- Email: [email protected]
Performance of toe-nail connections under realistic wind loading
Engineering Structures 33 (2011) 69–76
Contents lists available at ScienceDirect
Engineering Structures journal homepage: www.elsevier.com/locate/engstruct
Performance of toe-nail connections under realistic wind loading Murray J. Morrison, Gregory A. Kopp ∗ Boundary Layer Wind Tunnel Laboratory, Faculty of Engineering, University of Western Ontario, London, ON, N6A 5B9, Canada
article
info
Article history: Received 15 July 2010 Received in revised form 2 September 2010 Accepted 16 September 2010 Available online 29 October 2010 Keywords: Wind loads Disaster mitigation Low-rise buildings Nails Wood-frame construction
abstract Using recently developed pressure loading actuators (PLA), ramp and realistic fluctuating wind loads are applied to toe-nail connections which are typically used in wood-frame residential construction. The failure capacity from the ramp and fluctuating wind load tests are found to be similar, and are comparable to capacities reported in the literature. However, under realistic wind loading, the toe-nail connections are found to fail in increments with the majority of the damage to the connection occurring intermittently at the peak pressures so that it takes many peaks for a connection to fail. In addition, the effects of construction defects, in this case missing nails, were also examined in order to determine the reduction in capacity. Considering the wind loads on a typical house, estimates for failure wind speeds were obtained, assuming a factor of safety, and compared to ASCE 7-05 wind regions. © 2010 Elsevier Ltd. All rights reserved.
1. Introduction Roof failures are commonly observed in extreme wind events, including roof-to-wall connections (RTWC) and sheathing in residential wood-frame housing. In addition to life safety issues, failure of a roof or its components can cause significant water ingress, with subsequent damage to the building contents, substantially increasing the insured losses [1]. Furthermore, with all or part of the roof missing, the walls are more susceptible to collapse which can become a significant life safety issue. Toe-nails are the most common type of RTWC in North America. While it is now common to use hurricane straps in hurricane prone regions such as Florida and the Gulf Coast of North America, in new construction, there are still a large number of existing structures with toe-nail connections as the primary hold down for the roof structure. Moreover, in non-hurricane regions toe-nails are still used in new construction as the primary roof-to-wall connection. These regions are susceptible to extreme wind events, such as tornadoes and downbursts, which are capable of causing complete roof failures, a recent example of which is discussed in [2]. The capacity of toe-nail connections to uplift loads has been the subject of several studies, including those by [3–6]. These tests applied load at a constant displacement rate (typically 2.54–6.35 mm/min) and measured the force required to keep the connection moving at the set rate. These tests are nominally static and quite different from the highly fluctuating loads generated by
∗
Corresponding author. Tel.: +1 519 661 3338; fax: +1 519 661 3339. E-mail addresses: [email protected], [email protected] (G.A. Kopp).
0141-0296/$ – see front matter © 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.engstruct.2010.09.019
real winds. Experiments conducted by [6] on in situ connections did apply a form of cyclical loading, although a maximum of only three loading cycles were used. The cycles were applied at a low displacement rate of 2.54 mm/min and the end of a cycle was based on a displacement threshold. The loading from these tests was also significantly different from that induced by real wind loads where there are a large number of cycles and the loads can double or increase more than that in less than a second and decrease just as quickly. The mean maximum withdrawal capacity from these studies is in the range from 1130 N to 2840 N, depending on the type, number of nails, age of the connection, type of wood, and the wood moisture content. To date there has been no study to document the effects, if any, that cyclical or realistic wind loads have on the withdrawal performance of toe-nailed RTWC. Hysteretic behavior and reduction of the failure capacity when subjected to cyclical shear loads, have been tested and implemented in finite element models of entire structures, e.g., [7], primarily for studies involving earthquake loading. The current study examines the withdrawal behavior of toe-nail connections under realistic wind uplift loads and compares the response to that found from static testing. 2. Experimental setup Toe-nailed RTWC are tested using a load control approach with the test rig shown in Fig. 1. The loads are applied to the specimen by controlling the pressure inside of the blue airbag, which is then mechanically attached to the toe-nail specimen, as shown in inset A of Fig. 1. The toe-nail connections consist of two 0.61 m long 2 × 4 segments, representing a typical portion of a top-plate for woodframed stud walls, mounted at either end to load cells to measure
70
M.J. Morrison, G.A. Kopp / Engineering Structures 33 (2011) 69–76
Fig. 1. Photograph of the nail test rig. Loads are controlled by altering the pressure in the air bag, while load cells and a displacement transducer measure the response of the toe-nail connection.
Fig. 2. Pressure tap layout for the wind tunnel model of the full-scale test house. Dimensions are provided in full scale.
the reaction at the connection. A 0.3 m long 2 × 6, representing a typical rafter or roof truss section is connected to the bottom of the airbag; the rafter is then toe-nailed to the top-plate using 3 12-d twisted shank nails using a pneumatic nail gun. In this paper the term ‘d-nails’ will refer to the side of the rafter with two nails, while the ‘s-nails’ will refer to the side of the rafter with just a single nail. For the current experimental setup the loads are applied normal to the top-plate (upward in Fig. 1), meaning that only withdrawal loads are applied to the toe-nail connection. It is noted that each nail will not be subjected to pure withdrawal loads due to the angle at which the nails are driven; however, this description of the loading is consistent with the National Design Specification for Wood Construction (NDS) [8]. In addition, flat roof construction is not typical in most types of residential construction; common roof slopes are typically between 3 on 12 and 6 on 12, although this range can vary substantially by region. As a result a portion of the uplift wind loads applied to the roof of the house will not apply a perfect withdrawal load to the toe-nail connections. In the case of a 4 on 12 roof slope this results in approximately 5% of the applied load to the roof being applied in shear to the connection. The amount of load applied in shear due to the angle of the roof is likely much smaller than the shear loads that would be applied to the toe-nail connections due to the loads applied to the wall. The purpose of the current investigation is to examine the withdrawal capacity of only toe-nail connections under realistic uplifting wind loads; as such, shear loads have been explicitly ignored and is consistent with how loads have been applied in previous studies. A displacement transducer is mounted to the rafter and measures the displacements of the rafter relative to the flat concrete floor, as shown in Fig. 1. Since the top-plate is kept stationary throughout the experiment, the measured displacement is equivalent to the displacement of the rafter relative to the topplate. The pressure in the airbag is controlled using a pressure loading actuator (PLA) which is able to accurately replicate temporally varying pressures from an input pressure trace. Further details regarding the PLAs, and their capabilities, can be found in [9]. As discussed above, the capacity of toe-nail connections varies substantially in the literature based on the type of nails, grade of lumber, moisture content and age of the connection. The goal of the current investigation is to determine the behavior and capacity of toe-nail connections under realistic wind loading. To achieve this objective, the static capacity of the connections used in the current study must first be determined using a loading method similar
to that of previous studies in order to obtain a baseline static withdrawal capacity. To obtain the static capacity, and to assess the effect of loading rate on the response of the connections, ramp loads are applied to the specimen using three different loading rates, viz., 1, 8, and 32 kN/min, resulting in expected failure times in the range from 5 to 200 s. These ramp loading rates were selected so that the slowest rate would result in a failure time similar to that of previous experiments where a constant displacement rate was used, while the highest loading rate would approximate a loading rate closer to that found in actual wind loads. A total of 21 toenail specimens were tested at each ramp loading rate in order to account for variability in wood properties and construction. In typical construction, there can be significant variability in the quality of a toe-nail connection due to construction errors and defects. To attempt to quantify the effect of these errors in a controlled way, experiments were conducted using a ramp loading rate of 8 kN/min using toe-nail connections that were missing a single nail, which results in two types of defect cases. Defect #1 will refer to tests where the‘s-nail’ is missing, while defect #2, will refer to tests where one of the ‘d-nails’ is missing. Since actual wind loads vary significantly based not only on the building geometry, but also on the location on the particular building as well, there are nearly infinite possible wind loading time series that could be applied. The realistic, fluctuating loading for the current experiments were obtained from wind tunnel experiments which measured the roof pressures on the ‘Three Little Pigs’ test house, described in [9], at a scale of 1:50 in Boundary Layer Wind Tunnel II at the University of Western Ontario (UWO). This house has full-scale plan dimensions of 9.0 m by 8.9 m, an eave height of 8.0 m and a gable roof slope of 4:12. The dimensions, along with the tap layout, can be found in Fig. 2. The flow simulation approximates a typical open country atmospheric boundary layer at a scale of 1:50 with an aerodynamic roughness length, zo , of approximately 0.03 m (equivalent full scale). The terrain simulation used in the current study is identical to that of [10,11], where a detailed discussion of the flow simulation and modeling approaches can be found. Pressure taps were connected to PSI pressure transducers using the tubing system presented in [12] and have a frequency response which is flat up to 200 Hz. The pressures were sampled nearly simultaneously (maximum lag of 0.0025 s) for a total of 180 s at a frequency of 400 Hz. In total, 18 wind angles were tested ranging from 0° to 90° at a mean roof height wind speed of 9.6 m/s and Reynolds number, Re = VH /ν = 1.0 × 105 .
M.J. Morrison, G.A. Kopp / Engineering Structures 33 (2011) 69–76
71
Table 1 Summary of the real wind load trace applied to the toe-nail connection. Scaling wind speed (m/s)
Peak force (kN)
Trace duration (s)
20 25 30 35
1.35 2.45 3.47 4.91
900 720 600 514
Fig. 3. Truss layout for the ‘Three Little Pigs’ test house at UWO. The spacing between trusses is nominally 0.61 m on center and each toe-nail connection is labeled in the figure.
Fig. 3 presents the truss layout for the full-scale house. Using this layout of the trusses, the force coefficient, Cf , time series at each RTWC of the house can be estimated using: Cf (t ) =
∫
Cp (x, y, z , t ) · I (x, y, z )
A
da A
(1)
where I (x, y, z ) is the structural influence function, Cp is the pressure coefficient over a small area da, and A is the area where the influence function is non-zero. In the current study, a geometric tributary area approach is used to evaluate I (x, y, z ), as it is not generally known for wood-frame houses. The RTWC with the largest peak magnitude of Cf in the time series was then chosen for these experiments, which corresponded to RTWC ‘S3’. Rather than using the entire 1 h (equivalent full scale) time history, a representative portion was selected for the current experiments. The full-scale force, F , was obtained using: F (t ) = 0.5ρ U 2 Cf (t )A − W
(2)
where U is the hourly mean wind speed at mean roof height, ρ is the density of air and W is the portion of the weight of the roof associated with RTWC ‘S3’. The current approach to apply realistic loads is similar to that used by [13]. A wind speed, U, of 20 m/s was chosen for the first test for each connection and the resulting force time series, using Eq. (2), was applied to the connection. If the connection did not fail, the scaling wind speed was increased by 5 m/s and the new force time history was re-applied to the test specimen. This process was repeated until failure occurred. It is noted for the current experiments that the loads at each scaling wind speed were applied sequentially with no pause for changes in the scaling wind speed. In addition, since the weight of the roof, W , has been removed in Eq. (2), the wind loads at different wind speeds will no longer scale with the square of the velocity for each test. This choice does not affect any of the conclusions, but aids in the interpretation of the results. Fig. 4 provides an example of the entire force time series that is applied to the specimen, with the locations where the scaling wind speed changes are also indicated. Table 1 provides details of the force time history at each scaling wind speed. It is also noted, because of the scaling laws, i.e.,
νT L
= model-scale
νT L
Fig. 4. Force time series of the realistic wind loads for all scaling wind speeds applied.
(3) full-scale
that, as the scaling wind speed increases, the length of the test decreases. It was decided that the same segment of wind tunnel data would be used so that an identical number of peaks occur for each scaling wind speed, but the actual duration is shortened. This can also be inferred from Fig. 4. In total 25 specimens were tested using the realistic wind loading time series presented in Fig. 4.
Fig. 5. Load vs. displacement relationship for a ramp rate of 8 kN/min.
3. Results 3.1. Ramp loading Fig. 5 shows a typical load–displacement relationship for one of the 8 kN/min ramp tests. The failure capacity of the connection is defined as the maximum measured reaction, as indicated in the figure. After this point, the measured reaction drops as the nails pull out. Since the current study uses a load control approach, the failure proceeds in an uncontrolled manner; therefore, the data beyond the failure capacity are not used in the analysis. Similar to [6], the log-normal distribution was found to best fit the maximum capacities when considering normal, Weibull or Gumbel distributions. The cumulative distribution for all three ramp loading rates is shown in Fig. 6. The distributions for the ramp rates of 8 and 32 kN/min are similar, while the 1 kN/min test has a lower failure capacity. The mean failure capacity for each ramp rate is nearly the same ranging from 2.7 kN at 1 kN/min to 2.9 at 32 kN/min. The difference in mean failure capacity based on loading rate is insignificant as compared to the range of failure capacities observed, which ranged from 1.2 to 4.7 kN from all
72
M.J. Morrison, G.A. Kopp / Engineering Structures 33 (2011) 69–76 Table 2 Summary of the types of failures from the ramp loading tests and there average failure capacity. Failure type
Mean failure capacity (kN)
Standard deviation (kN)
% of failures
All nails split d-nails split s-nail split All nails pull out
3.3 3.2 3.2 2.6
0.58 0.41 0.87 0.54
8 21 6 65
Table 3 Summary of failure capacity and failure modes of connections with defects for ramp tests.
Fig. 6. Cumulative distribution of the mean failure capacity for all 3 ramp loading rates and the mean maximum applied load for the realistic wind loading tests.
Fig. 7. Probability distribution for the same data as presented in Fig. 6. The inset in the top right of the figure presents the probabilities on a logarithmic scale.
ramp rates. In order to determine if data from each loading rate follows the same distribution, a null hypothesis is formed that each loading rate can be represented by the same distribution. Using a Kolmogorov–Smirnov test, the null hypothesis cannot be rejected between the 3 ramp loading rates using a 95% confidence interval. This indicates that, with the current data, the failure capacities for each loading rate can be represented by the same distribution. Moreover, using the same confidence interval and an analysis of the variance (ANOVA), the mean failure capacity from each loading rate are not significantly different. This means that the failure capacity of the toe-nail connections is independent of the loading rate within this range, which encompasses nearly all loading rates expected to occur under actual wind loading. This finding is consistent with that of [14], which found no loading rate dependence in the capacity of nailed connections (although toe-nails were not examined). Fig. 7 presents the probability distribution for the same ramp rate data used for Fig. 6, and shows that distributions become broader with increasing loading rate suggesting that there is an increased variability of the failure capacity. In addition, the inset of Fig. 7 shows that the distributions vary significantly in the tail regions of the distribution, which is likely the result of an imperfect fit of the log-normal distribution to the experimental data. Additional samples at the current loading rates along with investigating more loading rates between 1 and 8 kN/min would be needed to determine if the reduced variability and lower failure capacity observed for 1 kN/min is a true loading rate effect. This point is not investigated further herein. During the ramp loading experiments the nails were observed to fail in two different ways: either ‘‘pull-out’’, where the nails pulled out of the top-plate; or ‘‘splitting’’, where the wood of the
No defect Defect #1 Defect #2
Mean failure capacity (kN)
Standard deviation (kN)
#split/#pull outs
2.8 1.9 2.2
0.6 0.46 0.48
22/41 11/5 0/16
rafter splits with the nails remaining attached to the top-plate. As a result of these two types of failures, four different failure modes were observed, all nails split, d-nails split, the s-nail split, or all nails pull out. Fig. 8(a) shows an example of a pull-out failure, while Fig. 8(b) presents an s-nail split failure. The ramp test results sorted by failure type are shown in Table 2. Since the loading rate was found to have a minimal effect on the capacity of the toe-nail connections, all ramp rate test results have been used. These results show that the predominant failure mode is when all nails pull out together. This failure mode has a lower failure capacity than when splitting failures occur, which is consistent with the results found in [6]. However, there does not seem to be any significant difference in capacity between types of splitting failures, although the d-nails splits were much more common than the other two types of splitting failures indicating that the proximity of the two nails on one side may cause that side to be more prone to splitting. 3.2. Effect of missing nails In total, 16 tests were conducted for each defect condition discussed above. The mean and standard deviation of the failure capacity, along with the type of failure, are given in Table 3. For both defect cases, the failure capacity per nail is higher than that of the no defect case. The mean failure capacity for defect #1 is lower than that of defect #2. The predominant failure mode of defect #1 is splitting, while only pull-out failures are observed for defect #2. This is surprising, since, for the no defect case, splitting failures are generally associated with a higher mean failure capacity, although it is likely the result of the extreme asymmetry of the connection in the case of defect #1. 3.3. Fluctuating wind loading A total of 25 specimens were tested using the realistic, fluctuating wind loading trace shown in Fig. 4. A displacement time series for a typical specimen is shown in Fig. 9. Rather than the connection gradually being withdrawn from the top-plate as in the ramp loading experiments, partial permanent withdrawal of the nails accumulates over handful of peak loads. The peak loads that cause this incremental withdrawal of the nails will be referred to as ‘‘damaging peaks’’ in the following discussion. For the displacement time history shown in Fig. 9, the damaging peaks are indicated in the figure. Fig. 10 shows the load versus displacement relationship for the same test as shown in Fig. 9. Over the first portion of the test, prior to any damaging peaks, the load–displacement behavior is nearly linear; however, following
M.J. Morrison, G.A. Kopp / Engineering Structures 33 (2011) 69–76
a
73
b
Fig. 8. (a) Photograph of a toe-nail failure where all nails have pulled out of the toe-plate. (b) Photograph of an ‘s-split’ failure of the toe-nail connection.
to the maximum load applied was observed to range from 87% to 100%. The overall mean maximum load applied during the realistic fluctuating wind loading was found to be the same as the ramp loading rates, being 2.8 kN. Pull-outs are the most common mode of failure, representing 76% of the cases. This is slightly higher than that of the ramp loading tests. However, unlike the ramp loading, there is no observed change in the mean maximum applied load between pull-out and splitting failures under the realistic, fluctuating wind load. 4. Discussion of temporal effects
Fig. 9. Displacement time series of a typical realistic fluctuating wind loading trace applied to a toe-nail connection. The damaging peaks are indicated in the figure.
While the failure of toe-nail connections due to ramp loads was a gradual withdrawal of nails, the nails subjected to realistic fluctuating loading are partially withdrawn from the top-plate during the peak pressures in the applied trace. The ramp loading tests, as well as the dynamic experiments, have shown that there is no significant dependence on the loading rate in the response of these connections. As previously discussed, the highest ramp loading rate is comparable to loading rates applied under realistic wind loading. However, it should be noted that under realistic wind loads as the connection becomes stronger and requires larger scaling wind speeds to cause failure, the higher the loading rate will become as indicated from Eq. (3). Despite the difference in the observed behavior between the ramp and realistic wind loading the average failure capacity is remarkably similar. The following discussion will examine the loading time history in detail and discuss the peak loads that cause damage to the connection. 4.1. First damaging peak
Fig. 10. Load vs. displacement for the realistic wind loading trace shown in Fig. 4.
the first damaging peak (as indicated in the figure), the connection is permanently displaced by approximately 0.4 mm. This is observed by the load–displacement curve shifting to the right. Each subsequent damaging peak increases the permanent displacement of the connection until ultimate failure occurs. Surprisingly, between these damaging peaks, the load–displacement behavior remains similar to the undamaged case, following the same slope. This indicates that the stiffness of the connection remains unchanged despite having been partially removed from the topplate. Furthermore, because of this incremental removal of the nails, the connection no longer fails at the maximum load applied to the connection as shown in Fig. 10, where the failure load was 3.3 kN as compared to a maximum applied load of 3.4 kN. While this difference is not substantial, the ratio of the failure load
The first peak load to cause a permanent incremental withdrawal of the nails from the top-plate or ‘‘first damaging peak’’ occurred at one of 3 unique peaks in the loading time history for all 25 specimens tested. These 3 first damaging peaks are summarized in Table 4. It would be expected that the load required to cause the first damaging peak for a particular specimen would be a random variable that would follow a certain distribution, similar to the failure capacities shown in Figs. 6 and 7. However, under a single, repeated, fluctuating time series not all load levels are applied (as they are in a ramp) since the peak values may be significantly larger than any previous peak value applied, as shown in column 4 in Table 4. As a result, while the load required to cause initial damage may follow a distribution, the loads that actually cause the initial damage will cluster around specific peaks as observed for the particular chosen load time history. For example, for the first damaging peak, which occurs at 755.6 s, the true initial damaging threshold of the specimens can range from 1.09 to 1.47 kN, which is the range from just above the previous largest peak to the peak actually causing the damage.
74
M.J. Morrison, G.A. Kopp / Engineering Structures 33 (2011) 69–76
Table 4 Summary of the first damaging peaks under realistic wind loading. Trace time in Fig. 4 (s)
Number of specimens
Load at the peak (kN)
Largest load prior to damaging peak (kN)
Maximum load applied for tests with this as a first damaging peak (kN) Min, Mean, Max
755.6 999.3 1502.5
6 10 9
1.47 1.87 2.49
1.09 1.64 1.89
1.99, 2.63, 3.73 2.30, 2.87, 3.58 2.76, 3.24, 4.14
It might also be expected that the lower the initial damaging peak, the weaker the specimen, i.e., the lower the maximum load the specimen can resist. This has been found to be generally true, although the exception to this trend is the strongest specimen between the peaks at 755.6 s and 999.3 s; however, this discrepancy is likely the result of having insufficient specimens to account for the complete variability in strength of the connections. Furthermore, it is important to point out that none of the specimens failed at the first damaging peak for that specimen indicating that toe-nails subjected to hurricane wind loads will always have some form of damage to them prior to the wind loading that actually causes the failure of the connection. The average load that causes the initial damage to the connection (damage threshold) as a function of the mean maximum load applied to the connection ranges from 56% to 77%.
load until failure. However, for the current loading time history, there does not appear to be any fatigue to the connection that would cause the ramp loading experiments to be unconservative. It is possible that a trace with lower amplitudes, but with a much longer duration, could induce a fatigue in the connection, as it has been shown that following certain peak loads there can be a reduction in the capacity of the nails. However, the longer the duration of the wind event, the higher the probability that a larger peak will occur, even at the same wind speed. Furthermore, the failure loads were on average only 0.2 kN less than the peak capacity of the connection, this is significantly less than the reduction of capacity that can occur due to errors in construction such as a missing nail which, on average, reduces the capacity of the connection by 0.65 kN. 5. Discussion of current design standards
4.2. Damaging peaks Through all 25 fluctuating wind load tests, permanent, incremental damage to the connection due to local peak pressures was observed a total of 187 times or 7.5 peaks per connection, including the peak loads that caused failure. However, since the same loading time history was applied to each specimen, of the 187 damaging peaks, there were only 22 unique peak values. Out of these 22 peaks, 5 were found to damage 100% of the connections that saw that particular peak (obviously, the later on in the time series the peak occurred the lower the number of specimens that actually saw that particular peak due to failures of specimens). The load of these 5 peaks was found to always be of a higher magnitude than any previous peak in the load time history. However, peak loads that caused damage to greater than 40% of the test specimens were found to always have a larger magnitude than any peak load previously applied to that specimen. Among those peak loads that had a lower magnitude than the previous peak load, the average load difference was approximately 0.15 kN (with a maximum difference of 0.57 kN). The incremental pull-out of the nails for these peaks was always less than 1 mm. This suggests that the threshold load where damage occurs to the connection can be reduced following a damaging peak; the level to which it is reduced varies, as one would expect, for each specimen. However, it is only of order 5%. Figs. 6 and 7 show the statistical distributions of failure load from the ramp loading experiments along with the distribution of the maximum load applied during the fluctuating loading tests. The distribution for the maximum load applied to the specimen during the fluctuating load tests is shifted slightly to the right of the ramp load distributions in the plots indicating that, at least in a static sense, the current ramp loading tests are slightly conservative. Moreover, from Fig. 7 it appears that there is a lower probability of failure at lower load levels under realistic wind loading. That being said, the results from the realistic wind load experiments have shown that the connections will suffer permanent partial withdrawal at loads as low as 56% of the maximum applied load. Furthermore, during the realistic wind loading experiments, the connection does not necessarily fail once the peak load has been applied, but may fail at a latter point in time, and at a slightly lower load. This raises the question of duration effects and reduced capacity of the connection. The current study cannot answer this question completely since only one particular loading sequence was considered and was designed to, on average, increase the wind
5.1. Comparison of ramp test results with design values from the NDS The design capacity using Load Resistance Factor Design (LRFD) from the NDS [8] for a toe-nail can be obtained from: Wd = 47 690CM Ct Ceq Ctn KF φz λG2.5 DL
(4)
where Wd is the design capacity for the connection (in kN), CM , Ct , Ceq , Ctn , φz and λ are adjustment factors provided in the NDS, G is the specific gravity of the wood, D is the nail shank diameter and L in the nail penetration length in meters. For the current toe-nail connections, the toe-nail factor, Ctn , is equal to 0.67, the resistance factor, φz , is 0.65, the format conversion factor, KF , is 2.16/φz , and the remaining adjustment factors are all equal to 1. The nail shank diameter D, is 2.87 × 10−3 m, and the total penetration length, L, for three nails is approximately 0.15 m. Finally, the specific gravity, G, was assumed to be equal to 0.5 from the NDS for Douglas Fir-Larch timber used in the experiments. Using these values the design capacity, Wd , for a 3-nail toe-nail connections is 1.07 kN. This design value is lower than that required for any toe-nail connection to fail in the current investigation regardless of the loading type, including those connections that only used 2 nails, and has effectively a zero probability of occurrence as shown in Fig. 6. The above results indicate that the NDS may be overly conservative for toe-nail connections. As discussed by Sutt [15], the NDS [8] implicitly uses a factor of safety of 5 in Eq. (4), which would place the mean failure capacity on the order of 5 kN, which is stronger than any of tested toe-nail connections in the present study. Using the mean maximum capacity from the current test results and the capacity calculated from the NDS [8] the factor of safety in the NDS [8] for the current toe-nail connections would appear to be approximately 3. Considering an entire roof of a house, the capacity of each toe-nail connection used to connect the roof to the walls, is independent of each other; as a result it should be possible to statistically reduce the factor of safety as compared to that used for a single toe-nail connection, although this does not appear to be possible within the NDS [8] framework. It is noted that other factors, such as aging or greater variations in the quality of toe-nail connections outside the laboratory were not considered in the current study, and it is not clear if these factors are considered in the NDS [8]. For example, Shanmugam et al. [6] have shown that toe-nail connections in older construction, tested in the field, have a lower capacity than
M.J. Morrison, G.A. Kopp / Engineering Structures 33 (2011) 69–76
those tested in the laboratory. In addition, the NDS [8] may not be fully appropriate since it implies that the strength of the connection is directly proportional to the penetration length of the nails. In contrast, the current results, as well as those of previous studies, have shown that splitting of the timber is also a common failure mode. Moreover, it has been found that the capacity per penetration length of the nail increases with decreasing number of nails, which is not accounted for in the standard. 5.2. Comparison to ASCE 7-05 A comparison of test results with the ASCE7-05 [16] is performed herein to determine the design wind speeds based on the strength of the toe-nail connections. Due to the significant variability in toe-nail capacity, a factor of safety must be selected in order to make a useful comparison to a design standard. A study by [4] used factors of safety (FOS) of 2–3 along with the 5th and 10th percentile values from the failure capacity distributions, while [3] used a factor of safety of 2.5. The same house used to obtain the fluctuating wind loads described above was used as the basis to obtain the design wind forces from ASCE 7-05 [16]. It is unclear whether toenail connections should be treated as Main Wind Force Resisting Systems (MWFRS) or as a Components and Cladding (C&C) element in the ASCE 7-05 [16]. Consequently, the design forces at RTWC ‘S3’ were obtained by treating toe-nail connections as both a MWFRS and a C&C element. The design terrain was assumed to be category C, with wind directionality factor (Kd ), topographic factor (Kzt ), and importance factor (I) all equal to 1. The ASCE705 [16] wind speeds are then calculated and presented in Table 5, using the toe-nail capacities for the FOS used in [4] and a dead load due to the weight of the roof of 250 N per connection, for both MWFRS and C&C loads. It is noted that the 5th and 10th percentile peaks reported in Table 5 correspond well with the mean failure capacities found for toe-nail connections with missing nails presented in Table 3. In addition, since wind tunnel data for the building under consideration are available, GCpeq values are calculated from this data, using the procedure outlined by [17]. The GCpeq values are then used to calculate the wind tunnel design wind speed presented in Table 5. For the current test house the value ‘a’ in the ASCE 7-05 [16] that defines the different regions of the roof is 1.0 m and is defined by 10% of the least horizontal dimension. This means that, using a tributary area approach to calculate the loads on connection ‘S3’ as MWFRS, the tributary area (3.2 m2 ) is within Zone 3E, with a GCp value of −0.69. However, when using the C&C portion of the code, the tributary area for connection ‘S3’, is spread over three zones: Zone 1, 44%; Zone 2, 45%; and Zone 3, 11%. This results in an area average GCp of −1.47. The wind speeds shown in Table 5 show that the MWFRS loads calculated for this particular connection are lower than those obtained from the wind tunnel data. In contrast, the C&C loads are shown to be greater than the wind tunnel data. Considering the wind speeds calculated using the C&C loads and the wind tunnel pressure data, the toe-nail connections tested do not have sufficient hold down capacity for the highest loaded region of the test house, in any wind region given by the ASCE7-05 [16]. It is noted that the wind tunnel wind speeds meet the 90 mph wind speed region, provided no factor of safety is assumed. However, wind speeds calculated using the MWFRS coefficients from the ASCE7-05 [16] show that toe-nail connections can be used in several wind regions, when using the 5th and 10th percentile toenail strength. Caution should be taken however, since these results are significantly lower when compared to the wind tunnel data. In any case, to answer this question more definitively requires detailed experiments on the nature of the load sharing in the structure. Other factors, such as internal pressure and load sharing were not considered in the present analysis and both can have a
75
Table 5 ASCE 7-05 design wind speed calculated using both the MWFRS and C&C coefficients from the code as well as from wind tunnel pressure data. FOS
Toe-nail design capacity (kN)
1 2 3 5th percentile 10th percentile
2.8 1.4 0.93 1.9 2.2
ASCE7-05 design wind speed (mph) MWFRS
C&C Wind tunnel
108 80 67 91 97
64 47 40 53 57
92 65 65 76 82
Table 6 ASCE 7-05 design wind speed calculated using both the MWFRS and C&C coefficients from the code as well as from wind tunnel pressure data, assuming perfect load sharing between toe-nail connections on the roof. FOS
Toe-nail design capacity (kN)
1 2 3 5th percentile 10th percentile
2.8 1.4 0.93 1.9 2.2
ASCE7-05 design wind speed (mph) MWFRS C&C Wind tunnel 112 82 70 94 100
84 61 52 70 75
124 91 77 104 111
significant effect on the calculated failure wind speed. In the case of internal pressures, dominant openings in the windward wall can significantly lower the design wind speed [11]. At the present time, the amount of load sharing that occurs for wood-frame structures is unknown. The results presented in Table 5 assume that there is no load sharing between connections (i.e., the ‘tributary area’ approach), which would be a conservative assumption. The wind speeds presented in Table 6, have been calculated assuming perfect load sharing between all connections, whereby the total load applied to the roof is divided by the number of connections (i.e., all connections experience the same load). This assumption is likely unconservative for the most highly loaded connections. In contrast to the no load sharing case, by assuming perfect load sharing both the MWFRS and C&C loads from the ASCE7-05 [16] are conservative compared with the wind tunnel data for RTWC ‘S3’. The wind speeds for the MWFRS and C&C loads have increased in comparison to the no load sharing case shown in Table 5. For the MWFRS wind speeds the changes are small, while for the C&C wind speeds, the increase is larger. In the case of the wind tunnel data, the failure-inducing wind speeds are substantially higher. This result is not surprising since the relatively low spatial correlations of the wind loads on the roof are well known [13]. However, the added benefit of significant load sharing between connections is that the FOS could be reduced from that used when considering a single connection, as previously discussed. 6. Conclusions An experimental study was conducted on the uplift capacity of toe-nailed, roof-to-wall connections under both ramp and realistic, fluctuating wind-induced pressures. The failure capacity was found to be independent of loading rate, consistent with results from previous studies on nails [14]. Realistic wind loads were found to fail toe-nail connections incrementally, with pullout of the nails occurring due to a small number of peak loads in the time history. The failure capacity of the connections is found to be lower than the peak load applied to the connection, with an average reduction of 0.2 kN. However, the maximum load applied during fluctuating wind load experiments was found to be slightly higher than the failure capacity obtained from the ramp load experiments. This indicates that the ramp load tests are slightly conservative, compared to fluctuating wind load experiments. That being said, the fluctuating wind loading results have shown that
76
M.J. Morrison, G.A. Kopp / Engineering Structures 33 (2011) 69–76
the failure of the connection often occurs at a load lower than the maximum load applied to the connection. A comparison to the ASCE 7-05 has shown that assuming no load sharing between connections, failure of the most highly loaded connection would likely occur for all wind regions in ASCE 7-05 [16] during a design wind loading event. However, if perfect load sharing between all connections is assumed, it is likely that toe-nail connections could provide adequate hold down capacity for many wind regions in the ASCE7-05. Further work is required on the load sharing between connections to make a definitive statement; such work in on-going at the ‘Three Little Pigs’ Project at UWO. Acknowledgements M.J. Morrison gratefully acknowledges scholarship support from NSERC Canada. G.A. Kopp gratefully acknowledges the support provided by the Canada Research Chairs program. The authors gratefully acknowledge the help provided by Mr. Mitchell Cuddie in the development of the toe-nail rig. Equipment for this project was provided through grants obtained from the Canada Foundation for Innovation, Ontario Innovation Trust, and the University of Western Ontario. References [1] Sparks PR, Schiff SD, Reinhold TA. Wind damage to envelopes of houses and consequent insurance losses. J Wind Eng Ind Aerodyn 1994;53:145–55. [2] Kopp GA, Morrison MJ, Kordi B, Miller C. A method to assess peak storm wind speeds using detailed damage surveys. Eng Struct 2011;33:90–8.
[3] Cheng J. Testing and analysis of the toe–nailed connection in the residential roof-to-wall system. Forest Prod J 2004;54:58–65. [4] Reed TD, Rosowsky DV, Schiff SD. Uplift capacity of light-frame rafter to top plate connections. J Archit Eng 1997;3:156–63. [5] Riley MA, Sadek F. Experimental testing of roof to wall connections in wood frame houses. National Institute of Standards and Technology. NISTIR 6938. 2003. [6] Shanmugam B, Nielson BG, Prevatt DO. Statistical and analytical models for roof components in existing light-framed wood structures. Eng Struct 2009; 31:2607–16. [7] He M, Lam F, Foschi RO. Modeling three-dimensional timber light-frame buildings. J Struct Eng 2001;127:901–13. [8] American Forest & Paper Association, ASD/LRFD national design specification for wood construction. Washington (DC); 2005. [9] Kopp GA, Morrison MJ, Gavanski E, Henderson D, Hong HP. The ‘Three Little Pigs’ Project: hurricane risk mitigation by integrated wind tunnel and fullscale laboratory tests. Nat Hazard Rev 2010;11:151–61. [10] Kopp GA, Surry D, Mans C. Wind effects of parapets on low buildings: part 4. Mitigation of corner loads with alternative geometries. J Wind Eng Ind Aerodyn 2005;93:31–59. [11] Kopp GA, Oh JH, Inculet DR. Wind-induced internal pressures in houses. J Struct Eng 2008;134:1129–38. [12] Ho TCE, Surry D, Morrish D, Kopp GA. The UWO contribution to the NIST aerodynamic database for wind loads on low buildings: part I. Archiving format and basic aerodynamic data. J Wind Eng Ind Aerodyn 2005;93:1–30. [13] Surry D, Sinno RR, Nail B, Ho TCE, Farquhar S, Kopp GA. Structurally-effective static wind loads for roof panels. J Struct Eng 2007;133:871–85. [14] Rosowsky DV, Reinhold TA. Rate-of-load and duration-of-load effects for wood fasteners. J Struct Eng 1999;125:719–24. [15] Sutt Jr EG. The effect of combined shear and uplift forces on roof sheathing panels. Ph.D. thesis. Clemson (South Carolina, US): Clemson University; 2000. [16] ASCE 7-05. Minimum design loads for buildings and other structures. Reston (Virginia): American Society of Civil Engineers; 2006. [17] St. Pierre LM, Kopp GA, Surry D, Ho TCE. The UWO contribution to the NIST aerodynamic database for wind loads on low buildings: part 2. Comparison of data with wind load provisions. J Wind Eng Ind Aerodyn 2005;93:31–59.
Contents lists available at ScienceDirect
Engineering Structures journal homepage: www.elsevier.com/locate/engstruct
Performance of toe-nail connections under realistic wind loading Murray J. Morrison, Gregory A. Kopp ∗ Boundary Layer Wind Tunnel Laboratory, Faculty of Engineering, University of Western Ontario, London, ON, N6A 5B9, Canada
article
info
Article history: Received 15 July 2010 Received in revised form 2 September 2010 Accepted 16 September 2010 Available online 29 October 2010 Keywords: Wind loads Disaster mitigation Low-rise buildings Nails Wood-frame construction
abstract Using recently developed pressure loading actuators (PLA), ramp and realistic fluctuating wind loads are applied to toe-nail connections which are typically used in wood-frame residential construction. The failure capacity from the ramp and fluctuating wind load tests are found to be similar, and are comparable to capacities reported in the literature. However, under realistic wind loading, the toe-nail connections are found to fail in increments with the majority of the damage to the connection occurring intermittently at the peak pressures so that it takes many peaks for a connection to fail. In addition, the effects of construction defects, in this case missing nails, were also examined in order to determine the reduction in capacity. Considering the wind loads on a typical house, estimates for failure wind speeds were obtained, assuming a factor of safety, and compared to ASCE 7-05 wind regions. © 2010 Elsevier Ltd. All rights reserved.
1. Introduction Roof failures are commonly observed in extreme wind events, including roof-to-wall connections (RTWC) and sheathing in residential wood-frame housing. In addition to life safety issues, failure of a roof or its components can cause significant water ingress, with subsequent damage to the building contents, substantially increasing the insured losses [1]. Furthermore, with all or part of the roof missing, the walls are more susceptible to collapse which can become a significant life safety issue. Toe-nails are the most common type of RTWC in North America. While it is now common to use hurricane straps in hurricane prone regions such as Florida and the Gulf Coast of North America, in new construction, there are still a large number of existing structures with toe-nail connections as the primary hold down for the roof structure. Moreover, in non-hurricane regions toe-nails are still used in new construction as the primary roof-to-wall connection. These regions are susceptible to extreme wind events, such as tornadoes and downbursts, which are capable of causing complete roof failures, a recent example of which is discussed in [2]. The capacity of toe-nail connections to uplift loads has been the subject of several studies, including those by [3–6]. These tests applied load at a constant displacement rate (typically 2.54–6.35 mm/min) and measured the force required to keep the connection moving at the set rate. These tests are nominally static and quite different from the highly fluctuating loads generated by
∗
Corresponding author. Tel.: +1 519 661 3338; fax: +1 519 661 3339. E-mail addresses: [email protected], [email protected] (G.A. Kopp).
0141-0296/$ – see front matter © 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.engstruct.2010.09.019
real winds. Experiments conducted by [6] on in situ connections did apply a form of cyclical loading, although a maximum of only three loading cycles were used. The cycles were applied at a low displacement rate of 2.54 mm/min and the end of a cycle was based on a displacement threshold. The loading from these tests was also significantly different from that induced by real wind loads where there are a large number of cycles and the loads can double or increase more than that in less than a second and decrease just as quickly. The mean maximum withdrawal capacity from these studies is in the range from 1130 N to 2840 N, depending on the type, number of nails, age of the connection, type of wood, and the wood moisture content. To date there has been no study to document the effects, if any, that cyclical or realistic wind loads have on the withdrawal performance of toe-nailed RTWC. Hysteretic behavior and reduction of the failure capacity when subjected to cyclical shear loads, have been tested and implemented in finite element models of entire structures, e.g., [7], primarily for studies involving earthquake loading. The current study examines the withdrawal behavior of toe-nail connections under realistic wind uplift loads and compares the response to that found from static testing. 2. Experimental setup Toe-nailed RTWC are tested using a load control approach with the test rig shown in Fig. 1. The loads are applied to the specimen by controlling the pressure inside of the blue airbag, which is then mechanically attached to the toe-nail specimen, as shown in inset A of Fig. 1. The toe-nail connections consist of two 0.61 m long 2 × 4 segments, representing a typical portion of a top-plate for woodframed stud walls, mounted at either end to load cells to measure
70
M.J. Morrison, G.A. Kopp / Engineering Structures 33 (2011) 69–76
Fig. 1. Photograph of the nail test rig. Loads are controlled by altering the pressure in the air bag, while load cells and a displacement transducer measure the response of the toe-nail connection.
Fig. 2. Pressure tap layout for the wind tunnel model of the full-scale test house. Dimensions are provided in full scale.
the reaction at the connection. A 0.3 m long 2 × 6, representing a typical rafter or roof truss section is connected to the bottom of the airbag; the rafter is then toe-nailed to the top-plate using 3 12-d twisted shank nails using a pneumatic nail gun. In this paper the term ‘d-nails’ will refer to the side of the rafter with two nails, while the ‘s-nails’ will refer to the side of the rafter with just a single nail. For the current experimental setup the loads are applied normal to the top-plate (upward in Fig. 1), meaning that only withdrawal loads are applied to the toe-nail connection. It is noted that each nail will not be subjected to pure withdrawal loads due to the angle at which the nails are driven; however, this description of the loading is consistent with the National Design Specification for Wood Construction (NDS) [8]. In addition, flat roof construction is not typical in most types of residential construction; common roof slopes are typically between 3 on 12 and 6 on 12, although this range can vary substantially by region. As a result a portion of the uplift wind loads applied to the roof of the house will not apply a perfect withdrawal load to the toe-nail connections. In the case of a 4 on 12 roof slope this results in approximately 5% of the applied load to the roof being applied in shear to the connection. The amount of load applied in shear due to the angle of the roof is likely much smaller than the shear loads that would be applied to the toe-nail connections due to the loads applied to the wall. The purpose of the current investigation is to examine the withdrawal capacity of only toe-nail connections under realistic uplifting wind loads; as such, shear loads have been explicitly ignored and is consistent with how loads have been applied in previous studies. A displacement transducer is mounted to the rafter and measures the displacements of the rafter relative to the flat concrete floor, as shown in Fig. 1. Since the top-plate is kept stationary throughout the experiment, the measured displacement is equivalent to the displacement of the rafter relative to the topplate. The pressure in the airbag is controlled using a pressure loading actuator (PLA) which is able to accurately replicate temporally varying pressures from an input pressure trace. Further details regarding the PLAs, and their capabilities, can be found in [9]. As discussed above, the capacity of toe-nail connections varies substantially in the literature based on the type of nails, grade of lumber, moisture content and age of the connection. The goal of the current investigation is to determine the behavior and capacity of toe-nail connections under realistic wind loading. To achieve this objective, the static capacity of the connections used in the current study must first be determined using a loading method similar
to that of previous studies in order to obtain a baseline static withdrawal capacity. To obtain the static capacity, and to assess the effect of loading rate on the response of the connections, ramp loads are applied to the specimen using three different loading rates, viz., 1, 8, and 32 kN/min, resulting in expected failure times in the range from 5 to 200 s. These ramp loading rates were selected so that the slowest rate would result in a failure time similar to that of previous experiments where a constant displacement rate was used, while the highest loading rate would approximate a loading rate closer to that found in actual wind loads. A total of 21 toenail specimens were tested at each ramp loading rate in order to account for variability in wood properties and construction. In typical construction, there can be significant variability in the quality of a toe-nail connection due to construction errors and defects. To attempt to quantify the effect of these errors in a controlled way, experiments were conducted using a ramp loading rate of 8 kN/min using toe-nail connections that were missing a single nail, which results in two types of defect cases. Defect #1 will refer to tests where the‘s-nail’ is missing, while defect #2, will refer to tests where one of the ‘d-nails’ is missing. Since actual wind loads vary significantly based not only on the building geometry, but also on the location on the particular building as well, there are nearly infinite possible wind loading time series that could be applied. The realistic, fluctuating loading for the current experiments were obtained from wind tunnel experiments which measured the roof pressures on the ‘Three Little Pigs’ test house, described in [9], at a scale of 1:50 in Boundary Layer Wind Tunnel II at the University of Western Ontario (UWO). This house has full-scale plan dimensions of 9.0 m by 8.9 m, an eave height of 8.0 m and a gable roof slope of 4:12. The dimensions, along with the tap layout, can be found in Fig. 2. The flow simulation approximates a typical open country atmospheric boundary layer at a scale of 1:50 with an aerodynamic roughness length, zo , of approximately 0.03 m (equivalent full scale). The terrain simulation used in the current study is identical to that of [10,11], where a detailed discussion of the flow simulation and modeling approaches can be found. Pressure taps were connected to PSI pressure transducers using the tubing system presented in [12] and have a frequency response which is flat up to 200 Hz. The pressures were sampled nearly simultaneously (maximum lag of 0.0025 s) for a total of 180 s at a frequency of 400 Hz. In total, 18 wind angles were tested ranging from 0° to 90° at a mean roof height wind speed of 9.6 m/s and Reynolds number, Re = VH /ν = 1.0 × 105 .
M.J. Morrison, G.A. Kopp / Engineering Structures 33 (2011) 69–76
71
Table 1 Summary of the real wind load trace applied to the toe-nail connection. Scaling wind speed (m/s)
Peak force (kN)
Trace duration (s)
20 25 30 35
1.35 2.45 3.47 4.91
900 720 600 514
Fig. 3. Truss layout for the ‘Three Little Pigs’ test house at UWO. The spacing between trusses is nominally 0.61 m on center and each toe-nail connection is labeled in the figure.
Fig. 3 presents the truss layout for the full-scale house. Using this layout of the trusses, the force coefficient, Cf , time series at each RTWC of the house can be estimated using: Cf (t ) =
∫
Cp (x, y, z , t ) · I (x, y, z )
A
da A
(1)
where I (x, y, z ) is the structural influence function, Cp is the pressure coefficient over a small area da, and A is the area where the influence function is non-zero. In the current study, a geometric tributary area approach is used to evaluate I (x, y, z ), as it is not generally known for wood-frame houses. The RTWC with the largest peak magnitude of Cf in the time series was then chosen for these experiments, which corresponded to RTWC ‘S3’. Rather than using the entire 1 h (equivalent full scale) time history, a representative portion was selected for the current experiments. The full-scale force, F , was obtained using: F (t ) = 0.5ρ U 2 Cf (t )A − W
(2)
where U is the hourly mean wind speed at mean roof height, ρ is the density of air and W is the portion of the weight of the roof associated with RTWC ‘S3’. The current approach to apply realistic loads is similar to that used by [13]. A wind speed, U, of 20 m/s was chosen for the first test for each connection and the resulting force time series, using Eq. (2), was applied to the connection. If the connection did not fail, the scaling wind speed was increased by 5 m/s and the new force time history was re-applied to the test specimen. This process was repeated until failure occurred. It is noted for the current experiments that the loads at each scaling wind speed were applied sequentially with no pause for changes in the scaling wind speed. In addition, since the weight of the roof, W , has been removed in Eq. (2), the wind loads at different wind speeds will no longer scale with the square of the velocity for each test. This choice does not affect any of the conclusions, but aids in the interpretation of the results. Fig. 4 provides an example of the entire force time series that is applied to the specimen, with the locations where the scaling wind speed changes are also indicated. Table 1 provides details of the force time history at each scaling wind speed. It is also noted, because of the scaling laws, i.e.,
νT L
= model-scale
νT L
Fig. 4. Force time series of the realistic wind loads for all scaling wind speeds applied.
(3) full-scale
that, as the scaling wind speed increases, the length of the test decreases. It was decided that the same segment of wind tunnel data would be used so that an identical number of peaks occur for each scaling wind speed, but the actual duration is shortened. This can also be inferred from Fig. 4. In total 25 specimens were tested using the realistic wind loading time series presented in Fig. 4.
Fig. 5. Load vs. displacement relationship for a ramp rate of 8 kN/min.
3. Results 3.1. Ramp loading Fig. 5 shows a typical load–displacement relationship for one of the 8 kN/min ramp tests. The failure capacity of the connection is defined as the maximum measured reaction, as indicated in the figure. After this point, the measured reaction drops as the nails pull out. Since the current study uses a load control approach, the failure proceeds in an uncontrolled manner; therefore, the data beyond the failure capacity are not used in the analysis. Similar to [6], the log-normal distribution was found to best fit the maximum capacities when considering normal, Weibull or Gumbel distributions. The cumulative distribution for all three ramp loading rates is shown in Fig. 6. The distributions for the ramp rates of 8 and 32 kN/min are similar, while the 1 kN/min test has a lower failure capacity. The mean failure capacity for each ramp rate is nearly the same ranging from 2.7 kN at 1 kN/min to 2.9 at 32 kN/min. The difference in mean failure capacity based on loading rate is insignificant as compared to the range of failure capacities observed, which ranged from 1.2 to 4.7 kN from all
72
M.J. Morrison, G.A. Kopp / Engineering Structures 33 (2011) 69–76 Table 2 Summary of the types of failures from the ramp loading tests and there average failure capacity. Failure type
Mean failure capacity (kN)
Standard deviation (kN)
% of failures
All nails split d-nails split s-nail split All nails pull out
3.3 3.2 3.2 2.6
0.58 0.41 0.87 0.54
8 21 6 65
Table 3 Summary of failure capacity and failure modes of connections with defects for ramp tests.
Fig. 6. Cumulative distribution of the mean failure capacity for all 3 ramp loading rates and the mean maximum applied load for the realistic wind loading tests.
Fig. 7. Probability distribution for the same data as presented in Fig. 6. The inset in the top right of the figure presents the probabilities on a logarithmic scale.
ramp rates. In order to determine if data from each loading rate follows the same distribution, a null hypothesis is formed that each loading rate can be represented by the same distribution. Using a Kolmogorov–Smirnov test, the null hypothesis cannot be rejected between the 3 ramp loading rates using a 95% confidence interval. This indicates that, with the current data, the failure capacities for each loading rate can be represented by the same distribution. Moreover, using the same confidence interval and an analysis of the variance (ANOVA), the mean failure capacity from each loading rate are not significantly different. This means that the failure capacity of the toe-nail connections is independent of the loading rate within this range, which encompasses nearly all loading rates expected to occur under actual wind loading. This finding is consistent with that of [14], which found no loading rate dependence in the capacity of nailed connections (although toe-nails were not examined). Fig. 7 presents the probability distribution for the same ramp rate data used for Fig. 6, and shows that distributions become broader with increasing loading rate suggesting that there is an increased variability of the failure capacity. In addition, the inset of Fig. 7 shows that the distributions vary significantly in the tail regions of the distribution, which is likely the result of an imperfect fit of the log-normal distribution to the experimental data. Additional samples at the current loading rates along with investigating more loading rates between 1 and 8 kN/min would be needed to determine if the reduced variability and lower failure capacity observed for 1 kN/min is a true loading rate effect. This point is not investigated further herein. During the ramp loading experiments the nails were observed to fail in two different ways: either ‘‘pull-out’’, where the nails pulled out of the top-plate; or ‘‘splitting’’, where the wood of the
No defect Defect #1 Defect #2
Mean failure capacity (kN)
Standard deviation (kN)
#split/#pull outs
2.8 1.9 2.2
0.6 0.46 0.48
22/41 11/5 0/16
rafter splits with the nails remaining attached to the top-plate. As a result of these two types of failures, four different failure modes were observed, all nails split, d-nails split, the s-nail split, or all nails pull out. Fig. 8(a) shows an example of a pull-out failure, while Fig. 8(b) presents an s-nail split failure. The ramp test results sorted by failure type are shown in Table 2. Since the loading rate was found to have a minimal effect on the capacity of the toe-nail connections, all ramp rate test results have been used. These results show that the predominant failure mode is when all nails pull out together. This failure mode has a lower failure capacity than when splitting failures occur, which is consistent with the results found in [6]. However, there does not seem to be any significant difference in capacity between types of splitting failures, although the d-nails splits were much more common than the other two types of splitting failures indicating that the proximity of the two nails on one side may cause that side to be more prone to splitting. 3.2. Effect of missing nails In total, 16 tests were conducted for each defect condition discussed above. The mean and standard deviation of the failure capacity, along with the type of failure, are given in Table 3. For both defect cases, the failure capacity per nail is higher than that of the no defect case. The mean failure capacity for defect #1 is lower than that of defect #2. The predominant failure mode of defect #1 is splitting, while only pull-out failures are observed for defect #2. This is surprising, since, for the no defect case, splitting failures are generally associated with a higher mean failure capacity, although it is likely the result of the extreme asymmetry of the connection in the case of defect #1. 3.3. Fluctuating wind loading A total of 25 specimens were tested using the realistic, fluctuating wind loading trace shown in Fig. 4. A displacement time series for a typical specimen is shown in Fig. 9. Rather than the connection gradually being withdrawn from the top-plate as in the ramp loading experiments, partial permanent withdrawal of the nails accumulates over handful of peak loads. The peak loads that cause this incremental withdrawal of the nails will be referred to as ‘‘damaging peaks’’ in the following discussion. For the displacement time history shown in Fig. 9, the damaging peaks are indicated in the figure. Fig. 10 shows the load versus displacement relationship for the same test as shown in Fig. 9. Over the first portion of the test, prior to any damaging peaks, the load–displacement behavior is nearly linear; however, following
M.J. Morrison, G.A. Kopp / Engineering Structures 33 (2011) 69–76
a
73
b
Fig. 8. (a) Photograph of a toe-nail failure where all nails have pulled out of the toe-plate. (b) Photograph of an ‘s-split’ failure of the toe-nail connection.
to the maximum load applied was observed to range from 87% to 100%. The overall mean maximum load applied during the realistic fluctuating wind loading was found to be the same as the ramp loading rates, being 2.8 kN. Pull-outs are the most common mode of failure, representing 76% of the cases. This is slightly higher than that of the ramp loading tests. However, unlike the ramp loading, there is no observed change in the mean maximum applied load between pull-out and splitting failures under the realistic, fluctuating wind load. 4. Discussion of temporal effects
Fig. 9. Displacement time series of a typical realistic fluctuating wind loading trace applied to a toe-nail connection. The damaging peaks are indicated in the figure.
While the failure of toe-nail connections due to ramp loads was a gradual withdrawal of nails, the nails subjected to realistic fluctuating loading are partially withdrawn from the top-plate during the peak pressures in the applied trace. The ramp loading tests, as well as the dynamic experiments, have shown that there is no significant dependence on the loading rate in the response of these connections. As previously discussed, the highest ramp loading rate is comparable to loading rates applied under realistic wind loading. However, it should be noted that under realistic wind loads as the connection becomes stronger and requires larger scaling wind speeds to cause failure, the higher the loading rate will become as indicated from Eq. (3). Despite the difference in the observed behavior between the ramp and realistic wind loading the average failure capacity is remarkably similar. The following discussion will examine the loading time history in detail and discuss the peak loads that cause damage to the connection. 4.1. First damaging peak
Fig. 10. Load vs. displacement for the realistic wind loading trace shown in Fig. 4.
the first damaging peak (as indicated in the figure), the connection is permanently displaced by approximately 0.4 mm. This is observed by the load–displacement curve shifting to the right. Each subsequent damaging peak increases the permanent displacement of the connection until ultimate failure occurs. Surprisingly, between these damaging peaks, the load–displacement behavior remains similar to the undamaged case, following the same slope. This indicates that the stiffness of the connection remains unchanged despite having been partially removed from the topplate. Furthermore, because of this incremental removal of the nails, the connection no longer fails at the maximum load applied to the connection as shown in Fig. 10, where the failure load was 3.3 kN as compared to a maximum applied load of 3.4 kN. While this difference is not substantial, the ratio of the failure load
The first peak load to cause a permanent incremental withdrawal of the nails from the top-plate or ‘‘first damaging peak’’ occurred at one of 3 unique peaks in the loading time history for all 25 specimens tested. These 3 first damaging peaks are summarized in Table 4. It would be expected that the load required to cause the first damaging peak for a particular specimen would be a random variable that would follow a certain distribution, similar to the failure capacities shown in Figs. 6 and 7. However, under a single, repeated, fluctuating time series not all load levels are applied (as they are in a ramp) since the peak values may be significantly larger than any previous peak value applied, as shown in column 4 in Table 4. As a result, while the load required to cause initial damage may follow a distribution, the loads that actually cause the initial damage will cluster around specific peaks as observed for the particular chosen load time history. For example, for the first damaging peak, which occurs at 755.6 s, the true initial damaging threshold of the specimens can range from 1.09 to 1.47 kN, which is the range from just above the previous largest peak to the peak actually causing the damage.
74
M.J. Morrison, G.A. Kopp / Engineering Structures 33 (2011) 69–76
Table 4 Summary of the first damaging peaks under realistic wind loading. Trace time in Fig. 4 (s)
Number of specimens
Load at the peak (kN)
Largest load prior to damaging peak (kN)
Maximum load applied for tests with this as a first damaging peak (kN) Min, Mean, Max
755.6 999.3 1502.5
6 10 9
1.47 1.87 2.49
1.09 1.64 1.89
1.99, 2.63, 3.73 2.30, 2.87, 3.58 2.76, 3.24, 4.14
It might also be expected that the lower the initial damaging peak, the weaker the specimen, i.e., the lower the maximum load the specimen can resist. This has been found to be generally true, although the exception to this trend is the strongest specimen between the peaks at 755.6 s and 999.3 s; however, this discrepancy is likely the result of having insufficient specimens to account for the complete variability in strength of the connections. Furthermore, it is important to point out that none of the specimens failed at the first damaging peak for that specimen indicating that toe-nails subjected to hurricane wind loads will always have some form of damage to them prior to the wind loading that actually causes the failure of the connection. The average load that causes the initial damage to the connection (damage threshold) as a function of the mean maximum load applied to the connection ranges from 56% to 77%.
load until failure. However, for the current loading time history, there does not appear to be any fatigue to the connection that would cause the ramp loading experiments to be unconservative. It is possible that a trace with lower amplitudes, but with a much longer duration, could induce a fatigue in the connection, as it has been shown that following certain peak loads there can be a reduction in the capacity of the nails. However, the longer the duration of the wind event, the higher the probability that a larger peak will occur, even at the same wind speed. Furthermore, the failure loads were on average only 0.2 kN less than the peak capacity of the connection, this is significantly less than the reduction of capacity that can occur due to errors in construction such as a missing nail which, on average, reduces the capacity of the connection by 0.65 kN. 5. Discussion of current design standards
4.2. Damaging peaks Through all 25 fluctuating wind load tests, permanent, incremental damage to the connection due to local peak pressures was observed a total of 187 times or 7.5 peaks per connection, including the peak loads that caused failure. However, since the same loading time history was applied to each specimen, of the 187 damaging peaks, there were only 22 unique peak values. Out of these 22 peaks, 5 were found to damage 100% of the connections that saw that particular peak (obviously, the later on in the time series the peak occurred the lower the number of specimens that actually saw that particular peak due to failures of specimens). The load of these 5 peaks was found to always be of a higher magnitude than any previous peak in the load time history. However, peak loads that caused damage to greater than 40% of the test specimens were found to always have a larger magnitude than any peak load previously applied to that specimen. Among those peak loads that had a lower magnitude than the previous peak load, the average load difference was approximately 0.15 kN (with a maximum difference of 0.57 kN). The incremental pull-out of the nails for these peaks was always less than 1 mm. This suggests that the threshold load where damage occurs to the connection can be reduced following a damaging peak; the level to which it is reduced varies, as one would expect, for each specimen. However, it is only of order 5%. Figs. 6 and 7 show the statistical distributions of failure load from the ramp loading experiments along with the distribution of the maximum load applied during the fluctuating loading tests. The distribution for the maximum load applied to the specimen during the fluctuating load tests is shifted slightly to the right of the ramp load distributions in the plots indicating that, at least in a static sense, the current ramp loading tests are slightly conservative. Moreover, from Fig. 7 it appears that there is a lower probability of failure at lower load levels under realistic wind loading. That being said, the results from the realistic wind load experiments have shown that the connections will suffer permanent partial withdrawal at loads as low as 56% of the maximum applied load. Furthermore, during the realistic wind loading experiments, the connection does not necessarily fail once the peak load has been applied, but may fail at a latter point in time, and at a slightly lower load. This raises the question of duration effects and reduced capacity of the connection. The current study cannot answer this question completely since only one particular loading sequence was considered and was designed to, on average, increase the wind
5.1. Comparison of ramp test results with design values from the NDS The design capacity using Load Resistance Factor Design (LRFD) from the NDS [8] for a toe-nail can be obtained from: Wd = 47 690CM Ct Ceq Ctn KF φz λG2.5 DL
(4)
where Wd is the design capacity for the connection (in kN), CM , Ct , Ceq , Ctn , φz and λ are adjustment factors provided in the NDS, G is the specific gravity of the wood, D is the nail shank diameter and L in the nail penetration length in meters. For the current toe-nail connections, the toe-nail factor, Ctn , is equal to 0.67, the resistance factor, φz , is 0.65, the format conversion factor, KF , is 2.16/φz , and the remaining adjustment factors are all equal to 1. The nail shank diameter D, is 2.87 × 10−3 m, and the total penetration length, L, for three nails is approximately 0.15 m. Finally, the specific gravity, G, was assumed to be equal to 0.5 from the NDS for Douglas Fir-Larch timber used in the experiments. Using these values the design capacity, Wd , for a 3-nail toe-nail connections is 1.07 kN. This design value is lower than that required for any toe-nail connection to fail in the current investigation regardless of the loading type, including those connections that only used 2 nails, and has effectively a zero probability of occurrence as shown in Fig. 6. The above results indicate that the NDS may be overly conservative for toe-nail connections. As discussed by Sutt [15], the NDS [8] implicitly uses a factor of safety of 5 in Eq. (4), which would place the mean failure capacity on the order of 5 kN, which is stronger than any of tested toe-nail connections in the present study. Using the mean maximum capacity from the current test results and the capacity calculated from the NDS [8] the factor of safety in the NDS [8] for the current toe-nail connections would appear to be approximately 3. Considering an entire roof of a house, the capacity of each toe-nail connection used to connect the roof to the walls, is independent of each other; as a result it should be possible to statistically reduce the factor of safety as compared to that used for a single toe-nail connection, although this does not appear to be possible within the NDS [8] framework. It is noted that other factors, such as aging or greater variations in the quality of toe-nail connections outside the laboratory were not considered in the current study, and it is not clear if these factors are considered in the NDS [8]. For example, Shanmugam et al. [6] have shown that toe-nail connections in older construction, tested in the field, have a lower capacity than
M.J. Morrison, G.A. Kopp / Engineering Structures 33 (2011) 69–76
those tested in the laboratory. In addition, the NDS [8] may not be fully appropriate since it implies that the strength of the connection is directly proportional to the penetration length of the nails. In contrast, the current results, as well as those of previous studies, have shown that splitting of the timber is also a common failure mode. Moreover, it has been found that the capacity per penetration length of the nail increases with decreasing number of nails, which is not accounted for in the standard. 5.2. Comparison to ASCE 7-05 A comparison of test results with the ASCE7-05 [16] is performed herein to determine the design wind speeds based on the strength of the toe-nail connections. Due to the significant variability in toe-nail capacity, a factor of safety must be selected in order to make a useful comparison to a design standard. A study by [4] used factors of safety (FOS) of 2–3 along with the 5th and 10th percentile values from the failure capacity distributions, while [3] used a factor of safety of 2.5. The same house used to obtain the fluctuating wind loads described above was used as the basis to obtain the design wind forces from ASCE 7-05 [16]. It is unclear whether toenail connections should be treated as Main Wind Force Resisting Systems (MWFRS) or as a Components and Cladding (C&C) element in the ASCE 7-05 [16]. Consequently, the design forces at RTWC ‘S3’ were obtained by treating toe-nail connections as both a MWFRS and a C&C element. The design terrain was assumed to be category C, with wind directionality factor (Kd ), topographic factor (Kzt ), and importance factor (I) all equal to 1. The ASCE705 [16] wind speeds are then calculated and presented in Table 5, using the toe-nail capacities for the FOS used in [4] and a dead load due to the weight of the roof of 250 N per connection, for both MWFRS and C&C loads. It is noted that the 5th and 10th percentile peaks reported in Table 5 correspond well with the mean failure capacities found for toe-nail connections with missing nails presented in Table 3. In addition, since wind tunnel data for the building under consideration are available, GCpeq values are calculated from this data, using the procedure outlined by [17]. The GCpeq values are then used to calculate the wind tunnel design wind speed presented in Table 5. For the current test house the value ‘a’ in the ASCE 7-05 [16] that defines the different regions of the roof is 1.0 m and is defined by 10% of the least horizontal dimension. This means that, using a tributary area approach to calculate the loads on connection ‘S3’ as MWFRS, the tributary area (3.2 m2 ) is within Zone 3E, with a GCp value of −0.69. However, when using the C&C portion of the code, the tributary area for connection ‘S3’, is spread over three zones: Zone 1, 44%; Zone 2, 45%; and Zone 3, 11%. This results in an area average GCp of −1.47. The wind speeds shown in Table 5 show that the MWFRS loads calculated for this particular connection are lower than those obtained from the wind tunnel data. In contrast, the C&C loads are shown to be greater than the wind tunnel data. Considering the wind speeds calculated using the C&C loads and the wind tunnel pressure data, the toe-nail connections tested do not have sufficient hold down capacity for the highest loaded region of the test house, in any wind region given by the ASCE7-05 [16]. It is noted that the wind tunnel wind speeds meet the 90 mph wind speed region, provided no factor of safety is assumed. However, wind speeds calculated using the MWFRS coefficients from the ASCE7-05 [16] show that toe-nail connections can be used in several wind regions, when using the 5th and 10th percentile toenail strength. Caution should be taken however, since these results are significantly lower when compared to the wind tunnel data. In any case, to answer this question more definitively requires detailed experiments on the nature of the load sharing in the structure. Other factors, such as internal pressure and load sharing were not considered in the present analysis and both can have a
75
Table 5 ASCE 7-05 design wind speed calculated using both the MWFRS and C&C coefficients from the code as well as from wind tunnel pressure data. FOS
Toe-nail design capacity (kN)
1 2 3 5th percentile 10th percentile
2.8 1.4 0.93 1.9 2.2
ASCE7-05 design wind speed (mph) MWFRS
C&C Wind tunnel
108 80 67 91 97
64 47 40 53 57
92 65 65 76 82
Table 6 ASCE 7-05 design wind speed calculated using both the MWFRS and C&C coefficients from the code as well as from wind tunnel pressure data, assuming perfect load sharing between toe-nail connections on the roof. FOS
Toe-nail design capacity (kN)
1 2 3 5th percentile 10th percentile
2.8 1.4 0.93 1.9 2.2
ASCE7-05 design wind speed (mph) MWFRS C&C Wind tunnel 112 82 70 94 100
84 61 52 70 75
124 91 77 104 111
significant effect on the calculated failure wind speed. In the case of internal pressures, dominant openings in the windward wall can significantly lower the design wind speed [11]. At the present time, the amount of load sharing that occurs for wood-frame structures is unknown. The results presented in Table 5 assume that there is no load sharing between connections (i.e., the ‘tributary area’ approach), which would be a conservative assumption. The wind speeds presented in Table 6, have been calculated assuming perfect load sharing between all connections, whereby the total load applied to the roof is divided by the number of connections (i.e., all connections experience the same load). This assumption is likely unconservative for the most highly loaded connections. In contrast to the no load sharing case, by assuming perfect load sharing both the MWFRS and C&C loads from the ASCE7-05 [16] are conservative compared with the wind tunnel data for RTWC ‘S3’. The wind speeds for the MWFRS and C&C loads have increased in comparison to the no load sharing case shown in Table 5. For the MWFRS wind speeds the changes are small, while for the C&C wind speeds, the increase is larger. In the case of the wind tunnel data, the failure-inducing wind speeds are substantially higher. This result is not surprising since the relatively low spatial correlations of the wind loads on the roof are well known [13]. However, the added benefit of significant load sharing between connections is that the FOS could be reduced from that used when considering a single connection, as previously discussed. 6. Conclusions An experimental study was conducted on the uplift capacity of toe-nailed, roof-to-wall connections under both ramp and realistic, fluctuating wind-induced pressures. The failure capacity was found to be independent of loading rate, consistent with results from previous studies on nails [14]. Realistic wind loads were found to fail toe-nail connections incrementally, with pullout of the nails occurring due to a small number of peak loads in the time history. The failure capacity of the connections is found to be lower than the peak load applied to the connection, with an average reduction of 0.2 kN. However, the maximum load applied during fluctuating wind load experiments was found to be slightly higher than the failure capacity obtained from the ramp load experiments. This indicates that the ramp load tests are slightly conservative, compared to fluctuating wind load experiments. That being said, the fluctuating wind loading results have shown that
76
M.J. Morrison, G.A. Kopp / Engineering Structures 33 (2011) 69–76
the failure of the connection often occurs at a load lower than the maximum load applied to the connection. A comparison to the ASCE 7-05 has shown that assuming no load sharing between connections, failure of the most highly loaded connection would likely occur for all wind regions in ASCE 7-05 [16] during a design wind loading event. However, if perfect load sharing between all connections is assumed, it is likely that toe-nail connections could provide adequate hold down capacity for many wind regions in the ASCE7-05. Further work is required on the load sharing between connections to make a definitive statement; such work in on-going at the ‘Three Little Pigs’ Project at UWO. Acknowledgements M.J. Morrison gratefully acknowledges scholarship support from NSERC Canada. G.A. Kopp gratefully acknowledges the support provided by the Canada Research Chairs program. The authors gratefully acknowledge the help provided by Mr. Mitchell Cuddie in the development of the toe-nail rig. Equipment for this project was provided through grants obtained from the Canada Foundation for Innovation, Ontario Innovation Trust, and the University of Western Ontario. References [1] Sparks PR, Schiff SD, Reinhold TA. Wind damage to envelopes of houses and consequent insurance losses. J Wind Eng Ind Aerodyn 1994;53:145–55. [2] Kopp GA, Morrison MJ, Kordi B, Miller C. A method to assess peak storm wind speeds using detailed damage surveys. Eng Struct 2011;33:90–8.
[3] Cheng J. Testing and analysis of the toe–nailed connection in the residential roof-to-wall system. Forest Prod J 2004;54:58–65. [4] Reed TD, Rosowsky DV, Schiff SD. Uplift capacity of light-frame rafter to top plate connections. J Archit Eng 1997;3:156–63. [5] Riley MA, Sadek F. Experimental testing of roof to wall connections in wood frame houses. National Institute of Standards and Technology. NISTIR 6938. 2003. [6] Shanmugam B, Nielson BG, Prevatt DO. Statistical and analytical models for roof components in existing light-framed wood structures. Eng Struct 2009; 31:2607–16. [7] He M, Lam F, Foschi RO. Modeling three-dimensional timber light-frame buildings. J Struct Eng 2001;127:901–13. [8] American Forest & Paper Association, ASD/LRFD national design specification for wood construction. Washington (DC); 2005. [9] Kopp GA, Morrison MJ, Gavanski E, Henderson D, Hong HP. The ‘Three Little Pigs’ Project: hurricane risk mitigation by integrated wind tunnel and fullscale laboratory tests. Nat Hazard Rev 2010;11:151–61. [10] Kopp GA, Surry D, Mans C. Wind effects of parapets on low buildings: part 4. Mitigation of corner loads with alternative geometries. J Wind Eng Ind Aerodyn 2005;93:31–59. [11] Kopp GA, Oh JH, Inculet DR. Wind-induced internal pressures in houses. J Struct Eng 2008;134:1129–38. [12] Ho TCE, Surry D, Morrish D, Kopp GA. The UWO contribution to the NIST aerodynamic database for wind loads on low buildings: part I. Archiving format and basic aerodynamic data. J Wind Eng Ind Aerodyn 2005;93:1–30. [13] Surry D, Sinno RR, Nail B, Ho TCE, Farquhar S, Kopp GA. Structurally-effective static wind loads for roof panels. J Struct Eng 2007;133:871–85. [14] Rosowsky DV, Reinhold TA. Rate-of-load and duration-of-load effects for wood fasteners. J Struct Eng 1999;125:719–24. [15] Sutt Jr EG. The effect of combined shear and uplift forces on roof sheathing panels. Ph.D. thesis. Clemson (South Carolina, US): Clemson University; 2000. [16] ASCE 7-05. Minimum design loads for buildings and other structures. Reston (Virginia): American Society of Civil Engineers; 2006. [17] St. Pierre LM, Kopp GA, Surry D, Ho TCE. The UWO contribution to the NIST aerodynamic database for wind loads on low buildings: part 2. Comparison of data with wind load provisions. J Wind Eng Ind Aerodyn 2005;93:31–59.