Fatigue Stength of Dowel Joints in Timbe Stuctues Kjell Ane Malo Pofesso Nowegian Univesity of Science and Technology, NTNU Tondheim, Noway Åge Holmestad Diecto Moelven Limte A/S Moelv, Noway Pe K. Lasen Pofesso Nowegian Univesity of Science and Technology, NTNU Tondheim, Noway Summay The inceased use of lage span glulam timbe stuctues is in pat due to the development of moe efficient connections with slotted-in steel gusset plates combined with steel dowels. In timbe oad bidges, also the fatigue behaviou of such connections must be consideed. The pape shows that fatigue damage and failue may occu in timbe stuctues with dowel connections when subjected to fatigue loading. The fatigue stength is dependent on the stess atio, which is the atio between the minimum and imum stess. The fatigue stength can be consideed to be a linea decaying function of the logaithm of the numbe of cycles. Use of imum stess as fatigue stength paamete seems to be moe easonable than use of the stess ange. The bidge pat of the new Euocode 5 timbe code has poposed fatigue veification ules in an infomative annex, and the fatigue design veification is commented in view of the pesent test esults.. Intoduction Duing the last decades the use of glulam timbe in lage span tuss stuctues has inceased significantly. In Noway the athletic facilities built fo the 994 Winte Olympics at Lillehamme may seve as examples of this tendency whee the speed skating ink at Hama (The Viking Ship) may be best known. Moe ecently, the teminal building at the new Oslo Aipot was constucted using a glulam time oof system. In all cases the stuctual systems wee exposed and used as an achitectual featue. One of the key factos which made these lage timbe stuctues competitive economically is the development of joints with slotted-in steel plates (gusset plates) combined with steel dowels. This type of joints has now been used in many timbe oad bidges []. Evenstad Bidge was the fist lage timbe bidge in the Nodic counties, and was opened fo oad taffic in 996. Late, lage bidges have been built including an ach bidge with a span of 7 metes at Tynset, Noway, as well as a bidge fo vey heavy militay taffic at Rena, Noway. Wheeas buildings ae subjected to pedominantly static loading, the moving taffic loads on the bidges may cause fatigue effects. The numbe of load cycles and load amplitudes will depend on the taffic patten at the bidge site. Howeve, at pesent thee is vey little infomation available on the fatigue stength of timbe connections, and the design codes give few guidelines. In a compehensive liteatue suvey [2] only one pape on this subject was identified [3]. Hee, thee connections with 4 dowels in a ow paallel to the gain wee tested, and the esidual static stength afte a pescibed numbe of load cycles was detemined. The tentative conclusion stated that the
fatigue stength was good [3]. Fatigue of bolted connection is addessed in [2] and [4] while the fatigue behaviou of wood mateial unde vaious loading conditions has been addessed in [5]. Examples on applications involving metal fastenes and a compehensive teatment on the subject factue and fatigue of wood may be found in [6]. Clealy, in ode to extend the use of timbe oad bidges to lage spans moe data on fatigue of dowel joints in timbe stuctues ae needed. The pesent design ules ae vey limited and most likely insufficient o non-documented. Futhemoe, the theoetical basis fo the damage evolution and a pedictive analytical theoy seems to be lacking. Fo this easons a Nodic eseach pogamme was pefomed, whee tests have been caied out on dowel type connections with 2 o 3 dowels in one ow pependicula to the gain [7], [8]. The pesent investigation extends this wok to lage specimens and moe dowels in each connection. 2. Fatigue veification Fatigue design fo wood-based mateials ae summaised in [9]. The fatigue life of engineeing mateials is taditionally epesented in the fom stess-life o stain-life cuves following Wöhle s classical wok. The fatigue stength is nomally detemined on the basis of cyclic tests with constant amplitude. Such a stess o load pocess is uniquely defined by the stess ange Δ and the stess atio R with the following definitions: Δ = min and R min = [] whee and min ae the chaacteistic imum and minimum stesses espectively fom the fatigue action. The use of Δ as the govening stess paamete fo fatigue design is simila to the cuent pactice fo welded steel stuctues. Fo steel stuctues thee exists esidual welding stesses as high as the yield stess f y in tension at the location of a potential fatigue cack, and the mateial will hence expeience stesses that vay between f y and (f y - Δ ) at this position. Fo this eason Δ is used as the fatigue stess paamete. Fo timbe stuctues thee seems to be few physical easons and little expeimental evidence that justifies the use of Δ as the only stess measue. Based on esults epoted in [7] and [8], and on the geneal fatigue data fo wood [9] and [], it seems appaent that the stess atio R should be included when detemining the fatigue stength. Futhemoe, taking into account the effect of the duation of load on the stength of wood, it might be easonable to conside the use of as an altenative load paamete. Both the stess ange Δ and may be nomalized by a efeence stess which typically will be a tensile, compessive o a shea stength. A nomalized imum stess may then be expessed by: Δ f = = ( R) [2] To evaluate the fatigue stength of dowel joints in wooden membes the nomalised stess ange Δ f =Δ o nomalized imum stess f in possibly combinations with the stess atio R have been used heein. The fatigue design veification then may ead: Δ =Δf f o = f f [3] fat, d fat, d whee, fat d f is the design fatigue stength elating the fatigue stength to the numbe of cycles N, fo given joints lay-outs and mechanical fastenes.
3. Fatigue Testing of Dowel Joints Figue Specimen mounted in test ig. Figue 2 Specimen with two simila connections. Evenstad bidge [], which is used as a model fo the joints consideed, has five equal spans, each with a span width of 36 metes. Each span consists of timbe tuss with a cuved uppe cod. Depending on the atio of dead weight elative to the taffic magnitude and location of the loading, the axial foce in the membe will vay between tension and compession. In ode to limit the scale effects and to epoduce test conditions simila to that of a pototype, it was decided to use lage-scale specimen of a splice type connection with geomety esembling that of Evenstad Bidge. The tests wee done with uniaxial loading. The mateial used was Nodic pine (Pinus silvestis), and all specimen wee poduced fom a single batch of machine gaded timbe C3 (MT3) with a chaacteistic bending stength of 3 MPa. The membe length was 68 mm, and the coss-section width and thickness wee 2 mm and 4 mm espectively. The connections had two gusset plates and 2 dowels, see pictue on Figue and the dawing on Figue 2. This specimen design allowed two connections to be fatigue tested simultaneously. Dowels and holes with diametes 2 mm and 2.5 mm espectively wee used. The spacing of the dowels paallel to the gain was mm, and the distance to the fee end was 2 mm. In the diection nomal to the gain the spacing was mm and the edge distance was 5 mm. The thickness of the oute wood layes wee 3 mm while the middle laye was 6 mm. Note that the length of the dowels was 2 mm, which is less than the 4 mm thickness of the specimen. The dowel connection was designed such that the static stength was limited by the embedding stength of the timbe, without yielding in the dowels. The mateial popeties of the dowels wee not detemined, but based on pevious tests yield stess was estimated to exceed 6 MPa, with a ultimate stess f u of about 75 MPa.The test specimens wee peconditioned at 65 % RH and 2 ο C (2% moistue), and in the test laboatoy the tempeatue was kept constant at 2 ο C and the humidity was contolled. To study the impotance of the stess atio R on the fatigue stength, the tests wee pefomed with two stess atios, R = - and.. The most likely stess atio fo the connections of the tuss membes in bidges of the Evenstad type would be R >, as load evesals would be athe unlikely. Howeve, it was consideed that a stess atio of R = - would be the most exteme situation, and hence it was included. The tests wee conducted at the Nowegian Univesity of Science and Technology, NTNU. The tests wee pefomed with a load fequency anging fom -4. Hz. The loading fequency may be of impotance in timbe stuctues [5], as it may goven the tempeatue histoy and moistue content duing the tests. Specimen with installed gusset plates and dowels is shown in Figue 2. A view of the test specimen afte installation in the test fame is given in Figue. Taditional test fames and hydaulic actuatos wee used.
3.. Load paametes. The static efeence values wee detemined fom thee specimens andomly dawn fom the batch. The static tests gave a mean connection esistance F = 4 kn. The fatigue test esults ae given in tems of the associated numbes of cycles to failue N (log scale) vesus nomalised stess anges Δ f o imum stess f with espect to F. The load paametes wee detemined by ( ) Δ f = F F F [4] min / whee F and F min ae the imum and minimum loads, and the stess atio R given by R = Fmin F and the elation to the nomalised imum stess is f F Δf R = = F The fatigue loading conditions and esults ae descibed in moe detail in [], [2] and [3]. [5] [6] 3.2. Cycles to failue The esults fom 48 tests on dowel connections ae plotted fo R =. in Figue 3, and fo R = - in Figue 4, whee the nomalised stess anges Δ f is used as paamete. The same esults ae also plotted in Figue 5 and Figue 6, but hee the nomalised imum stess f has been used as load paamete in addition to stess atio R. The static efeence tests have been egaded as a quate cycle load with a vey low load fequency. Δf..4.3.2. R=. - 2 3 4 5 6 7 8 Figue 3 Stess ange vs. cycles to failue. Δf..4.3.2. R=- - 2 3 4 5 6 7 8 Figue 4 Stess ange vs. cycles to failue. In some of the specimens damage and failue developed vey simila in both ends and it was not possible to detemine the ode of the connection failues. In othe cases the specimen was too damaged to be stengthened such that the emaining end connection could be tested futhe. The numbe of cycles to failue in those cases was set equal fo the two connections. In a numbe of cases the factued connection was emoved and eplaced by a new stengthened connection,
making it possible to cycle the othe connection to failue. A small numbe of the connections failed due to fatigue in the steel gusset plates, followed by damage to the timbe. Note that the tests coesponding to the smallest Δ f o f wee teminated without failue due to limited equipment availability (Fatigue testing at low fequencies is time consuming!). Hence, using all the data in an estimation of fatigue stength will give esults on the safe side, since not all specimens actually have failed due to timbe fatigue at the numbe of cycles given. See efeence [3] fo futhe details. Howeve, in view of the vey few test esults available at pesent, it seems easonable to use all the acquied infomation. f..4.3 R=..2. - 2 3 4 5 6 7 8 Figue 5 Maximum stess vs. cycles to failue. f..4.3 R=-.2. - 2 3 4 5 6 7 8 Figue 6 Maximum stess vs. cycles to failue. 4. Discussion Based on the liteatue [9] and vaious poposals fo design ules fo dowel connections, the choice of fatigue stength paametes most appopiate fo design is not obvious. The issues addessed hee ae the shape of the fatigue stength cuves, the significance of the stess atio R and the choice of stength paamete Δf o f. The stess ange Δf and the imum applied stess f ae plotted as functions of the log cycles to failue (log N) in Figue 3 to Figue 6. The expeimentally obtained numbe of cycles to failue, N, fo high values of Δf o f do not suppot any assumptions of keeping the fatigue stength unaffected of N up to a given numbe, fo example 4 cycles. In fact, the fatigue stength of the timbe connections stats to degade afte vey few cycles. Futhemoe it has been supposed that thee is a fatigue limit at a given level of stesses, leading to the assumption that no fatigue will take place fo stesses below that limit. Based on the pesent esults fo dowel joints no conclusion can be made. Howeve, if such a fatigue limit exists, it will pobably be below nomalized stess level of.25 and only of inteest fo lage numbe of cycles than 7. This implies that fo many bidges a fatigue limit will be ielevant. The expeimental esults, iespective of using Δ f o f as stength paamete, give few aguments fo a moe complicated stength function than a linea dependency of. Note that this fomat is diffeent fom the Log-Log fomat used fo fatigue of metal stuctues. Fo futhe discussions, a best fit of a staight line detemined by the method of least squaes have been added to the plots of the expeimental esults. The acquied data points have been fitted to the linea equation f = Alog N + B [7]
with the estiction that the static value ( = - ) should equal unity. The obtained constants ae listed in Table. Hee, f is the expeimentally detemined stength paamete, eithe f o Δ f, while f is the coesponding quantity calculated by Eq. [7]. The nomalised esidual, in addition to the imum deviations d between expeimental and calculated stength, may seve as a measue of the goodness of the fit. The computations ead: = ( f f ) 2 n n and d f f Table Resulting paametes to Eqs. [7] -[8]. 2 = ( ) [8] f R A B D The fatigue data have been plotted sepaately f fo the two R-values in ode to assess the effect. & - -.72 56.93.223 of the stess atio R on the numbe of cycles to Δ f. & - -.7 58.83.28 failue. Besides the static data, we have only f. -.66 6.59.26 seven data points fo R = -, which cannot be Δ f. -.74 55.46.3 consideed to be sufficient fo any clea f conclusions. In Table the two fist lines - -.98 4.38.7 contains esults fom computations accoding to Δ f - -.48 7.78.47 Eqs. [7] - [8] without making any distinction between the stess atios R. Compaing these two fist lines with the othe lines whee the two diffeent stess atios R ae accounted fo, we can see that the goodness of fit (paametes and d) is much bette when stess atio is taken into account. This is independent of the choice of stess paamete, f o Δ f. Compaing to the esults obtained fo R =., Figue 3 and Figue 5, with the plots fo R =, Figue 4 and Figue 6, it is seen that the slopes (A in Table ) of the fitted lines diffe consideably. It is theefoe appaent that the stess atio R should be included in fatigue design of dowel joints in timbe stuctues. In this investigation both Δ f and f have been used as altenative stength paametes. By compaison of Figue 3 and Figue 5, no significant diffeences may be obseved with espect to the epesentation and goodness of the fits. Note that thee is only % diffeence between the two stess measues f o Δ f fo R =.. Howeve, consideing R = - only, see Figue 4 and Figue 6, f should be the pefeed stength paamete. Fo this case the calculated esults fo and d in Table give fo f only half of the values obtained fo Δ f. Theefoe, the choice between the stength paametes Δ f and f seems to be ielevant fo R =., but might be impotant fo altenating loading e.g. R = -. In that case, the use of f gives a bette epesentation of the fatigue stength. It seems to be no physical easons to pefe Δ f instead of f as the stength paamete. Howeve, this conclusion is made on the basis of vey few available expeimental esults and moe expeiments should hence be conducted. 4.. Euocode 5 Fatigue Veification Based on the expeimental esults, included those epoted heein, a linea dependency of simila to Eq. [7] has been poposed as basis fo fatigue design in Euocode 5 [4], whee bidges in paticula ae coveed in Pat 2 [5].
Using the notations in Euocode 5, the expessions fo the fatigue design veification ae: kfat fk d, f fat, d and f fat, d = [9] γ M, fat f fat, d elates the fatigue stength to the coesponding chaacteistic static stength f k. γ M, fat is the patial safety facto fo mateial popeties in fatigue and equals.. The fatigue stength can then be given by a fatigue stength atio k fat, which depends on the stess atio R and the numbe of cycles N. k fat can be diectly compaed to the tem f used heein, and it is calculated by R kfat = logn [] ab ( R) Fo dowels connections the paametes ae a = 6 and b = 2. These paametes coespond, in Eq. [7], to the slopes A =.789 and A =. fo R =. and R = espectively, see Table fo compaison with the expeimental esults. In Figue 7 and Figue 8, Eq. [] has been plotted togethe with the pesent fatigue esults. Note that k fat and f ae a dimensionless paametes which elates the fatigue stength to the static stength. The design consideations in Euocode 5 ae a combination of the paametes a and b and the use of chaacteistic static stength f k. The effect of moistue vaiation is not pesent in the test esults. It is well known that highe moistue contents and vaiations in moistue can lowe the stength consideably. Euocode 5 does not seem to account fo this effect, which pobably should be done in the futue. f..4.3.2. - 2 3 4 5 6 7 8 Figue 7 Maximum stess vs. cycles to failue, R =., EC5 compaed to test esults. f..4.3.2. - 2 3 4 5 6 7 8 Figue 8 Maximum stess vs. cycles to failue, R =, EC5 compaed to test esults. 5. Concluding Remaks and Acknowledgements The static and fatigue stength of a dowel connection fo a glulam timbe membe have been tested. The esults fom six static component tests and 42 fatigue-tested connections have been pesented. The following concluding emaks ae made: Fatigue damage and failue may occu in timbe stuctues with dowel connections when subjected to fatigue loading. The fatigue stength depends on the stess atio, which is the atio between the minimum and imum stess. The fatigue stength can be consideed to be a linea decaying function of the logaithm of the numbe of cycles.
Use of imum stess as fatigue stength paamete seems to be moe easonable than use of the stess ange. A compaison with the fatigue design veification in Euocode 5 has been made. The pesent investigation was pefomed unde the poject Fatigue of Timbe Bidges, which constitute a pat of the Nodic Timbe Poject. The initiative and funding by the Nodic Timbe Council is gatefully acknowledged. 6. Refeences [] Solli, K.H; Timbe Tuss Bidges, Nodic Timbe Bidge Poject, Nodic Timbe Council, Stockholm, Sweden, 997. [2] Keuzinge, H and Moh, B; Holz und Holsvebindungen unte nicht vowiegend uhenden Einwikungen, Technische Unvesität, Institut fü Tagweksbau, München, 994. [3] Möhle, K and Maie, G; Untesuchungen übe das Daueschwingvehalten von Holzvebindungen. Kalsuhe TH, Vesuchanstalt fü Stahl, Holz und Steine, Abt. Ingenieuholzbau, Foschungsbeicht, 973. [4] Abendoth R.E. and Wipf, T.J; Cyclic Load Behavio of Bolted Timbe Joint. Jounal of Stuctual Engineeing, Vol. 5, No. (989) pp. 2496 25. [5] C.O.Cloius, M.Uhe Pedesen, P. Hoffmeye, L. Damkilde. Fatigue Damage in Wood Intenational COST 58 Wood Mechanics Confeence, May, 4-6, 996. [6] Smith, I., Landis, E: and Gong, M: Factue and Fatigue in Wood. John Wiley & Sons, England, 23. [7] Ljøing, J: Boe i tæ. Delposjekt 4.5. Fobindelsesmidle. Pøvningsappot. Dansk Teknologisk Institut, Denmak, 996 (in Danish). [8] Ellingsud, O. and Ljøing, J.; Joints fo Timbe Bidges, Nodic Timbe Bidge Poject, Nodic Timbe Council, Stockholm, Sweden, 997. [9] Ansell, M.P.; Fatigue design fo timbe and wood-based mateials, in Timbe Engineeing, STEP 2, lectue 22, STEP/EUROFORTECH,. ed. Centum Hout, The Nethelands, 995. [] Bonfield, P.W., Ansell, M.P. and Dinwoodie, J.M. Fatigue testing of wood: A det. guide fo develop. of life pediction fomulae fom fatigue data. Timbe Eng. Meeting (994) IUFRO. [] Malo, K.A: Fatigue Tests of Dowel Joints in Timbe Stuctues, Nodic Timbe Bidge Poject, ISBN 9-892-24-5, Nodic Timbe Council AB, Stockholm, Sweden, 999. [2] Malo, K.A., Holmestad, Å and Lasen, P.K, (2): Fatigue Tests on Dowel Joints in Timbe Stuctues, IABSE Confeence Innovative Wooden Stuctues and Bidges, August 29-3, Lathi, Finland. [3] Malo, K.A: Fatigue Tests of Dowel Joints in Timbe Stuctues, Pat II: Fatigue Stength of Dowel Joints in Timbe Stuctues. Nodic Timbe Bidge Poject, ISBN 82-72-35-6, Nodic Timbe Council AB, Stockholm, Sweden, 22. [4] Euopean Committee fo Standadization; Euocode 5 Design of timbe stuctues Pat - : Geneal ules and ules fo buildings, EN 995--: 25. [5] Euopean Committee fo Standadization; Euocode 5 Design of timbe stuctues Pat 2: Bidges, Annex A, EN 995-2: 24.