Functionally Graded Adhesive Patch Repairs in Civil Applications

Functionally Graded Adhesive Patch Repairs in Civil Applications Guilherme Miranda Silva de Oliveira Viana Master Thesis Report Supervisor: Prof. Lucas F. M. da Silva Co-Supervisor: Eng. Ricardo Carbas Faculdade de Engenharia da Universidade do Porto Mestrado Integrado em Engenharia Mecânica June 2013 i i Abstract Several investigations have been made concerning the fracture behaviour of scaled specimens of wood beams repaired with adhesively bonded carbon fibre reinforced plastic. However, one of the problems associated to these joints is the fact that the stress distribution (shear and peel) is concentrated at the ends of the overlap, leading to premature failure of the joint. Some solutions to this problem have been developed, such as hybrid joints, adherend shaping, adherend rounding and fillets at the ends of the overlap. Some of these methods tend to increase the weight of the structure and others are very expensive due to its complex manufacturing process. The stress concentration can be reduced with use of a functionally graded adhesive, in which, the mechanical properties vary along the bondlength. This can be achieved with a graded cure, in which the temperature varies along the bondlength. In order to perform a graded cure, induction heating was used. This technique has already been successfully tested in single lap joints to obtain a more uniform stress distribution along the bondlength, increasing the strength of the joint. In this project, the repair of wood structures with CFRP was made using a homogeneous cure and a graded cure. Two common types of defects on beams under bending solicitations were analysed. Scaled specimens of damaged wood beams were repaired and tested under four point bending. The results show that the beams repaired with a graded bondline were able to withstand higher loads. ii iii Reparação de Estruturas Usando Ligações Adesivas Graduadas Resumo Têm sido feitas várias investigações sobre o comportamento mecânico de vigas de madeira à escala, reparadas com compósitos de fibra de carbono. Contudo é sabido que um dos problemas associados às ligações adesivas se prende com a concentração de tensões tanto no adesivo como nos substratos (no que concerne à tensão de corte e tensão de arrancamento) e pode levar à rotura prematura da junta. Várias soluções têm sido desenvolvidas para minimizar este problema, tal como o método dos dois adesivos, arredondamento das arestas dos substractos, utilização de filetes, etc. Alguns destes métodos tendem a aumentar o peso da estrutura ou são demasiado dispendiosos devido à complexidade do processo de fabrico. O problema da concentração de tensão pode ser minimizado com recurso a um adesivo graduado, em que as propriedades mecânicas variam ao longo do comprimento de sobreposição. Isto pode ser conseguido com uma cura graduada, em que a temperatura de cura varia com o comprimento de sobreposição. Para a obtenção de uma cura graduada, o adesivo é aquecido por indução nas zonas de maior concentração de tensão, ficando, nessas áreas, com uma maior flexibilidade e ductilidade. Esta técnica já foi testada com sucesso em juntas de sobreposição simples e foi utilizada novamente na reparação de vigas em madeira com fibra de carbono. Os resultados da reparação da viga com junta graduada são comparados com os da viga com cura isotérmica. Foram analisados dois tipos de defeito comuns em vigas de madeira solicitadas à flexão. Foram fabricadas vigas de madeira à escala que tentam reproduzir esses mesmos defeitos. Esses provetes foram testados em flexão em quatro pontos. Os resultados mostram que as vigas reparadas com o adesivo graduado são capazes de aguentar cargas mais elevadas que as vigas reparadas com o adesivo homogéneo. iv v Acknowledgements I would like to express my gratitude towards my supervisor Lucas F. M. da Silva and cosupervisor Ricardo Carbas for their guidance, technical contribution, dedication and writing review during this work. I equally wish to thank engineers Eduardo Marques and Raul Campilho for their help with the finite element software Abaqus®. Thanks to engineer Filipe Chaves for his help with the DCB and ENF tests. Acknowledgements to my family and all my friends for their fellowship and encouraging support. Finally I would like to express my sincere gratitude towards Sara Duque, who has supported me through all stages of this work. vi vii Table of Contents 1 Introduction .......................................................................................................................................... 1 1.1 Background and Motivation .................................................................................................................. 1 1.2 Objectives ............................................................................................................................................ 3 1.3 Research Methodology ........................................................................................................................ 4 1.4 Outline of the Thesis ............................................................................................................................ 4 2 State-of-the-Art Review ....................................................................................................................... 5 2.1 Wood Structures .................................................................................................................................. 5 2.1.1 Common Defects in Wood Structures ................................................................................ 5 2.1.2 Repair of Wood Structures ................................................................................................. 7 2.2 Adhesive Joints .................................................................................................................................. 11 2.2.1 Factors Affecting the Strength of Adhesive Joints ........................................................... 11 2.2.2 Methods to Increase Joint Strength ................................................................................. 14 2.3 Strength Prediction ............................................................................................................................. 18 2.3.1 Stress/Strain Based Criteria ............................................................................................. 18 2.3.2 Fracture Mechanics Based Criteria .................................................................................. 18 2.3.3 Cohesive Zone Models .................................................................................................... 19 3 Characterization of the Adhesive ...................................................................................................... 23 3.1 Manufacture of the DCB and ENF specimens .................................................................................... 24 3.2 Pure Mode I Toughness ..................................................................................................................... 25 3.2.1 Results ............................................................................................................................. 26 3.3 Pure Mode II Toughness .................................................................................................................... 28 3.3.1 Results ............................................................................................................................. 29 3.4 Summary of Results ........................................................................................................................... 30 4 Repair of Wood Structures ................................................................................................................ 31 4.1 Mechanical Properties of the Pinus Pinaster Wood ........................................................................... 31 4.2 Geometry ........................................................................................................................................... 33 4.3 Specimens Manufacture..................................................................................................................... 34 4.3.1 Isothermal Cure ............................................................................................................... 35 4.3.2 Graded Cure .................................................................................................................... 36 4.4 Manufacture of the CFRP Patches ..................................................................................................... 37 4.5 Specimens Testing ............................................................................................................................. 38 4.6 Numerical Analysis ............................................................................................................................. 39 4.6.1 Stress Distributions in the Adhesive Layer....................................................................... 42 4.7 Strength Results ................................................................................................................................. 45 4.7.1 Undamaged Beam ........................................................................................................... 45 4.7.2 Compression Damage Specimens .................................................................................. 46 4.7.3 Cross Grain Tension Specimens ..................................................................................... 50 4.8 Discussion of the Results ................................................................................................................... 53 5 Conclusions and Future Work ........................................................................................................... 57 6 References ........................................................................................................................................ 59 xiv xv Nomenclature Acronyms CBBM Compliance Based Beam Method CFRP Carbon Fibre Reinforced Plastic CZM Cohesive Zone Model DCB Double Cantilever Beam ENF End Notched Flexure FEM Finite Element Method FEUP Faculdade de Engenharia da Universidade do Porto FPZ Fracture Process Zone FRP Fibre Reinforced Plastic GFRP Glass Fibre Reinforced Plastic 2D Two dimensions Symbols E Young’s Modulus G Shear Modulus P Load ν Poisson’s ratio δ Displacement GIC Mode I fracture toughness GIIC Mode II fracture toughness τ Shear stress τavg Average shear stress σ Peel stress xvi 1 1 Introduction 1.1 Background and Motivation Wood is one of the most ancient building materials, however, nowadays, in Portugal, its use in civil structures is not a common practise. Other countries, such as the United States of America, Russia, Norway, Sweden, Finland, due to the many advantages that it presents, extensively use wood in civil construction. Figures 1-5 show different examples of wood constructions. Figure 1: Church in Kizhi, Russia. Figure 2: Leonardo Bridge, Aas, Norway. 2 Figure 3: Houses, Alborg, Denmark. Figure 4: Stadium, Vantaa, Finland. Figure 5: Pavilhão Atlântico, Portugal. Wood is an inexpensive building material that can provide easy to build structures. Its thermal conductivity is low, providing good thermal isolation. It is also a good acoustic isolator. It has good specific mechanical properties (divided by its weight). However, if not properly preserved, it is easily degraded by atmospheric and biological agents, such as fungus and insects [1]. It is also susceptible to moisture changes and fatigue for low stress differential [2]. Little pollution is created during wood production. It is a renewable and recyclable product. Replacements of large timber sections are very expensive due to the complexity of the process and the limited availability of the material. When possible, damaged timber beams should be repaired rather than replaced. 3 Timber structures repairs using structural adhesives are both economically and structurally efficient. Among the several types of adhesives, epoxy adhesives are the best for this kind of joint as they do not require high pressure during application, they exhibit good adhesion to wood and several other materials (such as fibre reinforced plastic), little shrinkage during cure and are highly resistant to moisture and chemical products [3]. Several authors have witnessed the considerable improvement of the strength and stiffness of timber beams repaired with bonded Fibre Reinforced Plastic (FRP) patches [4-8]. However, one of the problems associated with this kind of joint is the stress concentration in both the adherends and the adhesive, which causes the premature failure of the joint [4][9]. In order to reduce this problem, several techniques have been developed, such as hybrid joints, adherend shaping, adherend rounding and fillets at the ends of the overlap, graded. These techniques are usually very expensive to implement due to the excessive steps in the manufacture process and tend to increase the weight of the structure. It is known that graded adhesives can provide a significant improvement in the strength of adhesive joints [10]. Many epoxy adhesives have different properties depending on their temperature of cure [11]. If the temperature of cure may vary along the overlap, a gradient in the rigidity of the adhesive can be obtained. This can be used to make a functionally graded adhesive joint. 1.2 Objectives The main objective of this research was to investigate the performance of a functionally graded adhesive in the repair of scaled damaged wood beams. The specific objectives are listed below:  Determine the toughness of the adhesive (mode I and II) for three different temperatures of cure. This can be made using DCB and ENF specimens. The energy release rate can be computed using the Compliance Based Beam Method (CBBM) method [12-13], a recently developed technique that does not require the measurement of the crack length during propagation;  Simulate the mechanical behaviour of the repaired beams using the FEM. CZM can be used in order to simulate crack initiation and propagation in both the adhesive and the wood;  Conduct bending tests on damaged and undamaged beams, beams repaired with the functionally graded bondline and the beams repaired with an adhesive with constant properties. 4 1.3 Research Methodology In order to achieve the objectives of this investigation, the following method was used:  A literature review on the repair of wood beams, methods to increase the strength of adhesive joints and on strength prediction was made;  Double Cantilever Beam (DCB) and End Notched Flexure (ENF) tests were made in order to assess the fracture toughness of the adhesive for three different temperatures of cure;  The beams repaired with the functionally graded bondline and the beams repaired with an adhesive with constant properties were tested under four point bending. The results were compared;  The repaired beams were numerically simulated using the Finite Element Method (FEM) and Cohesive Zone Models (CZM). 1.4 Outline of the Thesis This thesis is divided into five chapters:  Chapter 1 is an introduction to the proceedings used and the format of the thesis;  In Chapter 2, a review on the state of the art is made. It focus on the repair of wood beams, methods to increase the strength of adhesive joints and methods for strength prediction;  In Chapter 3, the experimental work concerning the determination of the energy release rate of the adhesive in modes I and II is presented;  Chapter 4 focuses on the manufacture, test and numerical simulation of the wood beams;  In Chapter 5 the final observations and references to possible future investigations are made. 5 2 State-of-the-Art Review 2.1 Wood Structures 2.1.1 Common Defects in Wood Structures Natural defects in wood include mainly knots, rots, wanes, shakes, checks and splits (Figure 6) [14]. Knots act as discontinuities in the wood, raising the stress concentration factor around it. In a beam under bending, knots should be placed in the compression face rather than in the tension face, as its contribution to reducing the strength of the beam is lower [1]. A wane is lack of wood on the edge or corner of wood members and shakes are lengthwise separations in the wood occurring between annual rings [1]. Checks and splits are cracks that occur along the direction of the tree’s growth and are caused by shrinkages and swellings due to changes in the moisture of the wood [1][15]. This kind of crack acts severely as stress concentration points especially if the beam is under bending. Heart checks (checks that extend deep into the interior of the wood) expose the inner part of the wood member to fungus and insects, causing rots and lowering strength [16]. 6 In-service damage of wood includes mainly compression failure, cross-grain tension and shear failure [1]. Compression failure is typical of low density wood or beams reinforced in the tension face, as the change of the neutral axis provides a larger cross section to bear the compressive load [1]. Cross-grain tension is common in beams whose fibres are not aligned with the main axis (Figure 7). This happens when the tree that is cut for lumber grows spirally or with a pronounced taper. The crack that is originated has usually an angle with the beam’s axis of less than 15 degrees [1]. Figure 6: Natural defects in wood members [1]. 7 Horizontal shear cracks normally propagate from checks and shakes (Figure 8). If the beam is under extreme low humidity or shows sharp changes in growth ring density, this kind of failure is more likely to occur [1]. Tension failure can also occur if there is a knot in the tension face of the beam, as it acts as a point of stress concentration [5]. Figure 7: A cross grain tension crack [1]. Figure 8: A horizontal shear crack [1]. 2.1.2 Repair of Wood Structures Radford et al. [17] studied the effect of nails, shear glass fibre rods and plates in the repair of wood beams. Holes were drilled in the wood beam and shear spikes impregnated in epoxy adhesive were inserted, reinforcing the beam in shear. The adhesive was used to improve the load transfer from the wood to the spike, exclude the water from the interface, due to its compatibility with the materials and also to spread into the gaps the beam might have. As the spikes are inside the wood, the aesthetics are only marginally affected and no preservatives can possibly affect the quality of the joint. 14 0 0 0,2 0,4 0,6 0,8 1 τ,σ x/L Long Overlap Short Overlap 2.2.1.4 Overlap Length In single lap joints, for example, peak stresses occur at the edges of the overlap. When rupture occurs, the stress concentration points are under the maximum stress while the middle of the overlap is underloaded [25] especially if the adhesive is stiff and brittle. This means that the total area of the overlap is not fully efficient. This phenomenon gets more critical if the overlap is long. In this case the stress may even become null, which means that if the length is further increased, no improvement will be achieved in the load bearing area and, consequently, on the strength of the joint [25] (Figure 19). This becomes critical if brittle adhesives are used. If the adhesive is very ductile, the strength of the joint is approximately proportional to the overlap length, as it has the capability to deform and redistribute the stress evenly as the load increases [9]. 2.2.2 Methods to Increase Joint Strength 2.2.2.1 Spew Fillets The spew is the portion of adhesive that is squeezed out of the lap as the two substrates are assembled (Figure 20). It is known that the stress concentration in the adherend-adhesive interface is reduced with the inclusion of these spews in the joint [26]. The reduction in the stress concentration factor is function of the ductility of the adhesive and the adherend properties and also of the geometry of the spew [9][27]. Several spew geometries have been tested by Lang and Mallick [27], including triangular, full triangular, square, half rounded, full rounded, full rounded with fillet, oval and arc. Figure 19: Typical stress distribution in a single lap joint 15 Despite being an efficient method to improve the strength of adhesive joints, in order to create spew fillets, several manufacture steps are required, thus increasing the cost of the bonding process. The use of this technique can also create more thermal stresses when used at low temperatures [9]. Figure 20: A triangular spew fillet. 2.2.2.2 Adherend Shaping Adherend shaping can reduce the stress concentration at the edges of the overlap, strengthening the joint (Figure 21). If the stiffness of the joint in the stress concentration points is reduced, a more uniform stress distribution can be obtained [9][18]. This technique can also be combined with the use of spew fillets. Adherend shaping techniques such as internal or external tapers are efficient in reducing the peel stress in adhesive joints. This is very important especially when FRP adherends are used [18], as their transverse strength and consequently their resistance to peel stresses, are very low. Due to its high cost, this technique is often not possible to be used. Figure 21: Different adherend shapes [9]. 2.2.2.3 Adherend Rounding If the adherend corners are rounded (Figure 22), the stress in the singular points (points where the stress concentration factor is infinite), may be reduced, increasing the joint strength. This effect is only local, as the stress distribution away from the rounded area is not affected 16 [9]. These singular points do not exist in practical, as the adherends are always slightly rounded in the manufacturing process. This phenomenon has been studied by Adams and Harris [28], who found that the point of maximum stress is slightly dislocated away from the corner. Zhao et al. [29] concluded that the improvement of the joint strength is smaller if ductile adhesives are used. Using a brittle adhesive an improvement in joint strength of 40% was achieved. Figure 22: Adherend rounding [9]. 2.2.2.4 Mixed Adhesive Technique A technique that is also used to reduce the stress concentration at the ends of the overlap is the mixed adhesive technique. Instead of using a single brittle adhesive along the entire overlap length, two adhesives are used. This second adhesive must be more flexible and is placed at ends of the overlap. This way, the stress at the edges of the bondlength, is decreased, resulting in a more uniform stress distribution and in a stronger joint [9][30]. One disadvantage of this kind of joint is the adhesive separation. Despite being difficult to find compatible adhesives, the best way to control the process is still to use film adhesives. Another way to do this is to use silicon strips. However, if this technique is used, there is a small decrease in the load bearing area [9]. Figure 23: The mixed adhesive technique [9]. Brittle adhesive Ductile adhesive 17 2.2.2.5 Graded Materials In order to decrease the stress concentration at the edges of the overlap, functionally graded materials can be used. The idea is to have an adherend or adhesive whose mechanical properties may vary along the bondlength. If the stiffness of the joint can be changed along the bondlength, a more uniform stress distribution can be obtained, thus strengthening the joint. Apalak and Gunes [31-32] and Gannesh and Choo [33] have used finite element analysis to simulate joints with functionally graded adherends, however, to the author’s knowledge, no joint has ever been experimentally tested. Stapleton et al. [10] studied the effect of functionally graded adhesives in single strap joints. Different concentrations of glass beads along the bondlength were used in order to create a gradient in the rigidity of the adhesive. A significant increase in the joint strength was achieved. Carbas et al .[34] obtained a variable modulus bondline through the use of a graded cure. The graded cure, which allowed the adhesive to have different properties along the bondlength, was made with induction heating. A considerable increase in the strength of single lap joints was achieved. The adhesive was modified to show a gradient in the rigidity along the overlap, being its stiffness maximum at the middle and minimum at the ends. This modification was achieved with a differentiated cure process. The temperature of cure was not uniform along the entire bondlength, causing the adhesive to develop different mechanical properties along the overlap. In order to obtain different temperatures at the middle and at the end of the overlap, a special apparatus was invented [35]. Single lap joints were tested with two different adhesives (Loctite Hysol® 3422 and Araldite® 2011) and with three different kinds of cure [34]:  Low temperature cure  High temperature cure  Graded Cure The results showed that the graded cure was able to increase the strength of the graded Loctite Hysol® joint by 210% in relation to the joints cured at low temperature (brittle behaviour) and by 62% in relation to the joints cured at high temperature (ductile behaviour). The strength gain of the graded Araldite® joint was of 70% in relation to the high and low temperature cure [34]. 18 2.3 Strength Prediction 2.3.1 Stress/Strain Based Criteria These criteria are based on the analysis of stresses and strains of the structures. With the help of the FEM or analytical models, it is possible to know the stress and displacement field around a certain point. However, real structures have points where the stress concentration factor tends to infinite. In these singular points, the solution provided by the FEM is highly mesh-dependent and not accurate. This kind of problem can be minimized with the use of the point stress criteria [36], in which the stresses are computed at a pre-defined distance from the singular point, and with the use of average stress criteria [36], in which an average stress is computed along a determined path. 2.3.2 Fracture Mechanics Based Criteria Unlike the Stress/Strain based criteria, the fracture mechanics based criteria have the ability to determine, taking into account the existence of singular points in the structure, when a crack may start to propagate. There are two kinds of formulation in the fracture mechanics based criteria: The one based on the stress intensity factor ( K ) and the one based on the energy release rate ( G ). 2.3.2.1 The stress intensity factor The stress intensity factor ( ) is defined as: √ ( ) Figure 24: Mode I crack solicitation. 19 In which is the remote tension and is the crack length (Figure 24). Crack propagation occurs as soon as equals . is the critical stress intensity factor, a property of the material and a measure of its toughness. is a dimensionless correction factor that depends on the mode of solicitation and geometry. There are three pure modes of solicitation, as can be seen in Figure 25. i) Mode I crack: Opening mode ii) Mode II crack: Sliding mode iii) Mode III crack: Tearing mode 2.3.2.2 The energy release rate The energy release rate is defined as: represents the work of external forces, represents the deformation energy and represents the area of propagated crack. As in the previous case, there is crack propagation when equals , the energy release rate, a property of the material. The energy release rate is related to the stress intensity factor: i) For plane stress: ( ) ii) For plane strain: ( ) ( ) Where and represent the Poisson’s ratio and the Young’s modulus respectively. 2.3.3 Cohesive Zone Models These models have the advantage of combining the stress/strain based criteria with fracture mechanics, accurately predicting the behaviour of the materials. CZM can predict the formation and propagation of cracks [36]. (2) Figure 25: Modes of crack solicitation. I II III 20 As soon as, in a given node, the strength of the material is reached, softening initiates. Depending on the properties of the material, several cohesive laws can be used to simulate the softening of the material. These include triangular, linear-parabolic, polynomial, exponential and trapezoidal laws. Although the cohesive laws can be adjusted to better fit the behaviour of the material, the triangular CZM, due to its simplicity, is very widely used and provides good results for most of the real situations [37]. In this project, every simulation was carried out with the use of a triangular cohesive zone model available in Abaqus® (Figure 26). The triangular cohesive law has an initial elastic behaviour. After the maximum stress is achieved, linear softening initiates. When the stress reaches the value of zero, no load can be transmitted, which is the same as saying that a crack has been created. The elastic domain is defined by a constitutive matrix [K] containing the stiffness parameters: { } [ ] { } ( ) For thin adhesive layers, the following approximations can be used [38]: ( ) ( ) ( ) For the damage initiation, the quadratic criterion was used [38]: {〈 〉 } { } ( ) Figure 26: Triangular cohesive zone model available in Abaqus®. Traction Separation Mixed Mode Law (m) Pure Mode Law -Tensile (n) -Shear (s) 21 〈 〉 are the Macaulay brackets, which indicate that compressive loads do not contribute to the damage initiation. The separation of the material was predicted using a linear energetic criterion [38]: ( ) 22 23 3 Characterization of the Adhesive In order to obtain a functionally graded adhesive, a graded cure was used. This means that the adhesive must have different mechanical properties (particularly the stiffness and ductility) for different temperatures of cure. The greater the range of these properties, the more effective the graded joint can potentially be. Carbas et al. [11][39] studied the tensile properties of different epoxy adhesives as a function of the temperature of cure: Araldite 2011® (Huntsman, Basel, Switzerland), Araldite® AV 138M (Hunstman, Cambridge, England), Sikadur®-30LP (Sika, Zurich, Switzerland) and Locktite Hysol® 3422. Of these adhesives, the one with the widest range of Young’s modulus is the Locktite Hysol® 3422 adhesive (Figure 27) and is therefore the one chosen for this project. Figure 27 Mechanical properties of the Locktite® Hysol 3422 adhesive. In order to accurately simulate the behaviour of the repaired wood specimens, the fracture toughness of the adhesive in pure modes I and II for three temperatures of cure was determined. For this, the DCB (pure mode I) and the ENF (pure mode II) tests were used. 0 20 40 60 80 100 20 40 60 80 100 Failure strain (%) Temperature of Cure (oC) 0 10 20 30 40 50 0 1000 2000 3000 4000 20 40 60 80 100 Temperature of Cure (oC) Young Modulus Yeld Strength Yield Strength (MPa) Young's Modulus (MPa) Young's Modulus Yield Strength 30 The results show that the mode II fracture toughness of the adhesive cured at 23oC is about 12.47 N/mm. Due to the high toughness of the adhesive when cured at 60oC and 100oC, GIIC could not be determined. 3.4 Summary of Results Table 3 summarizes the fracture toughness of the Locktite® Hysol 3422 adhesive in modes I and II, which were used in the repaired wood specimens simulations. Since the mode I toughness of the adhesive cured at 23oC and the mode II toughness of the adhesive cured at 60oC and 100oC could not be determined, these values were extrapolated and are written in bold. For the 100oC, the mode II energy release rate was fit to be four times greater than that of mode I. The values were extrapolated so that the toughness as a function of the temperature of cure could be two straight lines. The values are graphically represented in Figure 38. Table 3: Fracture Toughness of the adhesive in modes I and II. Temperature of Cure (oC) GIC (N/mm) GIIC (N/mm) 23 1.40 12.47±0.71 60 3.08±0.10 16.00 100 4.90±0.08 19.60 0 5 10 15 20 25 020 40 60 80 100 120 GiC (N/mm) Temperature of Cure Mode I Mode II Figure 38: Toughness in modes I and II as a function of the temperature of cure. 31 4 Repair of Wood Structures 4.1 Mechanical Properties of the Pinus Pinaster Wood On the macroscopic scale, wood is an orthotropic material with three orthogonal directions of symmetry [41-43] (Figure 39):  L, the longitudinal direction;  R, the radial direction of the annual rings;  T, the tangential direction of the annual rings. Six propagation systems are distinguished for each propagation mode (I, II and III): TL, RL, LR, TR, RT, and LT [41][43] (Figure 40). The first letter refers to the normal direction to the crack plane and the second letter to the direction of crack growth. Figure 39: Orthogonal directions of symmetry of wood [42]. 32 Figure 40: Crack propagation systems of wood. Wood failure in two propagation systems in modes I and II were considered. The elastic and fracture properties of the Pinus Pinaster wood are in Table 4 and Table 5 respectively. Because wood is a natural product, these values vary greatly from specimen to specimen, depending on the environment the tree grew in, among other factors. Table 5: Cohesive properties of the Pinus Pinaster wood for two propagation systems [44]. RL Plane LR Plane Pure mode I II I II σu,i [MPa] 16 16 65 16 GiC [N/mm] 0.2 1.2 25 1.2 Table 4: Elastic properties of the Pinus Pinaster wood [41]. EL =15.13 GPa νLR =0.47 GLR =1120 MPa ER =1910 MPa νLT =0.51 GLT =1040 MPa ET =1010 MPa νRT =0.59 GRT =170 MPa 33 4.2 Geometry Scaled specimens of Pinus Pinaster beams were repaired with bonded CFRP patches and tested under bending. These specimens were designed to simulate two common types of failure of wood beams under bending loads: compression failure (Figure 41) and cross grain tension failure (Figure 42). Table 6 shows the dimensions of the specimens. Figure 41: Schematic representation of the compression damage specimen. Figure 42: Schematic representation of the cross grain tension specimen. In order to compare the results obtained with the repaired specimens, unrepaired specimens were also considered. These had the same geometry of the repaired specimens, only lacking the bonded CFRP patches. The undamaged specimens that were also used as a comparison consisted of a [mm3] wood beam. The specimens were loaded under 4PB, as demonstrated in Figure 47. r 34 Table 6: Geometry of the compression and cross grain tension specimens. Compression failure specimen a=300 mm b=20 mm h=20 mm ta=0.2 mm th=1.2 mm r=4 mm Cross grain tension failure specimen a=300 mm b=20 mm h=20 mm ta=0.2 mm th=0.6 mm αc=15o This specimen geometry is the same used by Campilho et al. [7-8]. For each kind of beam damage, two bonded lengths were considered (L0): 20 and 30 mm for the compression damage and 40 and 60 mm for the cross grain tension damage. 4.3 Specimens Manufacture Right before the bonding of the patches, in order to raise the critical surface tension of wood and improve its wettability, the wood surface was abraded with sandpaper and cleaned with compressed air. It was not cleaned with acetone, as it could be absorbed and affect the bonding. The patches were also abraded with 220 grit sandpaper and cleaned with acetone. They were bonded using the Locktite Hysol® 3422 adhesive cured in three different ways:  Isothermal cure at room temperature (23oC): this allows the adhesive to have a stiff, brittle and high strength behaviour;  Isothermal cure at high temperature (100oC): this allows the adhesive to have a ductile, flexible and low strength behaviour;  Graded cure: this allows the adhesive to be stiff where stresses are normally low and flexible where stresses are normally high. This way a more uniform stress distribution can be obtained. 35 4.3.1 Isothermal Cure In order to guarantee the correct adhesive thickness, two spacers were used (Figure 43). The pressure was applied through the use of grips. Many specimens can be made with the use of a single pair of grips (Figure 44). This allows several specimens to be made at the same time. Figure 44: Several specimens being held by one pair of grips. Figure 43: Way to guarantee the adhesive thickness. Spacer Steel plate 36 0 0 1 0 0 1 The specimens were left to cure either at room temperature (low temperature) or in the oven (100oC temperature of cure) for one hour. One week after the application, the excessive adhesive was removed. 4.3.2 Graded Cure Induction heating was used to raise the temperature of the adhesive at the ends of the overlap, where the stress concentration exists. As this adhesive is flexible when cured at high temperatures, this allows the stresses in the joint to be more uniform. This technique has already been successfully used in single lap joints by Carbas et al. [34]. A recently invented apparatus was used to locally heat the adhesive and perform a graded cure [35]. Figure 45 shows the approximated distribution of temperature along the overlap length for the compression and cross grain tension specimens. The temperature gradient was kept during one hour and was monitored using a thermographic camera. As at ends of the overlap there is a greater gradient in the stress than in the middle. In order to obtain a more uniform distribution, these areas must also receive a considerable gradient in the rigidity. To achieve this, a great gradient in the temperature of cure was used at this area. At the middle of the overlap there is not a great variation in the stress. This is why a small gradient in the temperature of cure was used here. T T x/L0 x/L0 a b 23oC 100oC 23oC 100oC Figure 45: Approximated temperature distribution during the graded cures: a. Compression specimens (half of the beam) b. Cross grain tension specimens 37 4.4 Manufacture of the CFRP Patches The wood beams were repaired using CFRP patches. This is an orthotropic material, whose elastic mechanical properties can be seen in Table 7. Table 7: Elastic properties of the CFRP patches [25]. Ex= 1.09E5 MPa νxy=0.342 Gxy=4315 MPa Ey=8819 MPa νxz=0.342 Gxz=4315 MPa Ez=8819 MPa νyz=0.380 Gyz=3200 MPa Two 300x300 mm2 CFRP plates were manufactured (one was 1.2 mm thick for the compression specimen, the other was 0.6 mm thick and was used in the repair of the cross grain tension specimen). The manufacture process had five steps: 1. Cutting the pre impregnated carbon fibres in squares (300x300 mm2). Each square becomes a layer. Each layer is 0.15 mm thick. For the 0.6 mm thick plate, four layers are needed and for the 1.2 mm thick plate, eight layers must be cut. 2. The Teflon® films involving the pre impregnated carbon fibre should be removed. Each plate should be placed on the top of the other. In this step attention must be paid to the orientation of the fibres. A hot air gun should be used to heat up the plates and activate the epoxy resin. The air bubbles that are created can be removed with the use of a heavy trowel. 3. Using duct tape, a steel strip with the thickness of the plate was attached perpendicular to the fibres. This prevents the deformation of the plate during the pressure cycle, as well as the resin loss. 4. The plate was subjected to the heat and pressure cycle illustrated in Figure 46. This was made using a hot plates press. It is very important that the cooling is slow, avoiding the formation of residual stresses. 5. After the cure of the epoxy resin, the plate was cut into the CFRP patches. This was done with a diamond tip saw. 38 4.5 Specimens Testing The specimens were tested under four point bending in an INSTRON® model 3367 universal test machine with a capacity of 30kN. This allows a constant bending moment to be created on the repaired area (at the middle of the beam), as is shown in Figure 47. The displacement rate was 2 mm/min. The specimens were placed with the repaired area at equal distances from the rollers. The compression specimens had the patch on the upper side (compression face of the beam) and the cross grain tension specimens were placed with the repaired area on the bottom side of the beam (tensile face). Figure 47: Bending moment (Mf) of a beam under four point bending. 1h, 130oC T ⁄ ⁄ 130 260 Mf 17 bar Figure 46: Thermal cycle of the CFRP plates. 39 Large rollers (60 mm in diameter) were used not to damage the wood beams. If small rollers had been used, indentation would have been created on the beams, affecting the P-δ curves. Five specimens were manufactured for each repair geometry. Only the valid tests were considered in the analysis of results. 4.6 Numerical Analysis The FEM analyses were performed in Abaqus® using CZM. 2D models were used. The CFRP patch and the wood beam were modelled with 4 node solid elements (CPS4R). The adhesive was modelled with 4 node cohesive elements (COH2D4). Cohesive layers were also added to the beam (Figure 48), so that the crack initiation and propagation could be simulated. Due to the symmetry of the compression specimen, in order to decrease the computational effort, only half of the beam was numerically simulated. The boundary conditions, as well as the place where each layer of cohesive elements was created are represented in Figure 48. Figure 48: Boundary conditions and cohesive elements location: a. Compression failure b. Cross grain tension failure In the compression specimens, to simulate failure at the symmetry axel of the beam and bellow the loading cylinder, cohesive layers of wood were added to these locations. In the Wood in the LR plane Wood in the RL plane Adhesive δ δ δ a b 46 0 1000 2000 3000 4000 5000 0246810 P (N) δ (mm) Experimental Numerical Figure 61 shows the experimental and numerical P-δ curves of the specimens. Failure occurred when P reached about 4171 N. The numerical simulation was able to accurately predict the strength of the beam. As in the numerical model the grain is perfectly aligned, the numerical failure, like the experimental failure, occurred by simple tension under a loading cylinder. 4.7.2 Compression Damage Specimens The unrepaired compression failure specimens failed in the wood mostly in the symmetry plane by pure tension (Figure 62-a). The specimens that exhibited a slight cross graining failed by cross grain tension (Figure 62-b). Figure 60: Failure mechanisms observed in the undamaged beam: aSimple tension bCross grain tension a b a Figure 61: Experimental and numerical P-δ curves of the undamaged beam. 47 0 1000 2000 3000 4000 0 2 4 6 8 P (N) δ (mm) Experimental Numerical Figure 63: show the experimental and numerical P-δ curves of the unrepaired compression specimens. The average strength of the beams was about 2852 N. The failure mechanisms reported for both the 20 mm repair and for the 30 mm repair were similar. Fracture occurred away from the repaired region, either by simple tension or, in the beams that exhibited slight cross graining, by cross grain tension (Figure 64). The failure of the 20 mm patch specimens that failed by simple tension initiated in the symmetry axel of the beam, while the 30 mm patch’s started under one of the loading cylinders. As failure occurred away from the repaired area, no conclusions could be made about the effectiveness of the different kinds of cure in the performance of the beams. Figure 62: Failure mechanisms observed in the unrepaired compression damage specimens: aSimple tension bCross grain tension a b Figure 63: Experimental and numerical P-δ curves of the unrepaired compression specimen. 48 0 1000 2000 3000 4000 5000 0246810 P (N) δ(mm) Experimental Numerical 0 1000 2000 3000 4000 5000 02468 P (N) δ(mm) Experimental Numerical Figure 64: Failure mechanisms observed in the repaired compression damage specimens: aSimple tension (symmetry plane) bSimple tension (below the loading cylinder) cCross grain tension a b Figure 65: Experimental and numerical P-δ curves of the 20 mm repair compression specimens. Adhesive cured at 23oC (left) and at 100oC (right). c 49 0 1000 2000 3000 4000 5000 02468 P (N) δ (mm) Experimental Numerical 0 1000 2000 3000 4000 5000 0 2 4 6 8 P (N) δ(mm) Experimental Numerical 0 1000 2000 3000 4000 5000 0246810 P (N) δ(mm) Experimental Numerical 0 1000 2000 3000 4000 5000 6000 0 2 4 6 8 10 P (N) δ(mm) Experimental Numerical Figure 66: Experimental and numerical P-δ curves of the 20 mm repair compression specimen. Graded adhesive. Figure 68: Experimental and numerical P-δ curves of the 30 mm repair compression damage specimens. Graded adhesive. Figure 67: Experimental and numerical P-δ curves of the 30 mm repair compression damage specimens. Adhesive cured at 23oC (left) and 100oC (right). 50 0 500 1000 1500 2000 2500 0246 P (N) δ (mm) Experimental Numerical In the 20 mm and 30 mm compression failure specimens, failure occurred in the wood. The strength of the 20 mm compression failure specimens was lower than the strength of the 30 mm compression failure specimens because fracture occurred in the symmetry axel of the beam, influenced by the damaged area. The FEM failure also shows the fracture occurring in the wood beam by simple tension in the symmetry axel of the beam for the 20 mm repair and below the loading cylinder for the 30 mm repair. The beams repaired with the adhesive cured at 100oC are slightly more compliant, as the adhesive cured at this temperature is more flexible. The numerical results regarding the 23oC and the graded cure 30 mm patch repairs did not match exactly the experimental results. As all other simulations were reasonably accurate, this is probably due to the natural variability of wood mechanical properties. 4.7.3 Cross Grain Tension Specimens The damage in the unrepaired cross grain tension specimens initiated in the RL plane (Figure 69), on path 1 (Figure 71). Despite the drop in the stiffness of the beam, P continued to rise until a crack appeared on path 2 (Figure 71), leading to a drop in P. This is visible in the P-δ curves (Figure 70). Figure 69: Failure mechanism of an unrepaired cross grain tension specimen a Figure 70: Experimental and numerical P-δ curves of the unrepaired cross grain tension specimens. 51 0 1000 2000 3000 0 1 2 3 4 P (N) δ (mm) Experimental Numerical 0 1000 2000 3000 01234 P (N) δ (mm) Experimental Numerical Numerical failure also occurred first on path 1, with the consequent loss of stiffness, and then on path 2. Failure in the repaired beams (for both the 40mm patch and the 60 mm patch specimens) occurred suddenly. Cracks appeared instantly, at the same time, in the adhesive-wood interface and paths 1 and 2 (Figure 72). Figure 72: Failure mechanism of a repaired cross grain tension specimen. Figure 73: Experimental and numerical P-δ curves of the 40 mm repair cross grain tension specimens. Adhesive cured at 23oC (left) and at 100oC (right). Path 2 Path 1 Figure 71: Schematic representation of paths 1 and 2. 52 0 1000 2000 3000 4000 0 1 2 3 4 P (N) δ (mm) Experimental Numerical 0 1000 2000 3000 4000 0 2 4 6 P (N) δ (mm) Experimental Numerical 0 1000 2000 3000 4000 0 2 4 6 P (N) δ (mm) Experimental Numerical 0 1000 2000 3000 4000 5000 0 2 4 6 8 P (N) δ (mm) Experimental Numerical Figure 74: Experimental and numerical P-δ curves of the 40 mm repair cross grain tension specimens. Graded adhesive. Figure 75: Experimental and numerical P-δ curves of the 60 mm repair cross grain tension specimens. Adhesive cured at 23oC (left) and at 100oC (right). Figure 76: Experimental and numerical P-δ curves of the 60 mm repair cross grain tension specimens. Graded cure. 53 Figures 73-75 show that the strength of the beam increases with the patch length (L0). The beams repaired with the ductile adhesive were stronger than the beams repaired with the brittle adhesive. The strongest beams were those repaired with the graded adhesive. Numerical failure started in the adhesive-wood interface. At this stage, a drop in P occurred. The crack propagated first to path 1 and then to path 2. 4.8 Discussion of the Results Table 8 summarizes the results of the 4PB tests Table 8: Average strength of the specimens. Specimen L0 Tcure Pmax [N] Compression Specimen 20 mm 23oC 3973.5±467.4 100oC 3734.0±303.1 Graded 3981.7±289.4 30 mm 23oC 4926.6±647.0 100oC 3859.7±392.8 Graded 4447.3±149.3 Unrepaired 2852.4±190.,7 Cross Grain Tension Specimen 40 mm 23oC 2504,7±320,8 100oC 2677.1±191.9 Graded 2745.1±403.3 60 mm 23oC 2872.1±329.6 100oC 3391.4±175.6 Graded 3579.8±608.7 Unrepaired 2485.3±174.2 Undamaged Beam - 4171.3±277.2 54 0 1 2 3 4 5 6 7 8 9 10 Low Temperature Cure High Temperature Cure Graded Cure 0 5 10 15 20 25 30 35 40 -41 -40 -39 -38 -37 -36 -35 -34 -33 -32 -31 Low Temperature Cure High Temperature Cure Graded Cure -35 -30 -25 -20 -15 -10 -5 0 Figure 77 and Figure 78 show the improvement in the strength of the specimen versus an unrepaired beam and an undamaged beam respectively. Figure 77 shows that the adhesive cured at high temperature (flexible behaviour) was more efficient in repairing wood beams damaged by cross grain tension than the adhesive cured at low temperature (brittle behaviour). The specimens repaired with a graded bondline showed the greatest strength and are the best choice when repairing wood structures. However, as Figure 77: Strength gain of the cross grain tension specimens versus the unrepaired beam. The 40 mm repair (left) and the 60 mm repair (right). Figure 78: Strength gain of the cross grain tension specimens versus the undamaged beam. The 40 mm repair (left) and the 60 mm repair (right). Percentage of strength gain versus the damaged beam Percentage of strength gain versus the damaged beam Percentage of strength gain versus the damaged beam Percentage of strength gain versus the undamaged beam Percentage of strength gain versus the undamaged beam 55 0 5 10 15 20 25 30 35 40 45 Low Temperature Cure High Temperature Cure Graded Cure 0 10 20 30 40 50 60 70 80 -12 -10 -8 -6 -4 -2 0 Low Temperature Cure High Temperature Cure Graded Cure -10 -5 0 5 10 15 20 failure occurred in the repaired region, these patches were not able to restore the full strength of the beam, as can be seen in Figure 78. In order to fully repair the cross grain tension wood beams, longer patches should be used. The great advantage of the graded joint is that the adhesive is ductile where there is great stress concentration and resistant where the stress concentration factor is low, however, as fracture occurred in the wood-adhesive interface, the adhesive was not allowed to develop its full ductility or strength. The improvement on the strength of the beams was due to the more uniform stress distribution, obtained with this kind of bondline. Figure 79: Strength gain of the compression specimens versus the unrepaired beam. The 20 mm repair (left) and the 30 mm repair (right). Percentage of strength gain versus the damaged beam Percentage of strength gain versus the damaged beam Percentage of strength gain versus the undamaged beam Percentage of strength gain versus the undamaged beam Figure 80: Strength gain of the compression specimens versus the undamaged beam. The 20 mm repair (left) and the 30 mm repair (right).