Development of a structural composite using Irish-grown timber and fibre reinforcement

Transcription

1 Development of a structural composite using Irish-grown timber and fibre reinforcement Gilfillan, Raymond 1, Gilbert, Stephen 2 and Russell, Derek 3 ABSTRACT In comparison with many European countries Ireland does not make the most effective use of native timber. It is good economic and environmental practice to promote the more efficient utilization of this natural resource. This paper describes a pilot project the objective of which was to demonstrate that laminated beams fabricated with Irish timber and incorporating reinforcement are substantially stiffer and stronger than unreinforced laminated beams. The approach combined analytical and experimental methods. Bond and small scale beam tests were carried out in the laboratory; longer span beams were fabricated commercially. It has been established that the most effective form of reinforcement is a thin pultruded unidirectional fibre strip. For modest reinforcement ratios a considerable increase in both bending stiffness and ultimate load capacity has been achieved. In addition, the behaviour at ultimate load, depending on the compressive behaviour of the timber, shows much improved consistency compared with the inherent variability of the non-reinforced Sitka Spruce beams. 1. INTRODUCTION Ireland is self-sufficient in timber and the volume is expected to double in the next ten years. In comparison with other European countries which have an abundance of this renewable resource, native timber is not widely used for primary structural elements. An examination of the potential of home-grown timber for higher value applications can make it more competitive with other materials and can substitute imported timber. Many traditional prejudices associated with home-grown timber, such as accuracy of sawing, poor seasoning and presentation, have largely been overcome by increased investment in sawmilling equipment, by kiln drying and by the introduction of grading systems. However, home-grown timber, particularly Sitka Spruce which is the most widely planted species, compares unfavourably with imported softwood in terms of strength, stiffness and dimensional stability. Glue laminating is obviously an attractive technique to produce an element with higher structural performance than one of natural timber. Laminating will also minimise distortion which can be a problem associated with the kiln drying of Sitka Spruce. However, it is an expensive process, even for straight beams and further enhancement of the performance would be desirable by the addition of reinforcement. Enhancement of the behaviour of timber by the addition of reinforcement is not a new concept. Extensive work has been carried out in North America and Europe for over 30 years (Bulleit 1984). In recent years increasing attention has been paid to high performance fibre reinforcement (Van de Kuilen 1991, Davalos & Barber 1991, Tingley 1996). The objective of this programme was to demonstrate that glued laminated beams fabricated with home-grown Sitka Spruce and incorporating fibre reinforcement will exhibit considerably higher stiffness and strength properties than unreinforced beams of the same material and that this can be achieved without a significant increase in fabrication (labour) costs. The work was carried out in two stages. The first involved theoretical analysis and laboratory based experimental work, the overall objective of which was to identify a general reinforcing medium and matching adhesive. The second stage work concentrated on the fabrication and testing of reinforced beams. This progressed from small span clear natural beams, to 3m span laminated beams and eventually to 6m span glulam beams produced under commercial workshop conditions. 1 Senior Lecturer, School of Architecture, Queen s University of Belfast, N. Ireland. 2 Senior Lecturer, School of Civil Engineering, QUB. 3 Research Engineer, School of Civil Engineering, QUB.

2 2. THEORETICAL STUDIES A study of rectangular cross section beams, with and without tension reinforcement, was carried out for both elastic and ultimate behaviour states. The analyses were converted to a spreadsheet format, enabling a parametric study of the various cases to be made. The first case dealt with assumed elastic behaviour of three layers : timber in compression; timber in tension; reinforcement in tension. It was possible to predict neutral axis position, flexural stiffness, shear capacity and moment of resistance. An ultimate load approach was developed in which various assumptions were made regarding the depth and shape of the compressive stress block ranging from partial yield to complete yield over various depths. The most important outcome of this part of the overall study was to indicate the influence of reinforcement type and percentage area on the strength and stiffness behaviour of beams. 3. EXPERIMENTAL PROGRAMME 3.1 Sitka Spruce The timber, sawn section size 75mm x 35mm, was sourced from a large local sawmill selling to Irish and overseas markets. It was kiln dried and machine graded direct to C16 class. This is a typical specification for constructional timber. From this consignment small clear samples were prepared and subjected to compression, tension and shear block testing in accordance with the appropriate standard. A careful selection was made to produce clear 600mm and 1400mm span beams for flexural testing. The majority of the consignment was used to produce laminates for 6 no. 6m span glulam beams. All this material had been pin-stacked and top weighted in the laboratory to reduce the moisture content further. Individual laminates were weighed and their flexural stiffnesses measured at 600mm increments using a laboratory rig which applied a single point load over a gauge length of 1200mm. The set of E values obtained from this flexural testing was adjusted to a uniform moisture content (14%) using an established relationship (Lavers 1974). A frequency diagram was prepared to examine the variability of E. Laminates were then taken at random and assembled in beam sets to be labelled and photographed, thus providing a map of the faces of all laminates. 3.2 Reinforcement system Three reinforcement materials were considered: glass fibre, carbon fibre and steel. Four forms of the fibres were considered: separate fibres or strands, matting including unidirectional, prepreg sheet and pultruded rod or strip. The first two forms were rejected following preliminary beam tests in which the theoretical predicted stiffness was not achieved. This can be explained by crimping and lack of straightness of the fibres in the resin adhesive matrix, samples of which when removed from an experimental beam exhibited an E value similar to that of the timber. The prepreg form was thought to be impracticable and attention centred on pultruded fibre reinforced strip (FRP). Strip was the shape most compatible with the laminating process. Various types of adhesives, epoxy resins ( filled and unfilled ) and conventional wood glues, were evaluated using a shear bond test. Cube specimens (40mm side) were prepared to test first timber/timber, then timber/reinforcement/timber junctions and finally timber/multi-layer reinforcement/timber junctions (this final set anticipated the need for several layers of reinforcement for the larger beams.) 3.3 Flexural tests : clear beams Beams of nearly clear quality were prepared : 600mm span for flatwise bending and 1400mm span for edgewise bending, so maintaining a span to depth ratio of 20. The flexural stiffness (EI) of all of these beams was measured. Various types and amounts of reinforcement were then added and the beams retested to failure. 3.4 Flexural tests : laminated beams A 3m span beam was fabricated in the laboratory. The laminates were sawn from a single solid (natural) piece 150 x 75mm which had been tested elastically. The finished size of the laminated beam was also 150 x 75mm, made up by adding an extra laminate. After flexural testing of this beam a pultruded glass fibre reinforced strip (GFRP) was bonded to the soffit. This composite beam was tested elastically. The strip was then removed, the soffit lightly planed and a pultruded carbon fibre strip (CFRP) bonded on. Electrical resistance strain (ERS) gauges were bonded to both edges of each laminate at mid-span. The resulting beam was tested to failure.

3 For the 6 no. commercial beams the 6m dimension was chosen as it typifies the span at which laminated beams become viable in practice. In addition, the choice was influenced by handling and access to the testing machine in the laboratory. 10 no. laminates of sawn section 75 x 35mm were used to produce a finished beam of section 300 x 70mm, maintaining the span to depth ratio of 20. The beams were fabricated using conventional laminating practice. All six were laid-up in a double sided hydraulic press which produced a glue line pressure of 0.7N/mm 2. For these beam tests it was decided to use CFRP strip reinforcement only which, of all the materials considered, is shown to have the highest performance at ultimate load in the theoretical studies. ERS gauges were fitted as before. All beams were tested elastically in a twopoint loading configuration. Finally, the beams were tested to failure: some unreinforced, some 'under-reinforced' and some 'over-reinforced'. These last two descriptions refer to the ultimate load analysis and the assumptions made there. 3.5 Properties of in-grade Sitka Spruce The strength and stiffness properties of structural size in-grade samples from axial tests were considered important for the theoretical prediction of elastic and ultimate load behaviour of large beams based on a multi-layer section. Coupons were cut out of all laminated beams from which in-grade compression samples were prepared and tested. For the tension properties a set of laminates was selected from the remaining timber stock which exhibited E values (from flexural testing, section 3.1) close to the median value of the distribution of laminates in all beams. This set was tested in a purpose made axial loading rig. Strains were measured at 200mm intervals using a Demec gauge. Compression tests were then carried out on samples cut from the laminates at every tension gauge length. 4. SUMMARY AND DISCUSSION OF RESULTS 4.1 Theoretical studies Assuming elastic behaviour, the effect of reinforcement type and its percentage area on moment of resistance and flexural stiffness is illustrated by Figure 1. This example of the parametric study was based on a beam section size 70 x 300mm of C16 timber, representing the expected size of the commercial beams. The analysis was developed as a two-layer timber system with E in tension set at a higher value than E in compression. The moments of resistance are based on the permissible bending stress for timber, i.e., tension equal to compression. For all types of reinforcement the moments of resistance are limited by the timber in compression. In the ultimate load analysis of reinforced beams the assumptions made with regard to the shape of the stress profile approaching failure and the depth of the compression zone, were critical. A variety of stress profiles was examined, as illustrated in Figure 2 (ii) (v). Profile (v) is a case of (ii), (iii) and (iv) in which full compression yield of the timber extends to the total depth of the section. Figure 3 shows, for profile (iii), the variation of ultimate moment of resistance with reinforcement content. This ultimate moment of resistance is factored by setting the ultimate timber stresses at values which, for an unreinforced beam, produce the same moment of resistance as that assessed by conventional elastic theory, as shown in Figure 1. Thus the moments in Figures 1 and 3 are comparable and it is interesting to note that the advantageous position of steel by elastic theory is now replaced by carbon in Figure 3. Two assumptions for the depth of the compression block are illustrated in Figure 3: full yield to total section depth and full yield to 2/3 of section depth. For the latter assumption the factored moment of resistance is reduced to 70% of the total depth value. The concept of over, balanced and under reinforcement is clear in Figure 3. The precise area of reinforcement depends on the assumption made for the depth of the compression timber. Clearly the assumption of compression yield to total depth gives a conservatively high balanced reinforcement area. It is obvious that strength assessment by ultimate load assumptions is very different to that based on elastic behaviour. 4.2 Behaviour of the reinforcement system With all reinforcing materials the various resins gave good results when the strip surfaces were sanded as well as solvent cleaned. Some resins were unfilled, the rest filled. With these latter resins, wetting of the surfaces was more difficult, but assembly was easier due to the higher friction within the resin film. All the shear failures took place in the timber. Glass and carbon fibre strip were tested with RPF wood glue. Failure occurred at the interface at values between 50 and 75% of the timber shear values. A filled resin was chosen for all reinforcement bonding. 4.3 Behaviour of clear beams Sets of beams were tested; unreinforced, over reinforced (60mm 2 and 24mm 2 of CFRP for 600mm and 1400mm spans respectively) and under-reinforced (12mm 2 of CFRP for each span). The balanced amount of reinforcement, derived

4 using stress profile (v) of Figure 2, was 21mm 2 CFRP, for both section orientations. The elastic flexural behaviour showed good agreement with the predictions based on profile (i) of Figure 2. During ultimate load testing, all the over-reinforced (CFRP) beams exhibited a failure mode governed by compression in the top of the beams. Definite bands of fibre crushing were clearly visible in the regions of maximum moment. The beams failed in a ductile manner without any rupture or debonding of the carbon strip. Even with careful sorting of timber the longer span beams still contained knots associated with slope of grain. When knots were present in the top margin the compression banding was associated with these knots. Generally for the under reinforced beams, failure was initiated by the tension timber and for, the larger span beams, failures always occurred due to knots and/or grain slope in the tension zone close to the reinforcement. 4.4 Behaviour of 3m and 6m span laminated beams Table 1 presents the flexural stiffnesses and ultimate loads of these beams and compares experimental results with theoretical predictions. For the 3m span reinforced beams, 3(c) had 225 mm² (2%) GFRP and 3(d) 72 mm² (0.64%) CFRP. The CFRP area produced an under-reinforced section: ultimate load analysis indicates 110 mm² (0.98%) for a balanced section using profile (v) of Figure 2. On the 6m span beams, 6/2 with 84 mm² (0.4%) CFRP was underreinforced and 6/3 with 252 mm² (1.2%) was over-reinforced. For this section the balanced area of CFRP is 206 mm² (0.98%). All the measured stiffnesses of the reinforced beams agreed closely with predictions and showed increases over the basic beams ranging from 27% for beam 3(c) to 55% for 6/3. The collapse mode of the beams differed. Beam 6/1 (unreinforced) failed suddenly due to tension in the timber. For beam 6/3 (over-reinforced) there was clear evidence of fibre crushing in the upper laminates, followed by secondary failure, as deflection increased, which caused the timber to split longitudinally due to tension across the grain and the CFRP strip to delaminate internally. For the 'under-reinforced' beams, 3(d) and 6/2, there was less evidence of compression yielding: failure was due to the timber splitting longitudinally and tensile failure of the reinforcement. Table 2 compares the experimentally observed and predicted neutral axis positions at both elastic and ultimate load states. The neutral axis positions by elastic analysis (profile (i) of Figure 2) gave good agreement with those obtained from the strain profiles indicated by the ERS gauges. Figure 4 shows an example of the strain profile (for beam 6/3) developing from the elastic to the ultimate load states. For the prediction of moment Mu and position of neutral axis at failure, the analysis was based on either profile (ii) or profile (iii) (Figure 2). It should be noted that profile (v) is a case of (ii), (iii) and (iv) when compression yield extends to the total depth of the section and no tension timber is present. The timber yield stress values used were those obtained from axial in-grade testing (section 3.5), 30N/mm 2 (f c ) and 35 N/mm 2 (f t ), compression and tension respectively. These values correspond closely to the flexural values derived from the experimental ultimate moment of the beam 6/1 (unreinforced). The strain profile at failure of this beam showed that the neutral axis shifted down some 7 mm, indicating that f t was slightly greater than f c. In theory this is the result of a small zone of compression yield at the top edge. Nevertheless the 'modulus of rupture' approach, assuming a straight line stress distribution at failure, gives a reasonable estimate of f c and f t. Since beam 6/1 was not reinforced and compression yield was limited, profile (ii) was assumed for the ultimate moment and neutral axis prediction. For the two under-reinforced beams, 3(d) and 6/2, analyses based on profiles (ii) and (iii) gave similar results for Mu. The influence on Mu of changing the value of f t was examined. When f t was varied between 15 and 35 N/mm² the upper and lower limits for Mu differed only by 17% (based on the lower value) over both approaches. Profile (iii) gave a much better prediction for the neutral axis position. For the over-reinforced beam, 6/3, the stress profile deduced from testing showed that compression yield in the timber extended only to 2/3 depth of section. So, the balanced area of reinforcement, using profile (v), is an over-estimate. With this higher reinforcement content the analysis based on profile (iii) gave the better prediction of Mu and neutral axis position (for the trial variation f t, as before, limits for Mu differed only by 4%). By this method the predicted Mu at 2/3 yield depth was 79 knm, which can be compared with 94 knm by the complete yield assumption: a reduction of 16%.

5 5. CONCLUSIONS 1. Performance of epoxy adhesives in selection tests and in the reinforced beams has been good. The shearing stresses at the bond line due to ultimate beam loads are considerably lower than those measured in the test cubes. Shear cube results with conventional RPF wood glue and carbon strip reinforcement have been much less successful and no attempt has been made to use this combination in a beam. 2. The elastic behaviour of reinforced beams is predictable by conventional layer theory. Good agreement was found for flexural stiffness and position of the elastic neutral axis. 3. The concept of over, balanced and under reinforcement is clear in the ultimate load analyses proposed. The balanced amount of reinforcement depends on the assumption made for the area of timber available in compression. It is obvious that strength assessment by ultimate load assumptions is very different to that based on elastic behaviour. The experimental investigation of reinforced beams has confirmed the ultimate load theory used. The ductile compressive timber failure in over-reinforced beams was a familiar feature during testing. The experimental results obtained from the 6m span multi-layered beams has informed the development of ultimate load predictions. The evidence from the strain and the stress profiles supports the theoretical assumptions which have been made. 4. It appears that for over-reinforced beams compression yeilding in the timber does not extend below aproximately 2/3 of the beam depth. The original assumption of total yield produces an overestimate of the reinforcement area needed to mobilise all the timber in compression 5. Analyses assuming either partial or full compression yield profiles gave similar predictions for the ultimate moment. The full yield assumption tended to give somewhat better agreement with tests, particularly for the more heavily reinforced beams. In the case of all reinforced beams it was found that the inclusion of tension timber and the value assigned to the ultimate tension stress had only a relatively small influence on the ultimate moment predicted. 6. At the present stage of this investigation, the theoretical and experimental studies indicate the performance of home-grown Sitka Spruce beams can be significantly enhanced by the addition of fibre reinforcements. 6. ACKNOWLEDGEMENTS This research project was funded by the DETR Partners in Technology Scheme. Materials were provided by the NI Forest Service (in associaiton with Balcas Timber Ltd.) and Sika (UK) Ltd. 7. REFERENCES 1. Bulleit, W.M. 1984, Reinforcement of wood materials: a review. Wood and Fiber Science, 16(3), pp Davalos, J.F. and Barbero, E.J., 1991, Modelling of glass-fibre reinforced glulam beams, Proceedings of the International Timber Engineering Conference, pp Lavers, G.M., 1974, Chapter 1, The strength properties of timber, Medical and Technical Publishing Company. 4. Tingley, D.A. 1996, High-strength fiber-reinforced plastic reinforcement of wood and wood composite, International SAMPE Symposium and Exhibition, 41(1), p Van de Kuilen, J.W.G., Theoretical and experiemntal research on glass fibre reinforced laminated timber beams. Proceedings of the International Timber Engineering Conference, pp

6 Table 1: Test results and predicted behaviour for laminated beams Flexural stiffness Ultimate moment Beam Beam EI (x10 9 Nmm 2 ) Mu (kn.m) span / no. type Basic Reinforced Expt. Theor. Expt. Expt. Theor. 3(a) Solid na. na. na. na. unreinforced 3(b) Laminated na. na. na. na. unreinforced 3(c) 3(d) Laminated reinforced na. na. (225mm 2 GFRP) Laminated reinforced (72mm 2 CFRP) 6/1 Laminated na. na unreinforced 6/2 Laminated reinforced (84mm 2 CFRP) 6/3 Laminated reinforced (94)* (252mm 2 CFRP) Beam size: set 3(a) - 3(d) - 75x150x3000mm set 6/1-6/3-70x300x6000mm *this value was derived from the total yield assumption Table 2: Position of neutral axis Elastic Ultimate Beam a (mm) b (mm) Expt. Theor. Expt. Theor. 3(d) / / / b a

7 Mr (knm) Area of reinforcement (%) Figure 1 mild steel CFRP GFRP1 GFRP2 EI (Nmm 2 ) Area of reinforcement (%) Parametric plots, elastic behaviour <f c (i) (ii) (iii) (iv) (v) (vi) f t (i) Elastic (iv) Ultimate load - full compression yield, tension timber neglected (ii) Ultimate load - partial compression yield (v) Ultimate load - full compression yield of complete timber section (iii) Ultimate load - full compression yield (vi) Ultimate load - unreinforced section, ultimate tension stress less than ultimate compression stress Figure 2 Stress profiles used for the theoretical studies Mr (knm) Area of reinforcement (%) Figure 3 Full comp. yield of timber to depth d Full comp. yield of timber to depth 2/3d mild steel CFRP GFRP 1 GFRP 2 Parametric plots, ultimate behaviour

8 kn Dimension from bottom of beam (mm) Total neutral axis shift Compression Strain ( µe) Tension Figure 4 Strain profiles at mid-span for beam 6/3