Blowing in the wind - monitoring wind turbine health
With the growth in demand for renewable energy, wind turbines must compete on cost and reliability with other energy sources. Dr Stefanos Giannis, Senior Scientist, and Peter Hansen, Senior Engineer, at the Materials Engineering Research Laboratory Ltd, UK, consider the requirements.
Renewable energy is an essential part of the UK Government’s plan to address climate change and reduce CO2 emissions. According to the Materials UK Energy Review published in 2007, the renewables obligation requires that 15% of the UK’s supplied electricity should be sourced from renewables by 2015. It is anticipated that approximately 75% of this will be generated from wind energy. Currently, more than 2,000MW of wind energy is generated in the UK (1.5% of the total UK electricity supply), but a further 6,000MW is required to meet the interim target set by electricity suppliers to generate 10% of their supply from renewables by 2010.
One of the main challenges for the wind energy industry is to continually reduce through-life costs. The average production cost for a one megawatt onshore wind turbine stands at 0.041 euros/kWh. However, additional savings can be achieved in the complete lifecycle. For example, operational costs can be reduced with more efficient wind turbine designs or by cutting maintenance and inspection costs.
Most wind turbines are horizontal axis and have two or three blades. Offshore turbines are generally larger than those onshore and generate more power. The UK’s turbine industry is moving towards large offshore farms, where it is anticipated that the cost of inspections will be high and they will need to be limited to the essential.
There are many advanced material systems of importance to wind turbines as well as more traditional materials, such as steel. These include high temperature superconducting materials that replace conventional copper rotor coils in generators, carbonised steel alloys in gearboxes, and stealth materials used to control radar signature.
Wind turbine blades vary in length from 25-60m, with newer designs potentially reaching 100m. They are designed for high longitudinal stiffness and torsion, and mainly comprise glass fibre-reinforced polymer (GFRP) composites, wood or foam, and significant amounts of adhesives. As blades become larger the need for higher specific stiffness (stiffness/weight) materials such as carbon fibre-reinforced epoxy composites (CFRP) is increasing.
Blades are designed for a lifespan of at least 20 years and will undergo hundreds of millions of rotation or fatigue cycles. During their lifetime they are expected to contain defects from manufacture, or be damaged during installation or in service. However, the industry does not routinely employ the state-of-the-art inspection and damage tolerance approaches developed for the aerospace industry.
In aerospace, many structural elements made of composites are designed using a building block approach, which involves increasing the testing complexity and size from coupon tests to full-scale structural tests that contain the most probable defect or damage. The main reason for employing such an approach is to reduce programme cost and risk while meeting all necessary requirements of damage tolerance. Cost efficiency can be achieved by designing a programme where more of the less expensive small specimens are tested, and fewer of the more expensive full-scale elements are required. The use of computational tools in place of some tests would also reduce cost.
An industry consortium was formed in 2005 to address inspection, structural health and damage tolerance procedures for large composite wind turbine blade spars. Partly funded by the UK Technology Strategy Board, the three-year project, entitled Maintaining Structural Integrity inYacht and Wind Turbine Spars (MSI-SPAR) was led by the Materials Engineering Research Laboratory (MERL) Ltd, in Hitchin UK. The project investigated the structural integrity of composite wind turbine spars where defects and damage might be present, with the aim of providing methodologies for the design and manufacture of damage tolerant structures, thus ensuring structural integrity while in service.
Non destructive investigations
It is often difficult to identify the location of damage initiation on wind turbine blades and methodical inspections are required using non-destructive inspection techniques. During the MSI-SPAR project, a RapidScan2 system from NDT Solutions Ltd was used by Testsure Ltd. The device uses ultrasonic phased arrays that can be configured to meet specific criteria for each application and is particularly suitable for the inspection of CFRP/GFRP monolithic laminates as well as carbon/glass hybrid laminates. It is capable of producing very high resolution A, B and C-scans in real time and can detect defects as close as 0.5mm to the surface. The project has shown that there is considerable potential for this technology for assessing wind turbine composite structures.
Structural health monitoring (SHM) systems, which are increasingly used in large structures such as offshore platforms, pipelines, bridges, tunnels, ships and aircraft, comprise a sensor network that monitors the structure in real time to detect damage or predict faults, enabling responses to prevent failure. To do this effectively, the SHM system must be reliable in its operation and in terms of the data presented.
The simplest implementation of a SHM system has a network of sensors that measures strains throughout the structure and compares these against pre-set limits based on the strain and loading history. Although this does not directly measure internal damage, it can monitor the response of the structure under load and, by using sufficient sensors, the load distribution throughout. Significant damage in the structure can give rise to an unusual strain distribution which can then be recognised by suitable data processing algorithms.
Fibre Bragg Grating strain sensors can measure strain and can be either embedded within the composite laminate or surface mounted, enabling use on a wind turbine blade to measure edge- and flap-wise deflection along its length. A typical strain measurement range is of the order of ±4,500 micro-strain with a resolution less than one micro-strain.
The MSI-SPAR project assessed embedded sensors within a composite structure and studied the effect of impact damage on the embedded fibres. With sound manufacturing procedures, the effect of the sensor deployment within a material is negligible. Once the fibre is imbedded in the material it is very robust, with weakness occurring only where the fibre exits the composite.
The ability of an FRP structure to continue to perform with damage present is defined as damage tolerance. Under the principles of damage tolerant design, some small damage/defects in the structure may be difficult to detect, but the blades are designed so that these defects do not threaten their structural integrity.
Part of the MSI-SPAR project was the development of damage tolerance methodologies for composite wind turbine blade structures. The risk areas of the various features of wind turbine spars were identified and the different threats noted, along with potential damage modes.
To verify the damage tolerance methodologies, scaled down spar sections were manufactured by international composite maker Gurit Ltd with pre-determined manufacturing defects such as voids in the corners of the spars and peel ply inclusions in the main flat sections. Finite element (FE) models of the blade spar structure were built by MERL, and representative in-service loads were applied to determine the boundary conditions around the embedded manufacturing defects.
Quasi-static and fatigue tests were performed by MERL and the results combined with FE predictions for the specific locations. It was predicted that for the particular design of spar at the specified location within the length of the blade, defects would not grow. For the derivation of the predictions, the change in energy for crack propagation (strain energy release rate), predicted from the numerical modelling, was compared with experimental measurements to confirm this.