Report name: Fatigue Life Prediction and Strength Degradation of Wind Turbine Rotor Blade Composites. Rogier Pieter Louis Nijssen, Aerospace Engineer

Transcription

1 Report name: Fatigue Life Prediction and Strength Degradation of Wind Turbine Rotor Blade Composites PhD Thesis report by University Rogier Pieter Louis Nijssen, Aerospace Engineer Delft University of Technology, Faculty of Aerospace Engineering, Delft (NL) Full report numbers ISBN-10: ISBN-13: number Published and distributed by and Partners Knowledge Centre Wind turbine Materials and Constructions (KC-WMC), Kluisgat 5, 1771 MV Wieringerwerf (NL) Design and Production of Composite Structures Group, Faculty of Aerospace Engineering, Delft University, Kluyverweg 1, 2629 HS Delft (NL) - Oil & gas company Royal Dutch Shell (NL/UK) - Power utility NUON (NL); - Wind farm developer E-Connection (NL) Date 27 November 2006 Fatigue Life Prediction and Strength Degradation of Wind Turbine Rotor Blade Composites Modern wind turbine rotor blades are subjected to a wide variety of continuously variable loads, and thereby show two striking characteristics as illustrated in Figure 1. The illustration shows on the horizontal axis some distinct composite material engineering applications that include from left to right a bicycle, passenger car, aircraft, helicopter, bridge, and a wind turbine. The vertical axis shows the matching number of significant load cycles experienced during the structure s operational life, plus the actual degree of load cycle variation.

2 Load Variability Number of load cycles Figure 1: Wind turbine blade loads characteristics Predictability of Loads and Response With regard to specific and extreme loading patterns, wind turbine rotor blades are subjected to a huge number of load cycles during the targeted economic lifetime. Estimates on the number of rotor blade load cycles vary from 10 8 to 10 9 during this 20- year period. These loads also show a high degree of variability, which by itself is a consequence of the wind s stochastic nature. Different fatigue loading components can further be determined, and their nature is either deterministic or non-deterministic. Load fluctuations The wind s vertical velocity profile causes regular load fluctuations. But even these regular load components are dependent on rotor speed and thus on wind speed. This behaviour is by nature largely irregular, but it is important to note that the latter observation does not apply to fixed-speed machines. A wind turbine rotor blade is loaded in flap-wise as well as edgewise direction (Figure 2). Despite ongoing efforts to integrate structural and aerodynamic optimisations, current rotor blade geometries can be subdivided into load-carrying components and other parts that are geometrically focused at aerodynamic performance optimisation. The main spar thereby carries flap-wise blade loads, while edgewise loads are absorbed by the spar and structural reinforcements that may be typically integrated into the blade s leading and/or trailing edges. The remaining sections of a given blade design generally comprise multiaxial skin material composed into a sandwich-type structure. This development strategy aims at ensuring aerodynamic geometry retention, sufficient torsion resistance, lowweight properties and a high buckling resistance. Figure 2 provides an indication of the main rotor blade loads in two perpendicular structural load-bearing directions.

3 Edgewise bending moment Flapwise bending moment Load Type / Lay-up Axial / Unidirectional Shear / Multidirectional Overall wind direction Buckling / Foam Figure 2: Typical rotor blade cross-section Flap-wise blade loading originates mainly from wind loads acting perpendicular on the rotor plane. These loads vary strongly in amplitude as well as mean. Edgewise loads by contrast mainly originate from a blade being loaded by its own mass, which is supplemented by the actual torque load that drives a turbine rotor. The loading direction for edgewise loads changes twice during each rotor revolution, but their specific nature is rather regular. Research findings further indicate that for small turbines the edgewise blade root bending moment frequency distribution contains two distinct peaks. A first peak load originates from actual wind loading, while the second peak load is a result of a blade being loaded by its own mass. Life prediction With larger rotor blades, edgewise gravity-based fatigue loading becomes increasingly relevant for life prediction. Fatigue loads in wind turbine blades show a certain degree of statistical variability. Among materials variability and other uncertainties, actual load variation is an important source of scatter in research findings. Additional research efforts focused on numerical wind modeling conducted on 0.5 MW class wind turbines featuring a choice of different control strategies. One of the key findings was that turbulence intensity is the most significant fatigue-load contributing factor. Related research revealed that extra fatigue loading due to terrain complexity issues is usually in the 30% range. Typically, rotor blades are composed of fibre-reinforced polymers. A combination of main reasons explains this materials choice, including the excellent stiffness properties, a high stiffness-to-density ratio, and more than adequate fracture toughness characteristics. Glass-fibre reinforced composites are today most commonly applied, although some blade designers now move towards employing carbon because the stiffer fibres are slowly becoming more affordable for these low-cost composite structures. In both cases, the fibre material grade applied in wind turbine blades is, compared to aerospace grade composites, characterised by relatively coarsely woven laminates. The fibres themselves are embedded in a polymer matrix, which provides some resistance to compression loads, but mainly serves as body material to align and geometrically fix the fibres. Essentially almost all large commercial rotor blades are today manufactured

4 with thermo-set resins. Polyester resin is applied for a large percentage of commercial rotor blades, but many blade suppliers today manufacture their blades from epoxy-based composites. The density of both materials is very similar, but superior fatigue performance is attributed to composites composed of epoxy-based matrices, which also potentially enables lighter blade design. Moreover, the absence of toxic styrene vapours during manufacturing is regarded as a major advantage linked to epoxy-based composites. Lifetime reliability Figure 1 indicates the degree at which rotor blade lifetime estimates are reliable. Basically, the predictability of loads and their response is inversely proportional to loads variability. In other words the more strongly actual loads do vary the less predictable they turn out and the more difficult it becomes to calculate the response (dynamic, fatigue) of both the materials and structure. However, reproducibility of the structure is also a main issue. Relatively small mass reproduced articles composed of a homogeneous material can be made to comply with the original design more easily, hence their behaviour prediction is superior. In terms of manufacturing challenge, wind turbine rotor blades are among today s largest series produced composite structures, while at the same time blade-to-blade variation has been characterised only to a limited extent. Fatigue loads and properties Fatigue loads are generally represented by a typical cyclic sinusoidal waveform. These waveform characteristics are represented in Figure 3, which describes the relation between the generic quantities S and N. For S, stress, load (per unit of width or thickness), strain, or displacement can be applied. Typically, stress as a term is applied for homogeneous materials, while the term strain is preferred for composites. N further represents the number of cycles leading to material failure. Alternatively, terms like half cycles to failure, number of load sequences to failure, or the number of cycles to predefined stiffness degradation can all be applied. Results of fatigue experiments are mostly plotted in an S-N diagram, thus showing the relation S versus N (i.e. Figure 3). In general, fatigue properties are determined mainly by the slope of the S-N curves, as indicated in the same figure. A flat S-N curve (small slope angle) is often considered to represent superior fatigue properties over a steep S-N curve. The appraisal of fatigue behaviour depends also on the location of the intercept with the abscissa. S max S mean S amp S range S S-N curve flat S min R=S min /S max steep N Figure 3: General fatigue terminology

5 For determining static strength and modulus, many industrial standards are readily available. These documents describe standard test conditions, specimen geometry for optimal performance and reporting format. For composites fatigue testing even to date only a limited amount of fatigue test prescriptions have become available. For instance ASTM has published a standard for composites tension-tension testing, while ISO has produced a general fatigue-testing standard. Some general observations on composite fatigue can be made. At first, by virtue of their flat S-N curve composite fatigue properties have shown to be superior to those of many other (competing) materials. S-N curve slope is generally applied as a fatigue resistance descriptor, and the composite material fatigue slope is thereby very small. This feature can be typified as good fatigue behaviour, since sensitivity to materials fatigue loads drops significantly when the load range drops. This observation simultaneously implies that a small load increase results in substantial fatigue life reduction. Moreover, composite materials fatigue life scatter is typically higher compared to metals, and this can only partly be explained by variations in test conditions. It should also be noted that the combination of a flat S-N curve and large scatter hampers fatigue life predictability. Second, a composite material s S-N curve is characterised by the absence of a fatigue limit. As part of this given fact no cyclic load amplitude has been found below which fatigue life is infinite. Less positive, composite materials are infamous for their poor resistance to compression type loads. But their compressive fatigue properties can be considered quite superior compared to the tensile fatigue properties. This relation is both visualised by the S-N curve slope and strength degradation. With regard to compression-compression fatigue behaviour the S-N curve is typically very flat. Another favourable characteristic in terms of compression fatigue is strength conservation. As experiments show, composite materials strength degradation in terms of compressive fatigue is rather insignificant. It is also no secret that composite fatigue behaviour has presented researchers and designers with substantial modelling difficulties. Modelling composite de-lamination can be accomplished using existing single-crack model approaches, but these do not necessarily turn out well with all composite laminate geometries. Fatigue-induced damage finally often leads towards multiple, distinctly different, and interacting damage types that toghether complicate modelling. Material of choice The steady increase in wind turbine rotor dimensions has resulted in blade deflection challenges. Scaling effects lead to a mass increase that is nearly proportional to the cube of blade length, which as a consequence leads to gravity-type load issues during blade design and operation. These challenges have initiated investigations exploring carbon fibre application into rotor blade structural designs. Other major design considerations include a switch to thermoplastic-based rotor blade materials. However, despite certain potential advantages like superior recycling performance, and a self-repairing built-in capability, thermoplastic basic fatigue performance has been reported to be inferior to equivalent thermoset-based laminates. Early observations on damage modes, life and strength prediction date from the late

6 1960s and were documented for various composites since the second half of the 1970s. Historically, fatigue life prediction methodologies were often derived directly from existing metal fatigue know-how base. Another historical aspect of composites fatigue involvement is its strong aerospace connotation. And even though material fatigue has long been considered irrelevant as a design driver, a majority of early publications deal with fatigue related issues of aerospace grade composites containing predominantly carbon fibres. Fatigue research From the early 2000s researchers identified improved strength prediction as one of the key requirements for more successful composite materials implementation in general. Research further strongly indicated that long-term (compressive) durability should be investigated more extensively and in greater depth. For all materials, methods of fatigue life prediction found in existing research for constant or variable amplitude fatigue can be roughly classified into four distinct categories: Cumulative damage Progressive damage Phenomenological/empirical (Micro-) Mechanics Cumulative and Progressive damage The term damage is generally applied for all kinds of physical materials deterioration notwithstanding eventual effects on materials performance. The Miner s Sum method is said to describe a linear accumulation of damage. However, the exact nature of this damage is neither specified, nor its effect on materials performance such as stiffness or strength. In practice, damage should be application-related. In pressure vessels, a useful damage parameter would be matrix cracks causing leakages. Damage can also be expressed in terms of stiffness degradation, or a probability of failure. And in reliability driven design projects, probability of failure can be applied as a failure definition. In tension-compression load cases a first occurrence of composite de-lamination prior to final failure can serve as main failure criterion. For tensile fatigue cases that results in component breakage it can act as a failure definition. In the early 1970s research showed that prior to failure by specimen separation, two other distinct failure mechanisms occur much earlier in operational life. These distinct failure mechanisms show themselves at first when fibre de-bonding from the resin occurs, followed by the cracks formation in the resin. When plotted in the S-N diagram, the onset of de-bonding and resin cracking may be described using S-N curve type formulations. In this research framework, failure is defined as the instance when a specimen can no longer bear the intended load. This results in practice into considerable specimen damage, or even worse to complete specimen part separation. The research effort further aimed primarily at comparing different composite materials and investigating various influences on material performance, as this seems an appropriate failure definition.

7 Phenomenological Phenomenological models share in common that a physically measurable parameter is being tracked during fatigue loading trials. Typically, these models describe strength or stiffness degradation during constant amplitude fatigue and reapply these descriptions in spectrum load fatigue life predictions. A major drawback of this specific category is that extensive experimental data accumulation is required to describe these models. At the same time no veritable guarantee is given to the question whether given model parameters found appropriate for one composite specimen can also be reapplied for accurate life prediction in other cases. Micro-mechanical Advocates of the micromechanical models by contrast attempt to describe fatigue-related composite degradation by starting from the basic material and basic ply properties. Usually, these models are characterised by considerable modelling detail in the constituents, and implementing rather complicated models describing different, progressively developing and interacting, failure modes. Furthermore with regard to fatigue, implementation of micromechanics modelling for fatigue life prediction is still in its exploratory phase. To date the success of micromechanical approaches for fatigue related research is limited to single failure modes. PhD research focus A life prediction assessment contains different, partly independent, elements: The counting method, used for describing variable amplitude signals as a collection of constant amplitude cycles; Formulas for describing S-N curves which relate these stresses to the number of load cycles required up to composite material failure; Constant life diagrams made up of S-N curves for different stress ratios; Damage rules related to stress history and specimen life expectancy. Regarding damage description, two models were investigated during the project and compared with each other: the earlier described Miner s Sum method and a strengthbased life prediction. As part of the former method, results of a counting method and a constant amplitude fatigue behaviour description were converted into a damage parameter, known as Miner s Sum. Potential load order effects were ignored during this exercise. Moreover, the damage parameter value only indicates whether or not an actual failure occurred, but this is unrelated to physically quantifiable damage. These are also actual limitations to the model that have become suspect of causing inaccurate predictions. In a strength-based method by contrast, component life is predicted by calculating the effect of each load cycle on strength properties, until the actual load exceeds the remaining strength capability. An expected advantage of this cycle-by-cycle method is that sequence effects can be implicitly included. Moreover, the damage parameter is at all times related to a physically quantifiable parameter (viz. strength). Successful application of this strength-based method requires a description of post-fatigue strength, which by itself entails substantial experimental effort. In addition, a strength-based life prediction is much more computationally intensive than Miner s sum and cannot always rely on the

8 same counting methods. PhD research results summary During a Miner s sum and a strength-based method comparison, influences and significance of other life prediction elements, such as counting methods and description of constant amplitude fatigue behaviour on life prediction were included. The experimental part of the research involved a substantial number of static, and fatigue material tests. With respect to variable amplitude loading, both block loading sequences and (variants of) the WISPER spectrum were applied to the specimens. The tests aimed at providing a detailed image of static strength properties, constant and variable amplitude fatigue behaviour, as well as strength degradation for different glass-fibre reinforced laminates. A consistent database was thereby created by selecting single coupon geometry for combined material tests on a single material, an effort that included the definition and creation of standardized test conditions. Block-test experiments conducted confirm the existence of sequence effects on design life, even though more data are required as a necessary precondition to fully quantify these outcomes. Residual strength tests performed show strength degradation after fatigue-related exposure for a large range of fatigue load conditions. Significant tensile strength degradation could be observed in R=0.1 and R=-1 material fatigue related experiments. Generally, compressive strength remains within the boundaries of the initial static strength distribution. This favourable behaviour was observed for different laminate samples. The significance of an adequate constant amplitude behaviour description becomes evident from various life predictions. Commonly used simplifications, such as the Linear Goodman Diagram, result in a high degree of non-conservative predictions. The residual strength model yields more conservative predictions compared to Miner s sum for the investigated tension-dominated load sequences. Experimental efforts required for the determination of strength degradation, supplemented by computational efforts do not justify this relatively small advantage. The description of the constant life diagram showed to be the most influential on life predictions. Therefore, for design and future research, it is recommended to apply a description of a constant life diagram with as much detail as possible. And in addition accurately take into account the influence of loading type on fatigue life. The research effort finally focussed on wind turbine rotor blade composite material fatigue behaviour. Nevertheless, the results found are also relevant for other composite structures that are loaded in fatigue.