《木结构 Timber Engineering》课程教学资源（参考文献）An experimental and numerical study of the effect of friction in single dowel joints.pdf_大学文库

DOI 10.1007/s00107-008-0267-z ORIGINALARBEITEN · ORIGINALS Holz Roh Werkst (2008) 66: 363–372 An experimental and numerical study of the effect of friction in single dowel joints Johan Sjodin · Erik Serrano · Bertil Enquist ¨ Published online: 21 August 2008 © Springer-Verlag 2008 Abstract The effect of friction between the dowel and the surrounding timber was studied for single dowel-type joints. The joints tested were divided into two groups of joints, where the surface of the dowels differed. For the first group, the dowels had a smooth surface and for the second group the dowels had a rough surface. A contactfree measurement technique was used in the experimental investigation. In addition to that, numerical simulations were carried out aimed at predicting the load-bearing capacity of the joints tested as well as estimating the coef- ficient of friction between the dowel and the surrounding timber. Important conclusions from this study, which are supported by previous research, are that the load-bearing capacity of single dowel-type joints increases when the surface roughness of the dowel increases. A very small scatter in the results, in terms of the load-displacement behavior, was seen in the tests with rough surface dowels. For the joints with smooth surface dowels, the elastic response as well as the plastic response varied considerably between different tests. The current version of the European timber code EC5 does not explicitly take into account the effect of friction. In order to take the effect into account embedding tests have to be performed in order to be able to consider the benefits of using dowels with rough surface. J. Sjodin ( ¨ ✉) · E. Serrano · B. Enquist School of Technology and Design, Vaxj ¨ o University, ¨ Luckligs plats 1, ¨ 351 95 Vaxj ¨ o, Sweden ¨ e-mail: johan.sjodin@vxu.se Experimentelle und rechnerische Untersuchung des Einflusses von Reibung in Verbindungen mit einem Stabdubel ¨ Zusammenfassung Untersucht wurde der Einfluss der Reibung zwischen dem Dubel und dem umgebenden Holz. ¨ Zwei verschiedene Verbindungen mit jeweils unterschiedlicher Dubeloberfl ¨ ache wurden untersucht. In einer Verbin- ¨ dung wurden Dubel mit einer glatten Oberfl ¨ ache und in ¨ der zweiten Verbindung Dubel mit einer rauen Oberfl ¨ ache ¨ verwendet. In der experimentellen Untersuchung wurde ein beruhrungsloses Messverfahren eingesetzt. Zus ¨ atzlich ¨ wurden Simulationsrechnungen zur Berechnung der Tragfahigkeit der Verbindungen sowie der Sch ¨ atzung des Rei- ¨ bungskoeffizienten zwischen Dubel und umgebendem Holz ¨ durchgefuhrt. ¨ Als wichtigstes Ergebnis zeigte sich in Ubereinstimmung ¨ mit fruheren Untersuchungen, dass die Tragf ¨ ahigkeit einer ¨ Stabdubelverbindung mit zunehmender Oberfl ¨ achenrauhig- ¨ keit des Dubels zunimmt. Bei den Versuchen mit rauer ¨ Dubeloberfl ¨ ache wiesen die Kraft-Weg-Diagramme nur eine ¨ geringe Streuung auf. Das elastische und plastische Verhalten der Verbindungen mit glatter Dubeloberfl ¨ ache streute ¨ dagegen zwischen den verschiedenen Versuchen beachtlich. In der aktuellen Fassung des Eurocode 5 wird der Einfluss der Reibung nicht explizit berechnet. Um diesen Einfluss zu berucksichtigen, m ¨ ussen Versuche zur Lochleibungsfestig- ¨ keit durchgefuhrt werden, um den Vorteil der Verwendung ¨ von Dubeln mit rauer Oberfl ¨ ache nutzen zu k ¨ onnen. ¨ 1 Introduction The design rules contained in the European timber code EC5 (Eurocode 5 2004) for single bolted joints loaded par- 13

364 H6 z Roh Werkst(200866:363-372 oeopmec a given load.adhesives that differed markedly from on another.In Sjodin et al.(2006)the same measurement the timber and the dowels.Restrictions related to end and system was used for steel-timber dowel joints in order to nces and timber thickness are also included in study the strain distribution in the joint area Several ob failure de in that stud on of th opt th nt etr dete fi=0.0821-0.01d)p and shear strains and tensile strains,both parallel and per pendicular to the grain were found to be concentrated in the where is the dowel diameter andis the characteris areas most likely to influence the final failure mode of the tic timber d n has been derived from oint dard 1993BLaB2003 this and partly y numerical listed several narameters that affeet the embedment strenath of timber eg the timber density.the dowel and the hole adopted for studying the joint area in single dowel joint diameters;the angle between the load and the grain direc loaded in tension parallel to thegrain The joints studie tion and the mo sture content.Anothe r parameter, which were divided into two groups of joints The dowels o t paper, h a smooth su ace and the dowels of group 2 hac tim be al and s al w all aim of this w sto closely study the in this area (Rodd 1973.Siem 1999).Results have shown effect of friction between the dowel and the surroundin that increasing the surface roughness of the dowels increases timber by using contact-free measurements and numerical the embedment strength:changes the location of the ini- simulations.More specifically the overall aim can be di- tial crack/cracks and reduces the brittle tendencies of the vided into the following three aims: distribution in the dow to he (1973)and Siem (1990) plex (see es stress or strain is thus difficult to study by experimen- 2.Estimation of the coefficient of friction between the tal methods,implying that the dowel and the surrounding timber for each of the groups behavior of dowel typ of joints by comparing experimental results,based on oints are no yet h retical methods order to of he l sual in experimental studies of joints to of ioints b gathe other than that of a joint's global load-displac ement re. elastic fracture mechanics (LEFM concepts sponse.it has been difficult to verify such theoretical studies Thus.fulfilling aims 2 and 3 the present work is comple with experimental ones.Modern contact-free measuremen mentary to previous research about the effect of friction seems t toemploy order to between the dowel and the surrounding timber of joints. to perform suchm ts of the defo 2 Experimental of a surface have been developed and commercialized. These techniques are usually based on electronic speckle 2.1 Specimens and test setup interferometry (ESPI)with use of laser, are based on The type of joint used in the study is shown in Fig.1a rano an lagconelation(DICiehniques all joints test ialized dic. for timber joints.For example s uist (2005 520kg/m at 12%moisture used one such measurement system in order to study the of the annual rings for the timber can he seen in fig lh strain distribution along wood adhesive bonds.The re The steel dowels were 20 mm in diameter,with a tolerance sults showed that the measurement system was capabl .The diameter of the drill was20mm.End dis of distinguishing,in terms of their strain distributions at tances were set according to the minimum values as given in 鱼Springer

364 Holz Roh Werkst (2008) 66: 363–372 allel to the grain are based on the Johansen yielding theory (Johansen 1949), which assumes a rigid plastic behavior of the timber and the dowels. Restrictions related to end and edge distances and timber thickness are also included in EC5 in order to avoid brittle failure modes. One important parameter to adopt the design equations is the determination of the embedment strength of the timber. In EC5, this strength parallel to the grain is given by fh0k = 0.082 (1−0.01d ) ρk , (1) where d is the dowel diameter and ρk is the characteristic timber density. This expression has been derived from large experimental series, based on standardized tests according to EN 383 (European standard 1993). Blaß (2003) listed several parameters that affect the embedment strength of timber, e.g, the timber density, the dowel and the hole diameters; the angle between the load and the grain direction and the moisture content. Another parameter, which is considered in the present paper, is the effect of friction between the dowel and the surrounding timber. Extensive experimental and some theoretical work have been made in this area (Rodd 1973, Siem 1999). Results have shown that increasing the surface roughness of the dowels increases the embedment strength; changes the location of the initial crack/cracks and reduces the brittle tendencies of the joints. The stress distribution in the area close to the dowels for dowel-type joints is known to be highly complex (see e.g. Sjodin et al. 2006). A detailed state of ¨ stress or strain is thus difficult to study by experimental methods, implying that the behavior of dowel type joints are not fully understood yet. Theoretical methods have been used in the literature in order to study the joint area in more detail. But, because it has been rather unusual in experimental studies of joints to gather data other than that of a joint’s global load-displacement response, it has been difficult to verify such theoretical studies with experimental ones. Modern contact-free measurement techniques should be possible to employ in order to fill this gap that seems to exist between advanced theoretical methods and experimental approaches. Today, techniques to perform such measurements of the deformation field of a surface have been developed and commercialized. These techniques are usually based on electronic speckle interferometry (ESPI) with use of laser, or are based on white light digital imaging correlation (DIC) techniques (Serrano and Enquist 2005). Commercialized DIC-techniques have recently been used for timber joints. For example, Serrano and Enquist (2005) used one such measurement system in order to study the strain distribution along wood adhesive bonds. The results showed that the measurement system was capable of distinguishing, in terms of their strain distributions at a given load, adhesives that differed markedly from one another. In Sjodin et al. (2006) the same measurement ¨ system was used for steel-timber dowel joints in order to study the strain distribution in the joint area. Several observations of considerable interest were made in that study on the basis of numerical results in combination with results from the contact-free measurement system. A nonuniform strain distribution in the joint area was detected, and shear strains and tensile strains, both parallel and perpendicular to the grain were found to be concentrated in the areas most likely to influence the final failure mode of the joint. In this partly experimental and partly numerical study, the same contact free measurement system as used in Serrano and Enquist (2005) and Sjodin et al. (2006) was ¨ adopted for studying the joint area in single dowel joints loaded in tension parallel to the grain. The joints studied were divided into two groups of joints. The dowels of group 1 had a smooth surface and the dowels of group 2 had a rough surface. The overall aim of this work was to closely study the effect of friction between the dowel and the surrounding timber by using contact-free measurements and numerical simulations. More specifically the overall aim can be divided into the following three aims: 1. Verification of the effect of friction between the dowel and the surrounding timber that has been shown in Rodd (1973) and Siem (1999). 2. Estimation of the coefficient of friction between the dowel and the surrounding timber for each of the groups of joints by comparing experimental results, based on contact-free measurements, and numerical results. 3. Prediction of the load-bearing capacity of the two groups of joints by means of numerical analyses and using linear elastic fracture mechanics (LEFM) concepts. Thus, fulfilling aims 2 and 3 the present work is complementary to previous research about the effect of friction between the dowel and the surrounding timber of joints. 2 Experimental 2.1 Specimens and test setup The type of joint used in the study is shown in Fig. 1a. The timber for all joints tested was taken from the same board. Scots pine was used and the mean density was about 520 kg/m3 at 12% moisture content (MC). The orientation of the annual rings for the timber can be seen in Fig. 1b. The steel dowels were 20 mm in diameter, with a tolerance of ±0.2 mm. The diameter of the drill was 20 mm. End distances were set according to the minimum values as given in 13

Holz Roh Werkst (2008) 66: 363–372 365 Fig. 1 a Configuration setup, b end-grain view of timber specimen with the orientation of the annual rings, c a dowel belonging to group 1 with a smooth surface, d a dowel belonging to group 2 with a rough surface together with characteristics of the knurled surface Abb. 1 a Geometrie der Verbindung, b Ansicht des Hirnholzes des Prufk ¨ orpers mit ¨ Jahrringverlauf, c Stabdubel der ¨ Gruppe 1 mit einer glatten Oberflache ¨ d Stabdubel der ¨ Gruppe 2 mit einer rauen Oberflache und deren Struktur ¨ EC5. All the specimens were stored in a standard climate of 20 ◦C and 65% RH (relative humidity) until an equilibrium MC of 12.8% (COV = 0.3%) was reached, prior to further preparation. A total of eight specimens were employed divided into two different groups. The characteristics of the groups were as follows: for group 1, the dowels had a smooth surface in order to give the smallest possible friction between the dowel and the surrounding timber (Fig. 1c) and for group 2, the dowels were knurled in order to increase the friction between the dowel and the surrounding timber (Fig. 1d). After manufacturing, the joints were loaded parallel to the grain until failure. During the loading procedure, the dowel was placed in a specially made fork connection (Fig. 2). The other end was first reinforced with plywood and was then connected to a similar fork. During the loading procedure, the joints were subjected to a displacement-control rate of 1 mm/min. During the tests, the load was recorded and the relative displacement between the fork connection and the timber was measured by two inductive gauges attached on opposite sides of the joints (Fig. 2). 2.2 Contact-free measurement system The measurement system ARAMISTM manufactured by the company GOM was employed. The system is based on evaluating a random or regular pattern, which is applied to the surface and deforms along with the material. The placing of two CCD cameras (1280 by 1024 resolution) in front of the specimen at different angles enables stereoscopic pictures of the patterned surface to be taken. This is done during different occasions in time, during loading. The first digital-image processing step defines macro-image facets in the image pair in the original, unloaded state. For each stage of loading, the 3D coordinates of these facets on the specimen’s surface are calculated accurately using image correlation and photogrammetric principles. On the basis of these 3D coordinates, the 3D displacement field and the corresponding surface (in-plane) strain field on the specimen’s surface can be calculated with a high degree of spatial resolution. For a further description of the system, see for example Serrano and Enquist (2005). The CCD cameras were calibrated to a measurement volume that included the joint area and part of the dowel; see the area inside the dashed lines in Fig. 2. A random pattern was achieved in two steps. First a matt lightcolored paint was sprayed on the area around the dowFig. 2 Test setups Abb. 2 Versuchsanordnung 13

366 Holz Roh Werkst (2008) 66: 363–372 Table 1 Material property parameters for timber used in the numerical work. The elastic properties are taken from Dinwoodie (2000). The fracture and strength parameters are taken from Aicher et al. (2002) Tabelle 1 Materialeigenschaften des in der rechnerischen Untersuchung verwendeten Holzes. Elastische Eigenschaften nach Dinwoodie (2000), Bruch- und Festigkeitsparameter nach Aicher et al. (2002) Direction Young’s modulus Shear modulus Poisson’s ratio Fracture energy G Strength (MPa) (MPa) (–) (J/m2) (MPa) Grain direction (L) EL = 16 300 GLT = 680 VLT = 0.51 VTL = 0.015 GIC = 200 ft = 3 Tangential direction (T) ET = 570 GLR = 1160 VLR = 0.42 GIIC = 3.5GIC fv = 9 Radial direction (R) ER = 1100 GTR = 66 VTR = 0.31 els. This was done in order to obtain better contrast of the gray scale defining the facets and also to reduce the shininess of the timber. After this, small black dots were applied by spraying black paint on the surface from a distance. The cameras were triggered every five seconds, also logging analog signal readings of displacement and force from the loading machine. The image pairs captured during testing were then processed by the software included in the ARAMIS-system. For each pair of images, 3D coordinates for a large number of facets could be determined and the corresponding surface strains could be calculated. The calibration volume selected results in a spatial resolution of approximately 1.2 mm. According to the manufacturer, the strain accuracy is approximately 0.01%. 3 Numerical simulation The FE-software ABAQUS (ABAQUS Inc 2004) was used for the 2D numerical analyses of the joint. The element subdivision in the joint area, shown in Fig. 3, consists of approximately 8200 elements. Quadrilateral linear elements were employed. Plane stress state and linearelastic behavior was assumed for the wood. The moduli of elasticity, the shear moduli and the Poisson’s ratios used to define the orthotropic wood material were set according to Table 1. These properties were taken from Dinwoodie (2000) and should correspond to the timber used in the experimental work. The indices L, R and T in Table 1 denote the longitudinal, radial and tangential directions of the timber, respectively. As indicated in Fig. 3, the plane of analysis in the 2D calculations coincides with the LR-plane of the timber. This approximation is reasonable, c.f. the orientation of the annual ring pattern in Fig. 1b, especially when also considering the centrically placed dowel. The load was applied as a uniform deformation of the dowel in the longitudinal direction of the timber (Fig. 3). The interaction between the dowel, which was assumed to act as a rigid body, and the surrounding timber, was modeled by contact elements involving friction along the contact surface. This coefficient of friction (µ) was estimated by comparing the results in terms of the straindistribution using the ARAMIS-system and using the FE-model. Fig. 3 Element subdivisions for the numerical work Abb. 3 Elementeinteilung fur die rechnerische Untersuchung ¨ 13

Holz Roh Werk过(2008)66：363-372 367 tim ised assuming stress distributing capabilities of the wood The basic concept of this method,which is based on lin- ear elastic fracture mechanics (LEFM),is to evaluate the 4 Results and discussion acting across an area rather than to evaluate 4.1General observations 2002.s 。andCs0n2006LA The load-displacement cu then used capacity.The esforajointstesteadpoecnem in Fig.4.are based on the mean value of the disp work is given by readings of the two inductive gauges applied to each joint (Fig.2). ()+()s1 (2) results for the group 1joints. a smooth sur face This where f is the longitudinal shear strength of the timber in material,f is the tensile strength in the tangential di- of the 5a) rection of timber and and are the mean shear stress This led to a marked local deformation in the timber in the area close to the dowel.The presence of an initial crack nd mean n2002.Te previ (see e.g.Schmid width of the timber specimen times a certain le th (r)in 199 the longitudinal direction.The length (xo),which corres- ponds to the mean stress length,was calculated according plastic.Studying Fig.4.it an also be to seen that the load-bearing capacity is close to maximum at same instant for 3 out of 4.indicating that the initial 会（祭）品 crack has a big influence on the joints. The fi nal failure mode for the group was a splitting failure mode +4 that e th / of the dowels seemed to be pr (3) seemed to be initiated by a crack propagating from the end where grain The ioints mpared and As can be seen in Fig.4,the load-displacement behav- =(i.e.mixed mode ratio) (5) ior is almost identical for all the specimens in group s we the pla with EL.ET,Gu and n dif mode 1 (opening mode)and mode 2(in-plane shear mode) no initial crack was seen fothe 2 ioints instea respectively andf and f the strength roperties of tim only the embedment failure of the wood was noticed,see as mentioned above(Gustafsson 2002).The definition Fig.5a.The type of final failure mode was the plug-shea of the length xo is such that the failure criterion gives the failure mode (Fig.5b).also in contrast to the group 1 d carrying cap a t an L imental observations presented above confim k Th n the nu merical work ted in Table 1.The elastic in Rodd (1973)and Siem (1).Equations given in EC5 operties were,as mentioned above,taken from Dinwoodie (2000.The ure nd ngth pmelerm do not take into account these ady ges.In Siem(1999), equations of doing so are presented. Springer

Holz Roh Werkst (2008) 66: 363–372 367 Different models may be used for fracture analyses of timber structures. In this work a mean stress method was used assuming stress distributing capabilities of the wood. The basic concept of this method, which is based on linear elastic fracture mechanics (LEFM), is to evaluate the mean stresses acting across an area rather than to evaluate the stresses in a single point. The size of this area is governed by the fracture properties of the material (Gustafsson 2002, Serrano and Gustafsson 2006). A stress criterion is then used to evaluate the capacity. The criterion used in this work is given by σ¯ ft 2 + τ¯ fv 2 ≤ 1 , (2) where fv is the longitudinal shear strength of the timber material, ft is the tensile strength in the tangential direction of timber and τ¯ and σ¯ are the mean shear stress and mean normal stress, respectively, in the tangential direction of timber acting across a possible fracture area (Gustafsson 2002). The fracture area was calculated as the width of the timber specimen times a certain length (x0) in the longitudinal direction. The length (x0), which corresponds to the mean stress length, was calculated according to x0 = 2 π EIGIC f 2 t EL ET GIIC GIC 2 1 4k4 × ⎧ ⎪⎨ ⎪⎩ 1+4k2 EL ET GIC GIIC −1 ⎫ ⎪⎬ ⎪⎭ 2 1+ k2 fv ft 2 , (3) where 1 EI = 1 EL EL 2ET EL ET + EL 2GLT −vTL EL ET (4) and k = τ¯ σ ( ¯ i.e. mixed mode ratio) , (5) with EL, ET, GLT,vTL, denoting the elastic properties of timber, GIC and GIIC the critical energy release rate for mode 1 (opening mode) and mode 2 (in-plane shear mode) respectively and fv and ft the strength properties of timber as mentioned above (Gustafsson 2002). The definition of the length x0 is such that the failure criterion gives the same load carrying capacity as that predicted using classical LEFM for a body of brittle material and containing a sharp crack. The material property parameters for timber used in the numerical work is presented in Table 1. The elastic properties were, as mentioned above, taken from Dinwoodie (2000). The fracture and strength parameters were similar to the ones used in Aicher et al. (2002) and Davenne et al. (1996). 4 Results and discussion 4.1 General observations The load-displacement curves for all joints tested, presented in Fig. 4, are based on the mean value of the displacement readings of the two inductive gauges applied to each joint (Fig. 2). The results for the group 1 joints, with a smooth surface dowel, show a ductile behavior. This behavior was caused by the embedment failure of the wood in combination with a crack developing in front of the dowel (Fig. 5a). This led to a marked local deformation in the timber in the area close to the dowel. The presence of an initial crack close to the dowel is an observation that is supported by previous research (see e.g. Schmid et al. 2002, Daudeville et al. 1999, Jorissen 1998). Interesting here was that this crack was initiated at the time when the elastic displacements turned to plastic. Studying Fig. 4, it can also be seen that the load-bearing capacity is close to maximum at same instant for 3 out of 4, indicating that the initial crack has a big influence on the joints. The final failure mode for the group 1 joints was a splitting failure mode (Fig. 5b). This caused the load to drop suddenly and completely. Note that in some joints the initial crack in front of the dowels seemed to be propagating in a stable manner until final failure, whereas in some cases the final failure seemed to be initiated by a crack propagating from the end grain. The results for group 2 joints, with a rough surface dowel were very different as compared to the group 1 joints. Most important differences were that the loadbearing capacities were much higher for the group 2 joints (Table 2) and the results were also much less scattered. As can be seen in Fig. 4, the load-displacement behavior is almost identical for all the specimens in group 2. Normally, both the elastic response as well as the plastic response is expected to considerably vary between different tests, as seen for the group 1 joints. Furthermore, no initial crack was seen for the group 2 joints, instead only the embedment failure of the wood was noticed, see Fig. 5a. The type of final failure mode was the plug-shear failure mode (Fig. 5b), also in contrast to the group 1 joints. The experimental observations presented above confirm the advantages of using a rough surface dowel as reported in Rodd (1973) and Siem (1999). Equations given in EC5 do not take into account these advantages. In Siem (1999), equations of doing so are presented. 13

368 Holz Roh Werkst (2008) 66: 363–372 4.2 Estimation of the coefficient of friction between the dowel and the surrounding timber The coefficients of friction (µ) between the dowel and the surrounding timber were estimated by comparing the straindistribution (strain perpendicular to the grain) along a certain path as measured by the ARAMIS-system with the distribution calculated by numerical simulations. The location of the path is shown in Fig. 6. The comparison of the results was made for the elastic part of the response (based on the global load-displacement response of the joints, Fig. 4), involving a total load of approximately 11 kN for the group 1 joints and 17 kN for the group 2 joints. Three joints from each group were studied. For comparing the results from the ARAMIS-system with those of the numerical analyses, the same load levels were simulated. Fig. 4 Load-displacement results for all joints tested Abb. 4 Kraft-Weg-Diagramme aller untersuchten Verbindungen Several numerical simulations were made where µ was varied. In Fig. 6, the results are presented together with the experimental results obtained from the ARAMIS-system. By a rough estimation of the results in Fig. 6, µ was decided to lie between 0 and 0.3 for the group 1 joints and between 0.3 and 0.5 for the group 2 joints. Using a coefficient of friction (µ) of 0.1 for the group 1 joints and 0.4 for the group 2 joints in the numerical simulations, a rather close correlation between the measured and the theoretical values can be found, see Figs. 6b, 7 and 8. In Rodd (1973) tests were made to determine the coefficient of friction between a dowel and the surrounding timber. Different dowel surface characteristics as well as different diameters of the dowel were studied. The results in that study, with dowels similar to the ones used in this study, gave that the coefficients of friction µ was 0.339 for the smooth dowels and 0.558 for the rough dowels. Interesting to note is that these values are close to the upper values as mentioned above for respective group of joints tested in this study. There are several reasons why the results presented in Rodd (1973) are not on exactly the same level as obtained in this study. One parameter can, Fig. 5 Typical experimental observations for each group: a picture taken before final failure, b final failure mode Abb. 5 Typische Versuchsbeobachtungen bei beiden Gruppen, a Aufnahme vor dem Bruch, b Bruchbild 13

Holz Roh Werkst (2008) 66: 363–372 369 Table 2 Results of load-bearing tests Tabelle 2 Ergebnisse der Tragfahigkeitsversuche ¨ Load-bearing Mean capacity capacity COV Group Joint (kN) (kN) (%) 1 1 12.7 13.3 7.6 2 13.4 3 12.3 4 14.6 2 1 19.0 19.1 0.4 2 19.1 3 19.0 4 19.2 Note: COV = Coefficient of variation for example, be that the estimates of µ in this study were made at a load level close to the ultimate capacity of the joints. This was not the case in Rodd (1973). It is most likely so that local plastic deformation occurs in the timber, parallel and perpendicular to the grain, close to the dowel at higher load levels that affect the contact situation. Thus, the coefficients of friction estimated in this study should be considered as fictitious parameters. In numerical simulations, when assuming linear-elastic behavior for the wood, the value obtained for this fictitious parameter is the one to use for the modeling of the frictional behavior. Fig. 6 Strain distributions perpendicular to the grain along the x-axis in the picture at the right: a numerical results where the coefficient of friction (µ) is varied, b experimental results from the ARAMIS-system for each group of joints together with a numerical result for the best fit curve Abb. 6 Dehnungsverteilungen rechtwinklig zur Faser entlang der x-Achse im Bild oben rechts: a numerische Ergebnisse bei unterschiedlichen Reibungskoeffizienten (µ), b mit dem ARAMIS-System ermittelte Versuchsergebnisse der Verbindungen mit einer rechnerisch ermittelten Ausgleichskurve A general observation can be made when studying Fig. 6. It shows that the level of tensile strains perpendicular to the grain is quite similar between the two groups in spite of the fact that the load level is much higher for the group 2 joints. The numerical results in Fig. 6 indicate that the strains are rather uniform for the group 2 joints, but for the group 1 joints instead a peak of tensile strain can be seen. This is the reason why the initial crack was seen only for the group 1 joints. 4.3 Fracture analyses In order to predict the load-bearing capacity of the joints, the same numerical model as used above was adopted. The coefficients of friction (µ) were set to 0.1 and 0.4 corresponding to the results presented above for the group 1 and group 2 joints, respectively. Shear stresses and normal stresses in the tangential direction of the timber were analyzed along different paths starting from the dowel towards the end grain surface. Equations 2–5 were used to find a critical path and estimate the load-bearing capacity. The results showed that this path was located about 2 mm off the dowel centerline for group 1 and about 7 mm for group 2 (Fig. 9). This gave a load-bearing capacity of 14 kN for group 1 and 19 kN for group 2, respectively. Comparing these results to 13

Holz Roh Werkst (2008) 66: 363–372 371 Fig. 9 a Load-bearing capacities, b the location of the mean stress length (x0), c shear stresses and normal stresses perpendicular to the grain along the x-axis in the picture up to the right. NB, x = 0 is located at the outside of the dowel Abb. 9 a Tragfahigkeiten ¨ b Lage der mittleren Spannungslange (x ¨ 0) c Schubspannungen und Normalspannungen rechtwinklig zur Faser entlang der x-Achse gemaß Abb. 6. ¨ x = 0 befindet sich auf der Au- ßenseite des Dubels ¨ 5 Conclusion The experimental results of the joints tested show that the load-bearing capacity increases when rough surface dowels are used as compared to using smooth surface dowels. In addition, the scatter of the load-displacement response was lower for the rough surface dowels. These experimental observations confirm the advantages of using rough surface dowels as reported in previous studies. The current version of the European timber code, EC5, does not explicitly take into account the effect of friction. In order to take the effect into account embedding tests have to be performed in order to be able to consider the benefits of using dowels with rough surface. The results obtained using the contact free measurement system and the numerical analyses both indicate the strain distribution around the dowel to be affected by the surface characteristic of the dowel. By comparing the experimental results with the numerical results, the coefficients of friction (µ) between the dowel and the surrounding timber were estimated. For joints with a smooth surface dowel µ was estimated to lie between 0 and 0.3. For joints with a rough surface dowel µ was estimated to lie between 0.3 and 0.5. Predicting the load-bearing capacity of the joints by fi- nite element analyses, including a failure criterion based on LEFM, showed good correlations to the experimental results. Acknowledgement This research was made possible thanks to fi- nancial support obtained from the Development Fund of the Swedish Construction Industry (SBUF), the Wood Science Program at Vaxj ¨ o¨ University (WDAT- Wood Design and Technology), the foundation CBBT (“Centre for building and living with wood”) and Skanska AB. The work was carried out at Vaxj ¨ o University. Special thanks to Mr. ¨ Arne Emilsson of Limtrateknik AB for the valuable comments on ¨ the manuscript, and for the fruitful discussions during planning and execution of the experiments. References ABAQUS Inc (2004) ABAQUS/Standard and ABAQUS/CAE Version 6.5, User Manuals Aicher S, Gustafsson PJ (ed), Haller P, Petersson H (2002) Fracture mechanics models for strength analysis of timber beams with a hole or a notch – a report of RILEM TC-133. Report TVSM- 7134, Lund University, Sweden Blaß HJ (2003) Joints with dowel-type fasteners. In: Thelandersson S, Larsen HJ (eds) Timber Engineering. Wiley & Sons, Chichester, England, pp 315–331 Daudeville L, Davenne L, Yasumura M (1999) Prediction of the load carrying capacity of bolted timber joints. Wood Sci Technol 33(1):15–29 Davenne L, Daudeville L, Yasumura M (1996) Failure of doweled joints loaded parallel to the grain: experiment and simulation. Proceedings of the CIB W18 Meeting, paper CIB-W18/29-7-8 Dinwoodie JM (2000) Timber: Its nature and behavior. Second edition, E & FN Spon, London Eurocode 5 (2004) Eurocode 5 – Design of timber structures. EN 1995-1-1:2004(E) European standard (1993) EN 383:1993 Timber structures – Test methods. Determination of embedding strength and foundation values for dowel type fasteners Gustafsson PJ (2002) Mean stress approach and initial crack approach. In: Aicher S, Gustafsson PJ (ed) Haller P and Petersson H Fracture mechanics models for strength analysis of timber beams with a hole or a notch – a report of RILEM TC-133. Report TVSM-7134, Lund University, Sweden Gustafsson PJ (2003) Fracture perpendicular to the grain – structural applications. In: Thelandersson S, Larsen HJ (eds) Timber Engineering. Wiley & Sons, Chichester, England, pp 103–130 Johansen KW (1949) Theory of timber connections. Int Assoc Bridge Struc Eng 9:249–262 Jorissen A (1998) Double shear timber connections with dowel type fasteners, Delft University Press, ISBN 90-407-1783-4 Rodd PD (1973) The analysis of timber joints made with circular dowel connectors. Ph D thesis, University of Sussex, England Schmid M, Blaß HJ, Frasson RPM (2002) Effect of distances, spacing and number of dowels in a row on the load carrying capacity of connections with dowels failing by splitting. Proceedings of the CIB W18 Meeting, paper CIB-W18/31-9-1, Kyoto, Japan 13