Fully reversible reinforcement of softwood beams with unbonded composite plates

In this paper, results of flexure tests aimed at improving the structural behavior of softwood beams reinforced with unglued composite plates and at developing an effective alternative to the use of organic resins are presented. The addition of modest ratios of GFRP (Glass Fiber Reinforced Polymer) composite strengthening can prevent tension failure in timber beams. However the application of organic matrices presents problems of reversibility, compatibility and durability with timber and poor performance at high temperatures. The increment in capacity and stiffness and the analysis of the failure modes is the central focus of this paper. The experimental campaign is dealing with a significant number of un-reinforced and reinforced beams strengthened with unbonded GFRP plates. A 3-dimensional finite element model is also presented for simulating the non-linear behavior of GFRP-reinforced softwood beams. The ability of the numerical model to reproduce experimental results for the load–deflection curves is validated.


INTRODUCTION
Softwood is from gymnosperm plants and it is the basis of approx. 85 % of the world's production of wood elements. Softwood as a traditional building material has been extensively used from antiquity to the present and, among softwoods, fir wood is characterised by low weight density and good performance in terms of tensile strength, is likely to distort during seasoning. Knots or grain deviation are the main causes of tension failures. Timber construction constitutes a significant part of the infrastructure in many countries: its extensive use is essentially due to its excellent workability, good mechanical properties and low weight density. Splits caused during seasoning and natural defects may highly affect its mechanical properties and particularly cause high decreases of capacity. This reduction of the tensile strength may be as high as 90 % [1].
Softwood beams are usually replaced or reinforced with traditional methods involving the use of standard building materials such as steel or aluminum plates, or composite materials.
Timber reinforcement is often necessary for civil infrastructures: approx.47 % of US timber bridges is structurally deficient according to the National Bridge Inventory [2].
The application of composites for strengthening of softwood beams is not new. FRPs (Fiber Reinforced Polymers) have high tensile strength and stiffness. The structural use of Glass and Carbon FRP composites (GFRP and CFRP, respectively) is becoming common not only for new timber members, but also for reinforcement of structural elements belonging to the architectural heritage. Composites are usually used where at least two of their beneficial properties, e.g. high tensile strength and ease of application, may be exploited. In these situations, the total cost of using composite materials is similar to metallic alternatives such as steel and aluminum plates or replacement.
However the wide choice of composite products and their scattered mechanical properties can lead to serious problems for the designer. The selection of the reinforcement layout and the most appropriate material should be based on an accurate examination of the timber beams to be strengthened in order to avoid ineffective interventions [20][21]. The long-term durability of some FRPs also needs to be demonstrated [22][23].
Important issues remain to be solved. For example the use of FRP composites to reinforce timber beams without organic oil-based adhesives (e.g. epoxy resins) is less established.
Recently, the use of natural fibres with non-organic matrixes or mechanical metal connectors has been investigated [24], and it aims at developing an interesting competitor to the use of organic oil-based fibres (e.g. CFRP) or resins, which present problems of limited durability, low reversibility and poor performance at high temperatures [25]. Governmental and local conservation bodies do not often authorize an extensive use of organic adhesives on listed timber structures and this highly limits the use of composite materials on historic constructions. Ethical guidelines for conservation works on historic constructions often list the minimal intervention and the use of appropriate materials and fully-reversible methods [26].
In the field of green building, there are several positive aspects in this research. For example, any disposal process requires sorting materials based on composition and nature and, because timber is doubtless the ultimate green building material, its preservation and use is also desirable [27].
The reinforcement method proposed in this research meets the above requirements. The results of flexure tests on firwood beams reinforced using GFRP plates, applied on the tension side without the use of an organic adhesive, are presented in this paper. Plates have been fixed to the beam's tension surface using metal screws or bolts [28][29]. Reinforcement can be easily removed, if needed, because no organic adhesives have been used for the application of the GFRP plates. For small softwood beams, GFRP plates were reduced to a length of 1400 mm and were symmetrically applied on the beam's tension side using different types of steel screws. These beams were reinforced with a single GFRP pre-drilled plate (Fig. 2).

Timber
For reinforcement of large beams, two overlapping pre-drilled rectangular plates of dimensions 3600 × 80 × 9.5 mm (length × width × height) were used. GFRP plates were epoxy glued together and connected to the timber surface using metal screws or bolts. The mechanical properties of these plates are the same of the ones used for reinforcement of small beams.

EXTERNAL STRENGTHENING
Strengthening was performed with the application of the GFRP plates before the bending loads were applied. The timber surface was cleaned by air jet to rid it of loose particles and dust. The adhering face of the GFRP plate was also cleaned by acetone. Pre-drilled GFRP plates were fixed to the timber surface using commercially available metal screws ( Fig. 3a) applied according to different configurations.
For the geometrical arrangement of the screws on small beams, four configurations have been used. According to the first arrangement, (Fig. 4a), 28 woodscrews (50 mm length, thread diameter 4 mm) were transversally placed at a centre-to-centre distance of 100 mm. All screws were positioned 25 mm from plate's edges. According to the second arrangement ( Fig.   4b), 28 screws of the same type were placed diagonally (45°) at a centre-to-centre distance of 100 mm. Two or four U-shaped steel brackets have been used to increase the efficiency of the connection (3 rd and 4 th arrangement, respectively, Fig. 4c,d). The GFRP plate was epoxy glued to the steel bracket (Fig. 5).
For large beams, 8 mm-diameter metal screws or 18 mm-diameter bolts were used (80 and 100 mm in length, respectively). Three arrangements were adopted as shown in Figure 5.
According to the first configuration (5 th arrangement) 8 8mm-diameter (length = 80 mm) screws were applied (Fig. 5a). Screws have been placed at a centre-to-centre distance of 200 mm. For the second configuration (6 th screw arrangement), 6 8mm-diameter (length = 100 mm) screws were used for each plate end, for a total of 12 screws (Fig. 5b). The last configuration (7 th screw arrangement) is similar to first one: 6 18mm-diameter metal bolts were used (Fig. 5c). For all arrangements, screws or bolts were positioned 50 mm from GFRP plate's edge.
With the aim of comparing the effectiveness of the reinforcement technique, a limited number of both small and large softwood beams were strengthened by bonding the GFRP plate using an epoxy-resin. By ensuring perfect adherence between GFRP plate and timber, it was possible to define an upper limit to the capacity of reinforced beams.
Finally, for a small number of beams, steel brackets (Figs. 6 and 7) were used to increase the level of connection between and timber material and the GFRP plate. For each beam, 2 steel brackets were applied near the supports where the shearing force reaches the maximum value.
GFRP plate was epoxy-glued to the internal surface of the steel brackets and by drilling holes on them it was possible to apply a larger number of metal screws. The aim was to cause a better distribution of the shear loads and a reduction of stress concentration around the holes and of slippage between the GFRP and timber (Figs. 8 and 9).

SHEAR CONNECTORS
In this investigation the connection between the softwood beams and the composite plates was done using commercially available woodscrews or bolts. The natural consequence of this connection is the shear flow between the two structural components. If there is no connection or the connectors fail, the softwood beam and the GFRP plate would bend and slip reciprocally. The presence of a shear connection prevents slippage between the two materials and provides the means to achieve the reinforcement action, thus increasing both flexural strength and stiffness of softwood beams. If perfect bonding (zero slip) occurs the two components behave as one and the effect of the reinforcement is maximum.
Because the four point bending test was used for both small and large beams, the longitudinal shear forces V AB and V CD are constant between the end-supports and the loading points: where P is the total value of the bending force applied.
Using the elastic theory of the beam, because normal strains and stresses vary linearly from the neutral axis to the extreme tension or compression fibre, the shear flow P Sd at any level of a generic cross-section is: where Q* is the first moment about the neutral axis, I id is the second moment of area of the effective cross-section of the reinforced beam and i is the centre-to-centre distance between two shear connectors. For both I id and Q* the area of the GFRP plate was taken as its transformed area using a modular ratio of 4.55.
By using and adapting the formulation given in Eurocode 4 for design of composite steel and concrete structures [32], the resistance of the shear connector is the smallest of the following: . 0 (4) where f u is steel connector's ultimate tensile strength, d is the diameter of the shear connector, g V is a partial factor for design shear resistance (usually 1.25), f kw and E w are the ultimate (parallel to grain) compression strength and Young's modulus of wood, respectively. Eq. (3) derives from the design of shear connectors in concrete and it is used to study the effect of embedment of the connector. According to this theory: where h is the height of the shear connector. The total number of shear connectors needed is: The above theory has been used to estimate the needed number of connectors and it is based on the maximum shearing force values calculated by increasing of 50 % the maximum bending forces applied on un-reinforced beams. However clear limitations are present in this theory, making its application questionable and only possible for a preliminary design. The main limitations are: 1) The above formulas are based on the elastic theory of bending. This is acceptable for the GFRP plate, but not for timber. Non-linearity can occur (especially for low-grade timber) in wood due to yielding on the compression side. This will cause a shift of the neutral axis toward the compression side, a significant increment in deflections and shear flow.
2) The effectiveness of the reinforcement mainly depends on the bonding properties between the GFRP plate and the softwood beam and not on the resistance of the shear connector. Small values of slip may compromise the reinforcement's effectiveness.

METHODOLOGY AND TEST RESULTS
Four series of flexure tests were performed on unreinforced and reinforced beams (Tab. 3). It was decided to undertake four point bending tests, according to UNI EN 408 standard [33].
For small beams, a distance between the loading heads of 576 mm and a span of 1728 mm were used. In order to minimize local wood crushing, the beams were subjected to a doublepoint loading using two 42 mm-diameter steel cylinders by means of a compression hydraulic jack.
The simply-supported large softwood beams were monotonically loaded over a span of 3900 mm in four-point bending until failure occurred. A spreader steel H-shaped beam was used to apply the load to the beams, 1040 mm apart. Lateral supports were also used to preclude the lateral buckling of the beams. A load cell inside the 500 kN hydraulic actuator recorded the applied load.
For both small and large softwood beams, three inductive transducers (LVDT) were used to measure deflection at ¼, midspan and ¾ of the span. LVDT transducers were installed near the neutral axis (approx. the centre of the beam's height). Displacement-controlled loading with a crosshead speed of 2 mm/min was used.
The bending strength f m was calculated according to: where P max is the maximum value of the load; a is the distance between a loading head and the nearest end support (576 and 1430 mm, respectively for small and large beams) W is the modulus of resistance of the section (142.9 and 1333 cm 3 ). The global modulus of elasticity was calculated with the following equation: where b is width of softwood beam cross section, l is the beam's span; P 2 -P 1 is an increment of the bending load on a straight-line portion of the load vs deflections response curve; d 2 -d 1 is the deflection increment corresponding to P 2 -P 1 .
The beams' stiffness k 1 and k 2 , measured in the elastic range and at maximum load, respectively, was measured with the following: where P E was 7 and 50 kN, for small and large softwood beams, respectively. P i was a preload, applied to remove non-linear effects from the calculation of the beam's stiffness, of 2 and 5 kN, for small and large beams. and k 1,un are the mean maximum load, the corresponding mid-span vertical deflection and stiffness measured for un-reinforced beams.
For small beams, the test setup is shown in detail in Figure 10.

Un-reinforced beams
For small beams, ten un-reinforced elements were subjected to flexure in four-point-bending. between the load points where the bending moment is maximum.
The un-reinforced small and large softwood beams exhibited almost linear elastic behavior with brittle tensile failure. A very limited deviation from the value of 1 was calculated for the coefficient λ, given by the ratio between the stiffness in an elastic range and at maximum load (λ=1 denotes no variation in stiffness during loading and a perfect linear elastic behavior). For small and large beams, the value of λ was 0.812 and 0.88, respectively.

Small softwood beams
Reinforced beams were tested with the same test procedure used for unreinforced ones. For each of the 16 small beam tests, graphs of mid-span deflection versus vertical load have been drawn. These are presented in Figure 12. Numerical results are reported in Tables 5 and 6 By comparing the reinforced beams with control beams, it can be seen that the load at which the beams failed was not much higher for the reinforced beams, suggesting that the internal forces were only little shared between the GFRP plate and the timber. The use of an epoxy resin to bond the GFRP reinforcement caused a higher increase in beam capacity (+126 %) and demonstrated the importance to achieve a perfect bonding between the two materials. The load deflection plot is shown for beams reinforced with epoxy-glued and screwed GFRP plates in Figure 13. The composite action of the epoxy-glued GFRPs can be clearly seen here in the form of elevated stiffness (slope in the linear-elastic phase) and maximum bending load (beam's capacity). It is evident that the use of an epoxy resin is more reliable that the mechanical fasteners because it causes an higher increase in both stiffness and capacity.
However the long term behavior of an epoxy bond should be investigated.
In order to prevent the displacement of the fasteners and/or of timber, steel brackets have been applied near the supports to facilitate the transmission of the internal shearing force between the timber and the GFRP reinforcement. The most part of these timber beams failed because of timber cracking on the tension side; it was also observed that GFRP plates did not crack.
For this reason a residual bending strength has been detected following timber failure. The

Large softwood beams
As expected, the application of the GFRP plate also increased the bending strength of large beams, but results highly differed depending on the arrangement, diameter and number of screws used. A summary of the test results, obtained from the four-point bending tests, is presented in Table 7 and Figure 14. Reference UNL denotes a test on an un-reinforced softwood beam while the reference REL is used for GFRP-reinforced beams.
The application of only 8 8mm-diameter screws (5 th arrangement) did not cause an appreciable increment both in capacity (-0.04 %) and stiffness (5.7 %). The displacement of both the screws and timber material caused the detachment of the GFRP reinforcement and compromised its reinforcing action (Fig. 15).
By increasing the number of screws, the effectiveness of the reinforcement also increased, but very little. The application of GFRP plates according to the 6 th screw arrangement (by using  (Tables 8 and 9).
By replacing screws with larger diameter bolts (diameter 18 mm, length 100 mm) it was possible to reduce the stresses both in the metal and in the bottom timber lamination (7 th arrangement). Beams reinforced according to this arrangement exhibited a more linear elastic behaviour caused by the reduction of the displacement phenomena in the area around the holes. Yielding of timber material in compression and the low bearing resistance at the joint bolt-timber produced a plastic behaviour for high values of the bending load. Failure was first produced by shear rupture of the metal bolts (Fig. 16) and subsequently timber cracking. All beams failed in the pure moment region (between the points of application of the bending loads). Photographs of typical fractures are shown in Figure 17.
By applying steel brackets near the lateral supports (Fig. 18) Similarly to the previous tests on small beams, for two large beams the GFRP plate was applied using an epoxy adhesive. This produced an increment of the beam capacity δ 1 and stiffness δ 3 of 42.7 and 41 %, respectively and it represents an upper bound for the reinforcement effectiveness (Fig. 21 For the reinforcement, the Shell181 element was chosen to model the GFRP plate. This element material was also assumed to be orthotropic and the stress-strain relationship was assumed to be linear up to failure. GFRP plates were connected to joint points on the beam.
The same mesh size was used to model the GFRP plate and softwood beam in order to provide full overlap at the joints. Figure 22 shows the finite element model consisting of

Analysis results
In order to find the actual stress field at maximum load, a finite element analysis was conducted, in which the timber beams were subjected to both self-weight and a uniform load pressure under different load stages.
The results of the finite element analysis are summarized in Figure Table 10.
This comparison reveals good agreement between the theoretical predictions and experimental data for the collapse mode and for the corresponding load-carrying capacity.
The unreinforced beam was predicted to fail in tension in a brittle manner, as was also determined experimentally, and the deviations between the calculated and measured values was found to be no more than 14 %. A good agreement between theoretical and experimental results was also detected for both epoxy-bonded and unbonded GFRP-reinforced beams. The error of the model was in fact 12 % in both cases. As for the failure mode (Fig. 24), the FEM model properly simulated the experimental behavior of the beams, which failed because of timber cracking on the tension side (suggesting that the internal forces were not completely shared between the GFRP plate and the timber).

CONCLUSIONS
The mechanical behavior of low-grade (softwood) solid and laminated (glulam) beams reinforced using unbonded pultruded GFRP plates has been investigated. The pultrusion process is ideally suited to the economic production of prismatic composite profiles and the GFRP application without polymeric adhesives may be of interest to avoid irreversible interventions and to guarantee a more durable mechanical connection between timber substrate and reinforcement. The composite plates were anchored at the bottom of softwood beams using metal screws or bolts.
The fastener's and timber displacement, induced by both the low quality of commercially available fasteners and the limited parallel-to-grain timber compression strength, partially compromised the effectiveness of the reinforcement. Different fasteners' configurations were investigated. In order to reduce the stress concentration near the connection, the fasteners' number and/or diameter was increased: in this way it was possible to achieve an increase of bending strength up to 58.9 % and of the flexural stiffness up to 98.9 %. A further increment in both capacity and stiffness was measured when steel brackets were applied near the beam's end-supports to increase the level of connection between GFRP plate and timber.
The typical failure mode of GFRP-reinforced beams initiated from the displacement of the metal fasteners and timber material and subsequently resulted in timber cracking on the tension side without any significant damage to the GFRP plate. Strain measurements on the GFRP pultruded plates were always well below the plate's tensile strength.
The effectiveness of the reinforcement was studied by deriving an upper bound for capacity and stiffness (by preventing slippage between GFRP plate and softwood beams). This was achieved by using an epoxy adhesive to bond the two materials.
Initial prediction from the numerical model showed good agreement with test results for capacity, mid-span deflection and stiffness. However the modelling method of the connections between the metal fasteners and softwood needs to be further investigated in order to take into account the local effects in terms of shear stresses and displacements.