2.1. Image analysis and methodologies for determining the mechanical integrity of wind turbine blades.
Lead Partner: DU/Ørsted.
The main thrust of this PhD project will be to draw upon Durham’s established research expertise in developing innovative and novel new image analysis techniques. The aim of the workpackage will be to use this to develop new techniques for the detection of damage and surface degradation of wind turbine blades gathered from inspection vehicles or from remote locations (distances of order 100 m from the structures under inspection). The potential for using 3D imaging techniques capable of detecting faults in the sub-surface structure to depths of several millimetres will also be investigated in the study. The principal focus of the study will be on blade inspection, but opportunities to apply the new techniques to other parts of the wind turbine might also be considered.
New image analysis techniques available for detecting damage and surface degradation on wind turbine blades.
2.2. Integrating data from multiple on-turbine sensor networks to detect the most critical failure modes of wind turbine blades.
Lead Partner: UoS (DRG)/SGRE; Co-Partner: UoH.
The current trend in offshore wind is towards continuously increasing blade size (soon to exceed 100m); however, current industrial experience is showing that condition monitoring and structural health monitoring systems (CMS/SHMS) cannot rely on a single type of sensor for detection of all of the most relevant blade failure modes. This workpackage will focus on the design of a CMS/SHMS for monitoring of the most critical failure modes of the WT blades, occurring in the most structurally critical blade areas i.e. the root area. The main failure modes of interest are cracks and delaminations. A focus will be on selection of the most promising sensor technologies and design of a prototype measurement system for detection of the critical failure modes in the blade root area. Technologies considered will include acoustic emission, optical fibre transducers, guided ultrasonic waves, and potentially others. There will be close collaboration with Siemens’ Measurement Department, and the Siemens Blade Test Centre (Denmark) for this part of the research.
A further significant challenge is to determine optimal damage sensitive features for the SHM system based on data available from the Blade Test Centre and prototype WTGs. The success of any SHM strategy is critically dependent on these features; these should be low-dimensional vectors which concentrate information about damage, but are insensitive to benign operational and environmental variations. In this task, features will be optimised for the removal of any anticipated environmental and operation variations that would mask the presence of damage. New advanced machine learning algorithms will be developed to create robust feature data and new probabilistic models in combination with optimal combinations of diverse sensor technologies that will accommodate all anticipated sources of uncertainty
Design of prototype measurement system; proof of concept of SHMS for blade monitoring, algorithms and software for automatically-extracted optimal features for SHM, feasibility study report.
2.3 Integrated sensors for blade structural health monitoring.
Lead Partner: UoS (DRG)/SGRE, Co-Partner: UoH.
To develop analysis techniques for optimal extraction of SHM information from sensors embedded in blades and establish guidelines on the optimal placement of those optic sensors. The useful operational life of a wind turbine blade depends on the mechanical and thermal stresses it has experienced during manufacture, transport, installation and operation, while the repair or replacement of blades at sea is costly and dangerous and dependent on weather conditions. Current blades typically have little or no sensing capability and so there is very limited information available about blade stresses for use as an aid to predictive maintenance. There remain challenges around the process for embedding sensors during manufacture (addressed in WP 4.2) and extracting reliable and useful predictive information. This PhD will develop guidelines on the optimal placement of sensors within a blade (to be used in WP4.2) and develop SHM analysis tools and software to extract usable information from these sensors while eliminating background effects. Experimental data will be obtained from scale models of blades in the first instance, but if good progress is made in WP4.2, then a full scale blade with embedded sensors will be manufactured (outside of this project) and field data will be generated from either Siemens’ test site in Brande (DK) or one of Ørsted. Energy’s sites.
Design guidelines for the optimum placement of sensors within blades; SHM software that detects the onset of critical failure using embedded sensors and known failure modes, feasibility study report.
2.4. Improved modelling of wind turbine blade erosion on an individual turbine.
Lead Partner: DU/ Ørsted.
Ørsted Energy is interested in developing improved techniques for predicting erosion damage to the blades on offshore wind turbines during their service life. This PhD study will contribute to the work that they are undertaking to address this development need. Commercially available CFD and Solid Mechanics tools will be used to benchmark existing models for fluid-surface interaction models for erosion processes against blade erosion data owned by the Company. These models will then be further refined or new models proposed and validated against service data, that are optimised for predicting the erosion process under conditions that are typical of the offshore wind environment. The workpackage might also include some laboratory based testing in order to create new high fidelity erosion data under carefully controlled conditions to contribute to validation of the improved models.
Improved models for erosion of wind turbine blades in offshore environments.
2.5. Predicting critical failures in wind turbine blades by modelling populations of wind turbines.
Lead Partner: UoS (DRG)/SGRE, Co-Partner: UoH.
To evaluate if common patterns exist across WTs in a given farm and to determine optimal features for SHM based on historic data, so that critical failures can be predicted across the wind farm population in a more efficient way. This is a major shift away from standard CM/SHM damage detection strategies which focus on individual turbines. New ideas of population-based SHM offer the possibility that different types of data from individual turbines can be leveraged to give higher-level diagnostic information across an entire farm. Through data fusion, different types of data can be combined and analysed (i.e. low sampled SCADA data, high sampled CMS data, environmental data, alarm history, maintenance and inspection history, control strategy log, etc.) from individual turbines (healthy and damaged) in order to identify common patterns across whole farms. Methods for automatically updating diagnostic systems across farms from individual turbine data will be developed based on active/semi-supervised learning technology fused with advanced Bayesian inference. Sequential Markov chain models and Monte Carlo methods will be used to update generalised state-space models that will not only optimise diagnostic capability but allow access to prognosis.
Prognostic model/s and algorithm/s for wind farm critical failure/s prediction; model validation study; feasibility study report based on historic data from single wind farm.