Dowel connections in laminated strand lumber

Dowel connections in laminated strand lumber Cranswick, Chad J. 1, M c Gregor, Stuart I. 2 ABSTRACT Laminated strand lumber (LSL) is a relatively new structural composite lumber. As such, very limited research has been undertaken regarding the performance of connections in LSL. Specifically, the failure mechanisms and load capacities, as well as load sharing characteristics of multiple dowel connections have yet to be developed. This cursory study investigated the configuration and group effects of mild steel dowel connections in laminated strand lumber. In addition, a comparison was made between the tensile resistances of these connections and those predicted by the current standard CAN/CSA086.1-M94 (Engineering Design in Wood) for similar connections in sawn lumber. The investigation was limited to 4-dowel connections in laminated strand lumber. Three 4-dowel configurations and one single dowel configuration (control case) were loaded in tension to determine their relative tensile behaviours and resistances. Significant tensile strength differences and group factors were found in the three 4-dowel connection configurations. Further, significant differences were found between the experimentally determined tensile strengths and those predicted by the current standard. INTRODUCTION Laminated strand lumber (LSL) is a structural composite lumber (SCL) first developed in the late 1960 s as an alternative to sawn lumber and other traditional structural wood products. Its manufacture permits a product with a high degree of consistency and an efficient use of low-grade fiber. Consequently, LSL is an environmentally positive and structurally valuable material, providing the structural engineering community with an alternative to traditional sawn lumber. Having only been available on the market for approximately 25 years, there is a limited amount of research-based information available on LSL, especially on connections. The connection rules in the Canadian Standard CAN/CSA- 086.1-M94 (Engineering Design in Wood) are based on sawn and glued-laminated lumber research. Tabulated design values are provided for a variety of wood species and some engineered wood products, with the exclusion of LSL. While a fair amount of testing has been done on single and multiple fastener joints, little research specific to LSL is available. Limited information, as given in a CCMC report (Trus Joist MacMillan, 1999), such as specified strengths and vertical load capacity, are available from the manufacturer. However, no information regarding the performance of LSL in a bolted connection is included. OBJECTIVES AND SCOPE The objectives of this study were to: investigate the configuration and group effects of mild steel dowel connections in laminated strand lumber; and to compare the tensile resistances of these connections with code predicted values for similar connections in sawn lumber. This study was limited to 4-dowel connections in laminated strand lumber. Three 4-dowel configurations and one single dowel configuration (control case) were tested. 1 Student, Dept. of Civil Engineering, University of British Columbia, Vancouver, Canada 2 Student, Dept. of Civil Engineering, University of British Columbia, Vancouver, Canada

EXPERIMENTAL PROCEDURES All experimental procedures were carried out in the Materials and Structures laboratories in the Civil and Mechanical Engineering Building at the University of British Columbia. Materials And Preparation The Timberstrand laminated strand lumber was obtained in four 44.5 mm x 301mm x 3658 mm boards. The specimens were cut to 44.5 mm x 149 mm x 1086 mm. Connection configurations of one dowel, four dowels in a line, four dowels in a square, and four dowels in a diamond were used. The dimensions are shown in Figure 1. These dowel patterns were chosen such that the edge distance effects would be minimal. The configurations were laid out on the plates of the loading grips and the holes were drilled to a diameter of 10.5 mm (1mm oversize as per article 10.4.1.2 of CAN/CSA 086.1-M94). Each configuration was tested on each of the boards in an effort to prevent any skewing of the results caused by differences in individual board strengths. Readily available 9.5 mm diameter steel dowels were used as fasteners in the connections. The steel dowels were cut from 4 separate 9.5 mm hot rolled mild steel rod lengths, with specified strength of 350 MPa. The dowels were cut to a length of 100 mm, allowing adequate clearance of the dowel ends outside the grip plates. Figure 1 - Dowel Configurations Steel Dowel Tension Tests In order to determine the tensile strength of the steel dowels, two 300 mm length samples were cut from the manufactured steel rod lengths. A reduced diameter of 6.4 mm was used in the specimen gauge length of 51 mm. The samples were mounted in the Saetec Testing Machine, which was equipped with a 20 kips load cell and was configured for the tension test. The samples were loaded in tension at a constant rate of displacement of 0.9 mm/min. Both load and displacement values were recorded, at one-second intervals, until an ultimate load was reached. This testing was performed in accordance with ASTM A.370 92 (Standard Testing Methods and Definitions for Mechanical Testing of Steel Products). Steel Dowel Bending Tests In order to determine the bending strength of the steel dowels, three samples were randomly selected from the previously prepared dowels. The samples were placed in the Saetec testing machine, which was configured for the bending test. The dowels were tested in three-point bending, over a span of 90mm, at a constant rate of displacement of 0.9 mm/min. Both load and displacement values were recorded, at one-second intervals, until an angle of deflection of 45 degrees from the horizontal was reached. A constant rate of displacement for the loading head was set to a rate of 0.9 mm/min. This testing was performed in accordance with ASTM A.370 92 (Standard Testing Methods and Definitions for Mechanical Testing of Steel Products). Density And Moisture Content Tests The dry density and moisture content determination of the LSL specimens were based on ASTM D.4442-84 (Direct Moisture Content Measurement of Wood and Wood-Base Materials). The dry densities of the LSL specimens were measured on two occasions: before the day of testing and on the day of testing. These calculated dry densities were averaged to represent the dry density for each specimen. Samples were

obtained on both occasions by cutting a block from each of the four LSL boards. The specimens were placed in a convection oven at 102 o C for a minimum of 22 hours. LSL Tension Tests The LSL specimens were tested in tension using a custom testing machine, which included a hydraulic actuator mounted on an I-Beam frame. To avoid inducing torsional stresses, a swivel was used on the upper end of the apparatus. Further, pin connections were used on the upper and lower ends to simulate the behavior of a pinned connection. The specimen was secured in the lower loading grip with six 9.5mm bolts to ensure that failure occurred at the upper connection. The steel dowels were inserted through the holes of the grips and through the LSL specimen until they extended approximately 15 mm beyond both grip plates. Two LVDTs were mounted on the LSL specimens at opposing corners of the connection such that measurement of the displacement of the specimen relative to the plates could be recorded. This apparatus is presented in Figure 2. A 50 kips load cell was used to measure the load applied to the connection. A constant rate of displacement of 0.9 mm/min was set. The test was run until connection failure, during which time the load vs. displacement data was recorded using a computerized data acquisition system. The testing of the LSL was performed in accordance with ASTM D.1761-88 (Standard Test Methods for Mechanical Fasteners in Wood). RESULTS Figure 2 - Apparatus The following are the summarized results and analysis. The raw data and sample calculations can be found in Appendix A. Steel Dowel Tension Tests The yield stresses of the steel dowels, obtained from the stress versus strain data as generated by the tension tests, were as follows: Table 1 - Steel Dowels in Tension Specimen 1 Specimen 2 Average Yield Stress (MPa) 358 346 352 Steel Dowel Bending Tests The yield stresses of the steel dowels, obtained from the load versus displacement data as generated by the bending tests, were as follows: Table 2 - Steel Dowels in Bending Specimen 1 Specimen 2 Specimen 3 Average Yield Stress (MPa) 605 591 589 595 Density And Moisture Content Tests The results of the density and moisture content tests are as follows: Table 3 - Density and Moisture Content Board 1 Board 2 Board 3 Board 4 Dry Density (kg/m 3 ) 705 590 632 661 Moisture Content 8.9% 8.2% 10.4% 8.9% LSL Tension Tests The load readings for each of the five specimens of each configuration were averaged at 0.5 mm displacement intervals. These values were then plotted as follows:

120 100 80 Single Dowel 4 Dowel Line 4 Dowel Square 4 Dowel Diamond Load (kn) 60 40 20 0 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 Displacement (mm) Graph 1 - Load vs Displacement Since the serviceability limit of a connection in a typical building frame occurs at a sway corresponding to a brace displacement of approximately 5mm, the load values at this point were extracted as the design strength. The ultimate strengths, corresponding to the maximum load obtained were also extracted. These average strengths and relevant statistics, are shown in the following tables: Table 4-5mm Design Strengths Mean 5mm Strength (kn) 21.68 74.35 81.74 85.78 5th Percentile (kn) 16.5 55.1 70.4 71.6 Standard Deviation (kn) 3.16 11.69 6.92 8.59 Coefficient of Variation 0.146 0.157 0.085 0.100 Table 5 - Ultimate Strengths Mean Ultimate Strength (kn) 25.46 88.39 91.95 100.47 5th Percentile (kn) 15.1 71.7 79.5 82.7 Standard Deviation (kn) 6.27 10.14 7.59 10.78 Coefficient of Variation 0.246 0.115 0.083 0.107 Each 4-dowel configuration average load versus displacement curve was divided by the number of dowels. This allowed a comparison with the single dowel configuration. Table 6 - Load Per Dowel Mean 5mm Strength (kn) 21.68 18.59 20.44 21.44 5th Percentile (kn) 16.48 13.78 17.59 17.91 Standard Deviation (kn) 3.16 2.92 1.73 2.15 Coefficient of Variation 0.15 0.16 0.085 0.10

The loads for each of the configurations were normalized with respect to the single dowel control case, by dividing each by the single dowel load. This resulted in a suggested group factor and the following graph was obtained: 1.2 Group Factor (Load/dowel per Single dowel Load) 1 0.8 0.6 0.4 0.2 Single Dowel 4 Dowel Line 4 Dowel Square 4 Dowel Diamond 0 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 Displacement (mm) Graph 2 - Normailzed Load vs Displacement The group factors at displacements of 1mm and 5mm are estimated as follows: Table 7 - Group Factors Group Factor at 1 mm 1.00 0.48 0.71 0.72 Group Factor at 5 mm 1.00 0.86 0.94 0.99 DISCUSSION OF CONNECTION BEHAVIOR The configuration and group effects are discussed as follows: Tensile Behavior The tensile behavior of a dowel connection in LSL is characterized by Graph 1. The loading curves approximate linear behavior prior to displacements of 1mm. During this initial displacement, the steel dowels are taking up the tolerance of the 1mm oversized dowel holes. It is generally accepted (Turnbull, 1995) that a typical multiple fastener joint should perform elastically up to an initial displacement, which corresponds roughly to the oversize of the bolt holes. Up to 1mm displacement only one dowel is likely to have been correctly aligned in its hole, and therefore would be the only dowel bearing load. Once this initial displacement is reached, the remaining dowels become aligned and begin to bear load. After this point, the connection increases in stiffness and ceases to display linear behavior. Between displacements of 1mm and 3mm, the curves have a noticeable dip. This decrease in slope is a result of the steel dowels beginning to yield in bending. As the dowels become essentially fixed ended (i.e. restrained by the side plates) the slopes of the loading curves begin to increase. At displacements beyond 5mm the loading curves become increasingly horizontal as they approach an ultimate load. These ultimate loads are reached at displacements between 6mm and 7mm as tabulated in Table 5. After ultimate loads were reached, specimens began to fail. After failure of the first specimen, the average curve becomes invalid.

Failure Modes Generally, the connections tested exhibited ductile behavior, as indicated by the relatively large displacements prior to failure. This high ductility acts as an energy dissipation mechanism in the connection. Two major factors accounting for this energy dissipation are addressed. One energy dissipation factor is the deformation of the dowels. This deformation is due to the plasticity in the tseel, which absorbs energy as plastic work is done on the dowels. The ductile behavior of mild steel allows further displacement than in a similar connection with rigid fasteners. Typically, high strength steel bolts are used in similar connections, thus eliminating this energy absorption mechanism, since the bolts remain undeformed after connection failure. Had high strength steel dowels been used, smaller displacements for similar loads would be expected. The second factor is the extensive crushing of the wood fibers, resulting in elongation of the dowel holes. The crushing is due to the physical properties of LSL, which is composed of low-grade wood fibers oriented mainly in the longitudinal direction, with some fibers oriented transversely. The fiber orientation provides a consistent matrix inhibiting crack propagation and thus promoting ductile behavior. Successive crushing of the transversely oriented fibers allows the dowels to plow though the lumber, without (immediate) significant cracking. As the energy is absorbed as wood crushing progresses, the tensile stress perpendicular to the applied load (at the dowel hole) eventually exceeds the tensile capacity of the specimen. As this occurs failure is induced in the form of cracking. This cracking generally occurs at the ultimate load. The single dowel configuration was the most ductile of the configurations tested. There was limited brittle behavior as the specimens primarily experienced extensive crushing of the wood fibers. As a result, there was no pronounced failure point; ductile behavior continued up to displacements as high as 30 mm. Each of the 4-dowel configurations displayed significant dowel hole elongation prior to cracking. This cracking pattern was dependent upon the configuration of the dowels in the connection. The typical cracking pattern in the 4-dowel line configuration acted along the dowel line parallel to the applied load. There was no significant cracking perpendicular to the load, indicating that the edge distance had no effect on the failure mode. The cracking patterns in the 4-dowel square configuration varied. Some specimens cracked parallel to the load over the end distance, while others cracked in a combination of this and perpendicular cracking over the edge distance. The cracking pattern of the 4-dowel diamond also varied. The cracks originated at the furthest dowel from the end of the specimen, and extended towards the center two dowels. As in the square pattern, some of the cracks then extended to the edge, while some extended to the end. Configuration Effects The tensile resistance of a dowel connection in LSL is dependent on the configuration of the dowels. As illustrated in Graph 1, differences in strength were apparent in the three 4-dowel connections. The 4-dowel diamond configuration had the highest ultimate and 5mm strengths. Conversely, the 4-dowel line configuration had the lowest ultimate and 5mm strengths. The resistances of the 4-dowel connections were divided by the number of dowels in order to obtain a per-dowel resistance value. This allowed a comparison with respect to the single dowel resistance value. As expected, none of the 4 dowel connections exceeded the resistance of the single dowel case within a 5mm serviceability limit. This suggests the existence a group factor. To further demonstrate this group factor, the 4-dowel connection resistances per dowel were divided by the single dowel connection resistance and plotted in Graph 2. If there were no group effect, all of the lines would lie horizontally at value 1, as does the single dowel case. As the displacement increases, the group factor approaches 1 due to the load becoming equally distributed amongst the dowels. At a displacement of 1mm, the group factor ranges from approximately 0.50 (4-dowel line) to 0.70 (both 4-dowel square and diamond). This range becomes approximately 0.85 to 0.98 at a displacement of 5mm. Beyond a 5mm serviceability displacement, the group factor is not significant. These group factors are tabulated in Table 7. Group effects are a direct result of the load sharing characteristics of each configuration. At a 5mm displacement, the load-per-dowel strength in the diamond configuration is essentially equal to the single dowel strength. This indicates that the 4-dowel diamond configuration strength is equal to four times the single dowel configuration strength. This is due to the fact that the diamond is the most spread-out configuration, thus slightly isolating the dowels and reducing the

interaction between the loaded areas of each dowel (load sharing). Conversely, the four-dowel line is the most uniform configuration, thus maximizing the interaction between the loaded areas of each dowel. Statistical Analysis The relatively small sample size notwithstanding, the five 5mm resistance values for each configuration were fitted with a normal distribution such that a 5th percentile resistance could be determined for each. To estimate these 5 th percentile resistance values, the formula R 05 = R avg 1.645s was used. The 4-dowel diamond configuration had the highest 5th percentile resistance, and the 4-dowel line the lowest. The load per dowel 5 mm resistances were also fitted with a normal distribution, in order to determine the 5 th percentile values. Experimental Limitations The dowel holes in the grip plates sustained minor bearing failure, allowing the dowels to rotate further in the plates before behaving as fixed ended. This bearing failure became noticeable only after fifteen of the tests were performed. As such this effect was mitigated by performing tests on the different configurations randomly throughout the testing. Applying a normal distribution to a sample population of such limited size cannot be done with significant confidence. Typically 25 to 30 samples are the minimum population sizes to which a normal distribution should be fitted. By testing only five samples per configuration, relatively inaccurate values of both mean and standard deviation may have been obtained. As all of the subsequent statistical calculations were based upon the initial assumption of normally distributed data, there is potential for significant error to be introduced. EXISTING CODE PREDICTIONS Currently the CAN/CSA 086.1 (Engineering Design in Wood) combined with supplementary information in the Canadian Wood Council Wood Design Manual (the Code, hereafter) provides the standard for wood engineering. As of the 1995 edition, it contains no information on engineering connections in laminated strand lumber. One of this study s objectives involves the calculation of the tensile resistances of similar connection configurations as predicted by the Code. These calculated resistances are then compared to the 5 th percentile values of the 5mm strengths, experimentally determined, for LSL. With reference to Part 10 (Fastenings) in the Code, the resistance of a bolted connection parallel to the grain, is given as follows: Resistance, Pr = F P u n s n F J F Where: F = performance factor - 0.7; n s = number of shear planes; n F = number of fasteners; J F = number of rows, number of bolts in a row, and end distance factor; P u = p u (K D K SF K T ); K D = duration of load factor; K SF = service condition factor; K T = treatment factor; p u = unit resistance per shear plane. Predicted vs. Experimental The following table compares the Code predicted average lateral strength resistance with the 5 th percentile of the 5mm serviceability strengths as determined experimentally. Since limits states design, based on probability, often uses 5 th percentile strength values, the 5 th percentile of the 5mm design strengths were used. Also, the diamond configuration is not predictable by the Code; therefore, it is assumed to be equal to the square configuration as it exhibited similar resistances. Table 8 - Predicted vs Experimental Code Predicted Strength (kn) 7.5 18.7 18.4 18.4 5th Percentile 5mm Strength ( 16.5 55.1 70.4 71.6 % Gain in Strength 119% 194% 282% 289% Discussion of Code Predictions The experimental values of strength, for each of the configurations, are significantly higher than the values predicted by the Code. Percent gains of 119% up to 289% suggest that the Code predictions provide a very conservative estimate of the strength of LSL. The connections, however, differed slightly in nature and thus required engineering judgment to obtain comparable estimates.

As the Code does not contain any information on the use of dowel connections, a bolted connection was assumed. The use of bolts of similar properties would not have significantly affected the results. However, since the bolts considered in the Code are of high strength steel, and not mild steel, they will tend to plow though the wood without significant bending. Therefore, a significant amount of ductility would be lost. The Code predicted this plowing of the bolt as the governing (minimum) failure type in the calculation of the unit lateral strength resistance. This failure type depends on the embedding strength of the wood, which then depends on the density of the wood. The Code prediction of embedding strength of wood may not be accurate for LSL, for it may depend on the composition as well as the density. Another result of the assumption of a bolted connection is the fact that the Code uses the tensile strength of the bolts in the resistance calculations. As such, the tensile yield strength of the dowels was determined experimentally for use in these calculations (these values are listed in Table 1). Since, however, the dowels used were of mild steel, they failed mainly in bending. This was the reason for determining the flexural yield strength (these values are listed in Table 2). Since the flexural strength of the dowels was significantly higher than the tensile strength, the Code predicted values would have been higher, had the proper yield strength been used. Another notable difference in the calculated versus experimental values is the end distance consideration. It has been determined that for LSL, end distance effects are eliminated at distances greater than five times the bolt diameter. The experimental strengths are based on an end distance of 5d, whereas the calculated are based on 7d. This is because the Code does not permit end distances less than 7d (for sawn lumber). Using 5d in LSL may, however, be equivalent to using 7d in sawn lumber, which therefore eliminates the end distance factor. In consideration of these assumptions, the behavior of LSL in 4-dowel connections is much different than the behavior of sawn lumber in 4-bolt connections. The predicted values suggest that the 4-dowel line is roughly equivalent to the other 4-dowel configurations; however, the experimental values show that there are considerable differences in strength. The inconsistency between the predicted and the experimental values, assumptions aside, is mainly due to the differences between the physical properties of sawn lumber and LSL. CONCLUSIONS Significant tensile strength differences were found in the three 4-dowel connection configurations. For both 5mm serviceability and ultimate strengths the 4-dowel diamond configuration displayed the highest resistance values while the 4-dowel line configuration displayed the lowest. Significant group factors were found in the three 4-dowel connection configurations. At displacements of 1mm and 5mm the 4-dowel diamond displayed the highest group factors (and thus most efficient load sharing) while the 4-dowel line configuration displayed the lowest. Significant differences were found when the experimentally determined strengths were compared to those predicted by CAN/CSA 086.1 (Engineering Design in Wood). REFERENCES Canadian Wood Council, Wood Design Manual 1995. Canadian Wood Council: Ottawa, ON. Nelson, S.A. 1994. Design considerations for structural composite lumber and prefabricated wood I-joists. Wood Design Focus, 5(1), 8-13. Trus Joist MacMillan. 1999. CCMC Evaluation Report: Timberstrand LSL. CCMC 12637-R. National Research Council of Canada, Institute for Research in Construction. Turnbull, J.E. 1990. CSA Commentary, Wood Design Manual, Canadian Standards Association, Section 14-9, Fasteners. Cranswick, Farnworth, Hanson, McGregor, Pike, Seeton, and Soares, 2000. Dowel Connections in Laminated Strand Lumber. The University of British Columbia, Department Civil Engineering undergraduate project. The authors wish to acknowledge their fellow students who were instrumental in both testing and data reduction.