616_santoscl2009

8/10/2019 616_SantosCL2009

1/12

8/10/2019 616_SantosCL2009

2/12

models only give reasonable predictions for very specicextreme situations such as very thin or very thick timbermembers. Typically, dowel-type connections are threedimensional problems (non-uniform stress distributionsacross the thickness of members) that must be accountedfor a convenient modelling. Few 3D FE models can be

found in literature to predict the mechanical behaviour of single-fastener joints [11,12]. However, no comprehensiveFE models for multi-fastener joints are available in litera-ture. Many modelling issues are controversial in the FEmodels, such as the choice of the appropriate constitutivemodels for wood and the adequate failure criteria. Also,deterministic approaches are the common procedure,which represents an important limitation since the problemis governed by several important parameters with stochas-tic nature (e.g. wood properties, dimensions).

This paper presents results from monotonic quasi-statictensile tests of a double-shear single dowel wood connec-tion made of pine wood, namely the Pinus pinaster species,which is one of the species with large implantation in Por-tugal. Despite the abundance of this raw material, its usefor structural applications has been disregarded due to sev-eral reasons, such as the cultural and lack of data about thebehaviour of this material. The connection wood membersare loaded in the parallel-to-grain direction according tothe recommendations of the EN26891 standard [13]. Thewhole loadslip behaviour of the joint is illustrated untilfailure. In particular, the initial joint slip modulus, the ulti-mate strength and the ductility are evaluated and comparedwith corresponding values given by the Eurocode 5 (EC5).Also, embedding tests are carried out in parallel-to-grain

direction, both in compression and tension, according tothe EN383 standard. The resulting data is applied to pre-dict the behaviour of the wood connection.

Finally, a three dimensional FE model of the wood con-nection is built using the commercial FE analysis code,ANSYS [14]. The dowel is modelled as an isotropic elastic

material; the wood is considered as an orthotropic elasticmaterial. The interaction between the dowel and the woodis simulated using contact elements, namely surface-to-sur-face contact elements available in the ANSYS. The appli-cation of contact elements turns nonlinear the analysis,requiring an incremental load stepping. Since the materialsare considered as elastic, then it is only expected to modelthe initial joint slip modulus. Therefore, the initial joint slipmodulus is evaluated and compared with the experimentaldata, allowing the calibration of the proposed joint model.Additionally, stress elds are illustrated, especially aroundthe dowel.

2. Experimental results

2.1. Experimental details

This paper presents results of an experimental programwhich includes monotonic quasi-static tensile tests of adouble-shear single dowel wood connection, performedaccording to the EN26891 standard. Also, embedding testswere carried out using single wood members under bothtensile and compressive loads as suggested by the EN383standard. The load direction was parallel-to-grain for allexperiments. Fig. 1 depicts the tests congurations, includ-

Fig. 1. Test congurations: (a) parallel-to-grain compression; (b) parallel-to-grain tension; (c) double-shear dowel connection.

172 C.L. Santos et al. / Construction and Building Materials 23 (2009) 171182

8/10/2019 616_SantosCL2009

3/12

ing the geometries of the specimens and test rigs. A nomi-nal diameter of the dowel ( d ) equal to 14 mm was selected;the dimensions of the specimens were dened proportion-ally, according to the standards.

The specimens were manufactured from Portuguese pine(Pinus pinaster Ait.) trees harvested in the region of Viseu

(Portugal). Trees with straight stems (absence of reactionwood) and diameters at breast height of about 400 mmwere selected. Three meters long logs were cut from thesample trees, between three and six meters above the basalplane. The logs were live-sawn into thick boards whichwere kiln-dried to moisture content between 10% and12%. The specimens were cut from these boards aligningthe parallel-to-grain direction with the length of the speci-mens and the wood tangential direction with the thicknessof the specimens, as depicted on Fig. 2. Wood with knots,resin pockets or other kind of imperfections was excludedfrom specimens as an attempt to reduce the usual scatterobserved in timber testing and to stay closely to the clearwood idealization used in the FE model of a joint presentedlatter on this paper.

Table 1 summarizes the experimental program. Threeseries were tested: longitudinal compression (LC) ( Fig. 1a),longitudinal tension (LT) ( Fig. 1b) and a dowel connection(CON) ( Fig. 1c).

No attempt was made to match the samples for the LC,LT and CON series, except that the specimens in each ser-ies were made from the same board or the same log. Thedensity of the specimens was obtained at the current mois-ture content (1012%), measuring their dimensions (vol-ume) and weight, resulting averaged global density

values. Tests were performed on an INSTRON machine,model 1125, rated to 100 kN. Tests were carried out under

crosshead displacement control and were instrumentedwith LVDTs, model AML/EU 10-S10 (gauge range of 10 mm), from Applied Measurements . The experimen-tal data was acquired by a SPIDER 830 system.

2.2. Results and analysis

Fig. 3a and b exhibit the load-displacement recordsfrom the embedding tests, carried out according to theEN383 standard. The experimental loadslip curves fromthe connection tests are plotted in Fig. 3c. Both displace-ment and slip resulted from LVDTs average measure-ments. As illustrated on Fig. 1, displacements of LCseries were measured between the dowel and the base plateof the testing machine (single LVDT measure); the dis-placements of the LT series were measured between thedowel and a reference plane at 184 mm from the dowel(average of two LVDTs measures); the slip of the CONseries was measured between two parallel planes distant

of 149 mm, centred around the dowel (average of twoLVDTs measures). Specimens were loaded accordingly

Fig. 2. Procedure for selection of specimens from a tree trunk.

Table 1Experimental program

Series Number of specimens

Displacement rate(mm/min)

Density (kg/m 3)

Average Standarddeviation

LC 24 0.3 570.1 38.3

LT 26 0.5 573.2 49.8CON 25 0.3 610.7a 46.3a

617.8b 20.9b

a Central member.b Complete connection.

C.L. Santos et al. / Construction and Building Materials 23 (2009) 171182 173

8/10/2019 616_SantosCL2009

4/12

the strategy suggested by the EN26891 and EN383 stan-dards. Firstly, they were loaded until 40% of the maximumestimated load ( F est ), and the crosshead position held dur-ing 30 s. After this stage, specimens are unloaded until0.1F est and the crosshead position again maintained alongmore 30 s. Finally, specimens are reloaded until failure.Two stiffness (slip modulus) values were dened: an initialstiffness, k i, determined from linear regression analysis of the load-displacement/slip responses after the initial bed-ding-in displacement/slip and until 0.4 F est ; a stiffness, k ,determined from a linear regression analysis of the load-displacement/slip response during the unloading/reloadingstages. For all tests an ultimate load can be clearly dened.

Fig. 4a and b exhibit the initial stiffness, k i, and stiffness,k , against the wood density. As can be observed, a highscatter was registered. Moreover, there is no signicant

correlation between stiffness and density, using either a

power relation, suggested by the EC5, or a linear relation-ship (Fig. 4). It is worthwhile to refer that a short range of densities was registered in this work, which could hide asignicant correlation. Additionally, the measured wooddensity was a global average value, and the stiffness (beforeyielding) is more sensitive to the local density and morphol-

ogy of wood beneath the dowel. Fig. 4a also illustrates therelation proposed by EC5, which approaches the connec-tion experimental stiffness, being a lower limit. Table 2summarizes the average stiffness values for each test. Theinitial stiffness is appreciably lower than the stiffness mea-sured for the unloading/reloading stages, being the latterusually called elastic stiffness [5,13]. Clearly the longitudi-nal compression tests led to higher stiffness than the longi-tudinal tension and joint tests, which conrms thelongitudinal tension tests as the best choice for assessingthe joint behaviour. The differences between the stiffnessvalues from the longitudinal tension and the joint testscan be, in part, justied by the different member lengths(Fig. 1). The stiffness from the longitudinal tension testscan be used to estimate the joint stiffness using a proceduredepicted in Fig. 5. The stiffness of each joint members(k (L1), k (L2)) are estimated from the stiffness of the longi-tudinal tension tests ( k LT ) corrected in order to take intoaccount the differences between lengths. It is assumed thatmember stiffness is composed by two parts: a local bearingstiffness and a remote member stiffness, being the latteronly function of the cross sectional area ( A), longitudinalYoung modulus ( E ) and remote length ( L X ). Based ondimensions from Fig. 1 and assuming E = 15.1 GPa [15 17], one derive the following joint stiffness values, k i,CON =

18301.6 N/mm and k CON = 31659.9 N/mm, respectivelyfor the initial stiffness and stiffness. These values are about20% lower than the experimental ones, and as such arerough estimates of the joint stiffness. This difference canbe partially justied by the differences in average densities(Table 1 ) between the LT and CON series. In fact the LTseries presents an average density about 7% lower thanCON series. Also, the proposed model assumes a uniformstress distribution for each member composing the speci-men of the CON series, which is not totally veried forthe side members of the connection as will be demonstratelatter with the proposed FE model. It is interesting to notethat the estimated initial joint stiffness value closely agreeswith the average prediction carried out using the EC5guidelines (k EC5 = 18533 N/mm).

Table 2 also indicates the displacement/slip at failure forthe three test series, which is a measure of the ductility, i.e.,the capacity of a connection to withstand deformationbefore nal failure. A signicant scatter is observed as indi-cates the high standard deviations. The majority of the testsculminated with a fragile failure of shear-splitting type,however, before this failure occur, the test members experi-enced important deformations. This fragile failureappeared associated to an abrupt load reduction allowingan objective denition of the displacement/slip at failure.

For sake of clarity, the load/displacement slip curves

Fig. 3. Experimental load-displacement/slip records: (a) parallel-to-graincompression; (b) parallel-to-grain tension; (c) double-shear dowelconnection.


8/10/2019 616_SantosCL2009

5/12

depicted on Fig. 3 were truncated just before the abruptload reduction. Only two specimens, one from the LT

and another from the CON series, did not exhibit a fragile

failure, at least until the tests were interrupted to protectthe LVDT transducers. For these latter cases, the displace-

ment/slip at failure was assumed to be the values at the

Fig. 4. Experimental results: (a) initial stiffness; (b) stiffness; (c) embedding strength.

Table 2Experimental results

Series k i (N/mm) k (N/mm) f h (MPa) Displacement/slip at failure (mm)

Average Standard deviation Average Standard deviation Average Standard deviation Average Standard deviation

LC 41510.8 8864.6 53508.5 8919.1 46.4 4.2 6.09 2.05LT 25586.0 7215.8 42168.7 4101.6 42.8 5.3 6.62 3.16CON 22196.0 5224.2 39954.7 3663.7 7.32 1.96


8/10/2019 616_SantosCL2009

6/12

moment of test stopping (14.4 mm for LT and 13.4 mm forCON series). Fig. 6 shows the appearance of observed typ-ical failures. Although these reported failure modes arefrom the joint centre member, similar failure modes wereobserved for the LC and LT series.

Fig. 4c shows the embedding strength data, f h , resultedfrom the tensile and compressive longitudinal tests. Theembedding strength was evaluated dividing the maximum

load (F max ) of each test by the dowel diameter ( d ) timesthe thickness ( t) of the specimen, resulting a kind of aver-age compressive stress in the specimen. Linear regressionanalysis of the experimental data was carried out, beingrepresented in the graph the mean lines and respective95% condence bands. A clear correlation between theembedding strength and density can be observed. It isworth to note that based on the mean lines comparison,the embedding strength from longitudinal compressivetests is greater than the corresponding value from the ten-sile tests. Taking into account the condence bands, thiscomparison is only valid for the density range of 540 660 kg/m 3. A comparison between experimental correla-tions and the EC5 proposal are also carried out inFig. 4c. The EC5 does not distinguish the embeddingstrength from tension or compression tests; it underesti-mates both embedding strength values being conservativeas would be expected. However, if this comparison is madetaking into account the 95% condence bands, the EC5values are only signicantly different from the LT valuesfor densities above 540 kg/m 3. Both experimental meanregression lines and EC5 relation show the same trends.Based on the embedding strength relations proposed inFig. 4, the ultimate failure load of the tested joint is evalu-ated. Both estimations and experimental ultimate failure

loads are compared in Fig. 7. Additionally, this gure

Fig. 5. Procedure for joint stiffness estimation.

Fig. 6. Typical failure modes of the centre member of the joint: (a) fragilefailure two shear cracks; (b) fragile failure single shear crack; (c) ductile

failure.


8/10/2019 616_SantosCL2009

7/12

includes a linear regression analysis of the experimental fail-ure data with respective 95% condence bands. The exper-imental data reveals important scatter around the averagetrend. Take into account the average trend the EC5 still isa safe option. However, if the 95% condence bands areconsidered, the EC5 is no longer safe for densities above640 kg/m 3 . Clearly the longitudinal compression data arenot well suited for the joint failure load estimation; the lon-gitudinal tension data seems to be a more appropriatechoice, even though the different slopes observed. Table 2includes the average strength values of the three test series.

3. Numerical results and analysis

A 3D FE model of the tested joint was built and resultsare discussed in this paper. The commercial FE code,ANSYS 10.0 [14] was used for this purpose. The modelwas built using the ANSYS parametric design languagecapabilities APDL language. Both wood members andsteel dowel were accounted in the model. They were mod-elled using hexahedra isoparametric 20-node elements(SOLID95). The contact between the dowel and woodmembers was modelled applying the contact elementsavailable in ANSYS, using a surface-to-surface option. Inparticular, the elements CONTA174 and TARGE170 wereused to model, respectively the contact and target surfaces,forming the so-called contact pair. Both surfaces in contactwere assumed exible.

Once the joint geometry admits two planes of symmetry,only 1/4 of the joint was modelled (1/2 of a side memberand 1/4 of the central member). The displacements of thenodes located at the planes of symmetry were restrainedin the normal direction to these planes. Three contact pairswere considered, namely one between the dowel and thesurface of the hole in side member, another between thedowel and the surface of the hole in central member andnally a contact pair between the surfaces of the side andcentral wood members. The lengths of the members wereassumed equal to the LVDT gauge length as indicated in

Fig. 1. Fig. 8 shows the FE mesh built for the joint. The

nodes in the base of the side member were restrained inthe vertical direction ( y) and nodes at the top of the centremember were displaced by a maximum value,d max = 0.1 mm, in vertical direction.

Both materials were modelled as homogeneous and elas-tic. While steel was considered isotropic ( E = 210 GPa,m = 0.27), wood was assumed orthotropic. The propertiesof the wood were identied by Xavier [15], Pereira [16], Oli-veira [17] and Garrido [18]. These properties were derivedfor clear wood extracted from trees of same region as thoseused to extract the wood required for this work. The elas-ticity modulus ( E L , E R , E T ) and Poissons ratios ( mLR , mRT ,mTL ) were measured from Pereira [16] and are summarizedin Table 3 . This table also includes the average densities of

wood and the standard deviations. The shear moduli weremeasured by Xavier [15], Oliveira [17] and Garrido [18]using independent approaches, respectively the Iosipescu,Arcan and Off-axis tests. The corresponding values are pre-sented in Table 4 together with the average densities andstandard deviations. The shear moduli used in the FEmodel were an average of the three estimates given in Table4: G LR = 1.28, G LT = 1.12 and G RT = 0.23 GPa.

Some preliminary tests, namely tension and compres-sion tests of small prismatic clear wood specimens, carriedout with wood from the trees used in this work conrmedthe elasticity modulus in the longitudinal direction as wellas the Poissons ratio, mLR , included in Table 3 . The densityof these preliminary tests was 646.4 22.4 kg/m 3 for thecompression tests and 634.8 42.1 kg/m 3 for the tensiletests.

Since linear elastic behaviours for materials wereassumed, the model is not able to simulate the nonlinearbehaviour of the joint until failure. The model proposedin this paper is only intended for simulation of the jointelastic stiffness, which can be compared with the availableexperimental results allowing the calibration of the model.In spite of assuming elastic materials, the global problem isnonlinear due to the contact. Therefore, an incrementalloading is applied: a maximum displacement of 0.1 mm

was applied in 10 equal increments.

Fig. 7. Ultimate load of the tested joint: experimental data and predictions.


8/10/2019 616_SantosCL2009

8/12

The contact problem is governed by an important num-ber of parameters, such as the contact algorithm, frictionmodel, normal contact stiffness, tangential contact stiffness,maximum penetration, etc. Also, each contact pair requiresthe denition of the contact and target surfaces that will bemeshed with CONTA174 and TARGE170 elements,

respectively. Furthermore to the contact model parameters,

geometric parameters can also play an important role oncontact behaviour, such as the dowel-hole gap.

Simulations were carried out using the augmentedLagrange algorithm available in the ANSYS [14]. Sincethe contact pair behaviour is governed by several parame-ters, a set of 24 simulations were carried out in order to bet-ter understand the inuence of some of those parameters

on joint behaviour, especially on stiffness. For the contactpair between the two wood members, default ANSYS(version 10.0) parameters were used since it is expected thatthis interface behaviour will not govern the global jointbehaviour. For the other two contact pairs other parame-ters were used, but using the same values for both. All sim-ulations were carried out assuming the Coulomb frictionmodel, for two distinct friction coefficients: l = 0.0 (12 sim-ulations) and l = 0.5 (12 simulations). The augmentedLagrangian method requires the denition of normal andtangential contact stiffnesses. The amount of penetrationbetween contact and target surfaces depends on the normalstiffness. Higher stiffness values decrease the amount of penetration, but can lead to ill-conditioning of the globalstiffness matrix and to convergence difficulties. Lower stiff-ness values can lead to a certain amount of penetration andproduce an inaccurate solution. Ideally, it is desirable ahigh enough stiffness that the penetration is acceptablysmall, but a low enough stiffness that the problem will bewell-behaved in terms of convergence. On effect, a stiffnessrelationship between two bodies must be established forcontact to occur. Without contact stiffness, bodies will passthrough one another. The relationship is generated throughan elastic spring that is put between the two bodies, wherethe contact force is equal to the product of the contact stiff-

ness (j ) and the penetration ( d). The amount of penetration

Table 3Elastic properties of wood [16]

E L (GPa) 15.1 1.1Density (kg/m 3) 616 12.3

E R (GPa) 1.9 0.2Density (kg/m 3) 696 6.5

E T (GPa) 1.0 0.1

Density (kg/m3

) 647 6.7mLR 0.47 0.03Density (kg/m 3) 616 12.3

mRT 0.59 0.04Density (kg/m 3) 696 6.5

mTL 0.05 0.01Density (kg/m 3) 647 6.7

Table 4Shear modulus [15,17,18]

Arcan Iosipescu Off-axis

G LR (GPa) 1.33 0.09 1.41 0.15 1.11 0.08

Density (kg/m 3) 619 38.4 589 33.0 596 23.8G LT (GPa) 1.09 0.06 1.22 0.10 1.04 0.08Density (kg/m 3) 608 15.2 589 15.3 538 21.5

G RT (GPa) 0.24 0.05 0.29 0.05 0.16 0.01Density (kg/m 3) 650 16.3 578 37.6

Fig. 8. Finite element mesh of the double-shear single dowel joint.


8/10/2019 616_SantosCL2009

9/12

Table 5Model congurations and numerical results

Simulations Results

Congurations GAP (mm) FKN FTOLN l Number of Iterations k (N/mm) Stress, r y (MPa) Maximum of

Maximum Minimum Penetration (m

01 0.1 0.1 0.10 0.0 20 31407.0 11.3 18.1 0.014921 02 0.1 1.0 0.10 0.0 20 46117.0 14.8 21.4 0.002667 03 0.1 0.1 0.05 0.0 20 31407.0 11.3 18.1 0.014921 04 0.1 1.0 0.05 0.0 20 46117.0 14.8 21.4 0.002667 05 0.1 0.1 0.01 0.0 29 37475.0 12.9 18.6 0.007306 06 0.1 1.0 0.01 0.0 20 46117.0 14.8 21.4 0.002667 07 0.3 0.1 0.10 0.0 20 25273.0 9.1 20.0 0.017231 08 0.3 1.0 0.10 0.0 22 39823.0 12.8 25.7 0.003265 09 0.3 0.1 0.05 0.0 20 25273.0 9.1 20.0 0.017231 10 0.3 1.0 0.05 0.0 23 39823.0 12.8 25.7 0.003265 11 0.3 0.1 0.01 0.0 30 33964.0 11.2 22.4 0.007612 12 0.3 1.0 0.01 0.0 23 39823.0 12.8 25.7 0.003265 13 0.1 0.1 0.10 0.5 27 31420.0 11.2 17.7 0.015192 14 0.1 1.0 0.10 0.5 22 46499.0 14.9 21.1 0.002643 15 0.1 0.1 0.05 0.5 27 31420.0 11.2 17.7 0.015192 16 0.1 1.0 0.05 0.5 22 46499.0 14.9 21.1 0.002643 17 0.1 0.1 0.01 0.5 36 37610.0 12.8 17.9 0.007555 18 0.1 1.0 0.01 0.5 22 46499.0 14.9 21.1 0.002643 19 0.3 0.1 0.10 0.5 27 25254.0 9.0 19.6 0.017435 20 0.3 1.0 0.10 0.5 29 39901.0 12.7 26.3 0.003345 21 0.3 0.1 0.05 0.5 27 25254.0 9.0 19.6 0.017435 22 0.3 1.0 0.05 0.5 29 39901.0 12.7 26.3 0.003345 23 0.3 0.1 0.01 0.5 35 33902.0 11.1 22.7 0.007738 24 0.3 1.0 0.01 0.5 29 39901.0 12.7 26.3 0.003345

8/10/2019 616_SantosCL2009

10/12

(d), or incompatibility, between the two bodies is thereforedependent of the stiffness ( j ). Ideally, there should be nopenetration, but this implies that j = 1 , which will leadto numerical instabilities. The value of j that is computedby ANSYS depends on the relative stiffness of the bodies incontact (bulk modulus, K , of contacted element) existing

the possibility of scaling j through the FKN factor alsocalled normal penalty stiffness factor. The usual factorrange is from 0.01 to 1.0, with a default of 1.0. The defaultvalue is appropriate for bulk deformation. Present simula-tions covered FKN values equal to 0.1 and 1.0. Anotherrelevant contact parameter to be used in conjunction withthe augmented Lagrangian method is FTOLN. FTOLNis a tolerance factor to be applied in the direction of thesurface normal. The range for this factor is less than 1.0(usually less than 0.2), with a default of 0.1, and is basedon the depth of the underlying solid element. This factoris used to determine if penetration compatibility is satised.Contact compatibility is satised if penetration is within anallowable tolerance (FTOLN times the depth of underlyingelements). The depth is dened by the average depth of each individual contact element in the pair. If ANSYSdetects any penetration larger than this tolerance, the glo-bal solution is still considered unconverged, even thoughthe residual forces and displacement increments have metconvergence criteria. FTOLN values equal to 0.01, 0.05and 0.1 were simulated. For all other contact parametersnot mentioned here default values were adopted [14].

The contact performance is strongly inuenced by thegap between dowel and hole. This geometric parameter isnot easy to estimate due to the difficulty on dimensional con-

trol of manufactured holes. Also, bedding-in phenomenonis observed between the dowel and the hole, which elimi-nates any irregular surface roughness and misalignments,contributing, in practice, to increase this gap. Based ondirect measurements and on the unloading branch of theloadslip curves, we concluded that gaps in the range of 0.10.3 mm are realistic. Thus, simulations were carried

out considering the two extreme gap values of the referredrange.

Table 5 summarizes the congurations of the simulationsand respective numerical results. Results include the jointstiffness (slip modulus), the maximum and minimum stres-ses on the surface of the holes and some relevant contact

results, such as the maximum penetration, pressure andsliding distance. Also, the number of iterations is includedin the table, since it serves as a measure of the computa-tional cost, which is important for nonlinear simulations.

The joint stiffness is the only numerical result that can beassessed using the available experimental data, namely the joint stiffness obtained at the unloadingreloading stages(Table 2 ). The experimental average stiffness falls withinthe range of numerical results. The gap has a determinanteffect on stiffness values; its increase results in a systematicstiffness reduction. For frictionless contact and usingdefault values of FKN and FTOLN with GAP = 0.3 (con-guration 08), simulations yield about the same value of the average experimental stiffness. Considering frictioncoefficient of 0.5 (conguration 20) this stiffness slightlyincreases getting even better. Consideration of a non-nullfriction coefficient yields a slightly results change, with anincrease in computational cost, since the stiffness matrixof the structure become unsymmetric.

The FKN factor also plays an important role on contactsimulation as demonstrated by the numerical results of Table 5 . In fact, the reduction of this factor produces a sig-nicant reduction on the joint stiffness. Apparently theFTOLN parameter has a minor inuence on results; onlyfor small values of this parameter, changes in results are

visible since more iterations are needed to accommodatea smaller penetration. In all simulations the numerical pen-etration was kept very small as certied the maximum pen-etration value of Table 5 (17 l m).

Fig. 9 illustrates the stress eld acting in parallel-to-grain direction ( y) for the conguration 20. It is clear thatthe extreme values are observed around the dowel and the

Fig. 9. r y stresses in MPa obtained with conguration 20.


8/10/2019 616_SantosCL2009

11/12

through thickness stress distribution is not uniform. Themaximum compressive stress is veried in the inner woodmember, which is consistent with the experimental tests.In fact the failure occurred always in the inner member the most stressed one. Fig. 10 shows some contact results,in particular the contact penetration, pressure and sliding

distance, again for the conguration 20.

4. Concluding remarks

This paper reported both experimental and numericalwork about the mechanical behaviour of a double-shearsingle dowel wood connection made of Portuguese pinewood. The experimental work also included embeddingtests in order to generate strength data required for the joint analysis. As far as the authors are aware there areavailable in literature very few data about the Portuguesepine wood, hindering their application for structural pur-poses. The generated data was assessed using EC5 proce-dures and some discrepancies were observed, namelythere is no evident correlation between the slip modulusor stiffness and wood density. The embedding test dataderived under tensile loading is the most appropriate forassessing the tested joint; signicant differences can befound if compressive embedding test data is used in the joint assessment. An analytical model supported by statis-tical regression analysis data is suggested to estimate the joint slip modulus, based on data from the tensile embed-ding tests, which gave a reasonable prediction. Finally, a3D FE model is proposed for the joint which was cali-brated using the available experimental stiffness values.Since the model was able to simulate very accurate stiffnessvalues, authors believe that it will be also able to simulatethe stress eld in the hole vicinity. It is important to referthat there are only few 3D FE models for wood connec-tions available in literature, which makes the proposedmodel an important contribution.

Acknowledgements

The authors whish to express their gratitude to the grad-uation students, Vitor Costeira and Ricardo Leal, for theircollaboration in some experimental and numerical tasks,

and also to Mr. Armindo Teixeira for his contribute onspecimens preparation.

References

[1] Itany RY, Faherty KF. Structural wood research, state-of-the-art andresearch needs. NY: ASCE; 1984.

[2] Soltis LA, Wilkinson TL. Bolted-connection design. General techni-cal report FPL-GTR-54. Forest Products Laboratory USDA.Madison, WI, 1987.

[3] European Committee for Standardization. EN 1995-1-1 Design of timber structures. Part 1-1: General rules and rules for buildings.Brussels, 2004.

[4] Johansen KW. Theory of timber connections. Int Assn Bridge Struct

Eng Pub. 9. Bern, Iasbe, Zurich, Switzerland, 1949. p. 24962.Fig. 10. Contact values obtained with conguration 20: (a) penetration inmm; (b) pressure in MPa and (c) sliding distance in mm.


8/10/2019 616_SantosCL2009

12/12

[5] European Committee for Standardization. EN383: Timber struc-tures. Test methods. Determination of embedding strength andfoundation values for dowel-type fasteners. European Standard,1993.

[6] Patton-Mallory M, Pellicane PJ, Smith FW. Modelling boltedconnections in wood: review. J Struct Eng 1997;123:105462.

[7] Sawata K, Yasamura M. Estimation of yield and ultimate strengthsof bolted timber joints by nonlinear analysis and yield theory. J WoodSci 2003;49:38391.

[8] Chen CJ, Lee TL, Jeng DS. Finite element modelling for themechanical behaviour of dowel-type timber joints. Comput Struct2003;81:27318.

[9] Racher P, Bocquet JF. Non-linear analysis of dowelled timberconnections: a new approach for embedding modelling. Electron JStruct Eng 2005;5:19.

[10] Kharouf N, McClure G, Smith I. Elasto-plastic modelling of woodbolted connections. Comput Struct 2003;81:74754.

[11] Patton-Mallory M, Cramer SM, Smith FW, Pellicane PJ. Nonlinearmaterial models for analysis of bolted wood connections. J Struct Eng1997;123:106370.

[12] Moses DM, Prion HGL. A three-dimensional model for boltedconnections in wood. Can J Civil Eng 2003;30:55567.

[13] European Committee for Standardization. EN26891: Timber struc-tures. Test methods. Determination of embedding strength andfoundation values for dowel type fasteners. European Standard, 1993.

[14] Swanson Analysis Systems Inc. ANSYS. Version 10.0. Houston,2005.

[15] Xavier JC. Caracterizac ao do comportamento ao corte da madeirausando o ensaio de Iosipescu. MSc Thesis, University of Tra s-os-Montes and Alto Douro, Vila Real, Portugal, 2003.

[16] Pereira J. Comportamento mecanico da madeira em tracc ao nasdireccoes de simetria material. MSc Thesis, University of Tra s-os-Montes and Alto Douro, Vila Real, Portugal, 2003.

[17] Oliveira M. Caracterizac ao do comportamento ao corte da madeirausando o ensaio de Arcan. MSc Thesis, Universidade of Tra s-os-Montes and Alto Douro, Vila Real, Portugal, 2003.

[18] Garrido N, Identicac ao do comportamento ao corte da madeira,atrave s do ensaio de tracc ao fora dos eixos de simetria material. MScThesis, Universidade of Tra s-os-Montes and Alto Douro, Vila Real,Portugal, 2004.


616_santoscl2009

Documents