monitoring will be applied to both generators and converters, accounting for their interactions.
Reports on fault analysis techniques, health and condition monitoring techniques for both generators and converters.