(Asce) ST 1943-541X 0002219
(Asce) ST 1943-541X 0002219
Abstract: The performance of post-tensioned rocking cross-laminated timber (CLT) panels depends on the ability to maintain the post-
tensioned force in the strand. This post-tensioned force may change over time due to the creep behavior of wood, which, in turn, is a function
Downloaded from ascelibrary.org by Sergio Barreiro on 01/07/19. Copyright ASCE. For personal use only; all rights reserved.
of time and moisture content in the CLT panels. In this study, a numerical moisture content diffusion model was developed to predict moisture
content migration through CLT panels when the ambient relative humidity changes. Fick’s second law and the moisture content diffusion
coefficients were applied to derive the differential diffusion equation for use in a numerical model. A four-element creep model was included to
estimate the creep deformation of CLT panels over time under an axial load with changing environmental conditions. Data from a series of
moisture content and creep tests under different configurations and environmental conditions were used to calibrate the proposed moisture
content diffusion and creep model for CLT panels. The moisture content diffusion model was calibrated for two relative humidity steps,
50%–70% and 70%–90%. Then, a new creep model at material level that considers the change of moisture content in CLT panels was
introduced. The viscoelastic parameters and mechano-sorptive constants were recommended for the creep model based on creep test data.
Axial strain in CLT panels varied under 2% when ambient relative humidity switched between 50% and 70%. The axial strains in CLT
panels with three layers were more sensitive to variations in surrounding relative humidity than those of CLT panels with five layers.
DOI: 10.1061/(ASCE)ST.1943-541X.0002219. © 2018 American Society of Civil Engineers.
Author keywords: Cross-laminated timber (CLT); Moisture diffusion; Creep model; Wood panel; Mass timber.
∂ ∂Φ ∂ ∂Φ ∂ ∂Φ ∂Φ where MC = specimen moisture content; W 1 = weight of the speci-
Rx þ Ry þ Rz ¼ ð2Þ men measured during the tests; and W 0 is the oven-dry weight of
∂x ∂x ∂y ∂y ∂z ∂z ∂t
the specimen.
where Rx , Ry , Rz = diffusion coefficients (or MC capacities) in the The purpose of these tests was to estimate the equilibrium MC
x-, y-, z-directions, respectively. of CLT under different RH conditions. The results of these tests are
Using a finite-element method formulation, Eq. (2) can be con- listed in Table 1.
verted into the weak form for finite element approximation as
(Thompson 2005)
Moisture Content Diffusion Tests
½KfΦg þ ½CfΦ̇g ¼ 0 ð3Þ In the numerical model described previously, the diffusion coeffi-
where Φ̇ represents the nodal point values for the derivative of Φ cients Rx , Ry , and Rz of CLT panels need to be experimentally
with respect to time evaluated. A series of tests on MC diffusion were conducted to es-
timate these coefficients. In order to understand the effects of panel
∂Φ size on the diffusion coefficients, the CLT specimens used in the
Φ̇ ¼ ð4Þ
∂t tests were divided into three groups: 60.9 × 60.9 cm (2 × 2 ft)
three-layer panels, 30.4 × 60.9 cm (1 × 2 ft) three-layer panels,
and [K] and [C] are calculated by integrals: and 30.4 × 30.4 cm (1 × 1 ft) five-layer panels. The dimensions
Z · and number of specimens in each group are shown in Table 2.
½K ¼ ½N 0 T ½R½N 0 dV and ð5Þ In buildings, CLT panels mainly absorb/desorb moisture
V
through surfaces exposed to the environment, because panel edges
Z · have a small area and are often coved by adjacent panels or walls.
½C ¼ ½NT bNcdV ð6Þ For this reason, the experiments in this study focused on the
V
estimation of the moisture diffusion coefficient in the direction per-
where [N] is the element shape functions row matrix; [N 0 ] is the pendicular to the panel surface. To achieve this restriction, the panel
derivative of the element shape function matrix versus space coor- edges of each specimen were covered by aluminum tape, and the
dinates (Thompson 2005); [R] is the material properties matrix surfaces were exposed to controlled environmental conditions
(or the matrix of diffusion coefficients); and V = volume. The cen- during the tests, as shown in Fig. 3.
tral difference approximation was used to solve MC at each time The CLT specimens were then placed in environmental cham-
step using Eq. (3). At each time step, the MC was calculated as bers in which the temperature and RH were controlled. The envi-
ronmental chamber at the University of Alabama (UA) was built to
½CPKfΦgtþΔt ¼ ½CMKfΦ̇gt ð7Þ control temperatures ranging from 5°C (41°F) to 60°C (140°F) and
RH ranging from 20% to 98%. In this study, the chamber was set at
where a constant temperature of 22.8°C (73°F) and three levels of RH of
Δt 50%, 70%, and 90%, respectively. At the beginning, the environ-
½CPK ¼ ½C þ ½K and ð8Þ mental chamber was set at a temperature of 22.8°C (73°F) and a RH
2
of 50%. The CLT specimens were placed in the chamber until an
Δt equilibrium of MC was achieved (there was not significant change
½CMK ¼ ½C − ½K ð9Þ of weight in each specimen). The RH of the chamber was then
2
switched to 70%, and the temperature was maintained at 22.8°C
and Δt = time step. (73°F). The change in weight of each specimen was measured
Outcomes of the MC diffusion model for a five-layer spruce- to observe the moisture absorbing process in CLT. After switching
pine-fir (SPF) CLT specimen are shown in Fig. 2. The dimensions the RH, the CLT specimens absorbed moisture at a higher rate in
of the specimen were 17.1 × 60.9 × 60.9 cm (6.75 × 24 × 24 in:) the first two days because of the large difference in MC between
The original MC in the specimen was 10.6% and the MC in the new the surface and the inside of the CLT specimens. Over time, the
environment was 16.1%. absorption process slowed down as the MC inside the specimens
From Fig. 2, one can see that when the MC at the surface of the increased. In order to have a smooth curve of MC change in the
CLT panel changes to the higher value of 16.1%, the MC starts to CLT specimens, the weights of the CLT specimens were measured
travel inside the panel over time, as expected; this leads to higher every 6 h for the first two days; after that, the interval was changed
MC inside the panel. to every 12 h until the specimens were believed to be close to new
Fig. 2. Moisture content migration in a CLT panel with time: (a) day 1; (b) day 10; (c) day 19; and (d) day 28.
equilibrium conditions. The RH was then switched to a new level Because of the nature of CLT materials, moisture absorbing and
and the same data collection procedure was used for the rest of the desorbing rates may vary; this phenomenon was taken into consid-
MC tests. The tests were conducted for both increasing RH from eration by using two separate coefficients for the two processes.
50% to 70% to 90% and for decreasing humidity from 90% back to Fig. 4 shows the fitting data for one of the tested specimens for
70%. During the tests, the weight of each specimen was measured both absorbing and desorbing moisture (other specimens can be
using a scale that has a resolution of 0.002 kg (0.03% of the small- seen in the Appendix). In Fig. 4, Rxai is the diffusion coefficient of
est panel mass). This high-resolution scale allowed the accurate the absorbing process, and Rxd is the diffusion coefficient of the
capture of the change in MC of each specimen. desorbing process. The boundary condition of the MC at the sur-
To estimate the MC of each specimen, method A (oven-drying) face for different RHs is represented by MCi.
of ASTM D4442 (ASTM 2015b) was applied. All CLT specimens Since wood is a natural material, the moisture diffusion coeffi-
were dried in an oven after the test, and their dry weight was cients may vary from specimen to specimen. The final coefficient
measured following the specifications of ASTM D4442 (ASTM for a group was estimated by performing a statistical analysis to
2015b). The dimensions and dry weight of each specimen are obtain the mean and variance value of coefficients in specimens.
shown in Table 3. The MC of the samples could then be calculated Table 4 shows the MC diffusion coefficients for the three specimen
by Eq. (10). groups for different RH intervals. The MCs at the surfaces of spec-
imens predicted by numerical models at 70% RH and 90% RH
were also included In this table.
Fitting Moisture Content Diffusion Coefficients Statistical distributions were fitted to each population of three
In this study, a FEM subroutine was developed to estimate the MC MC diffusion coefficients; the lognormal distributions were found
diffusion coefficients. By adjusting the diffusion coefficients and to be adequate representations of the distributions of all coeffi-
the boundary conditions in the numerical model, the weight of each cients. Fig. 5 illustrates the lognormal distribution plots for the MC
CLT specimen was fitted to the experimental data from the tests. diffusion coefficient in different sample groups at different RH step
Fig. 3. Specimens in (a) Group 2; and (b) Group 3 for diffusion coefficient tests.
Table 5 shows that the absorption coefficients of the CLT where εðtÞ = total strain at time t; σ = constant applied stress; K e ,
specimens varied between 0.183 and 0:217 cm2 · day−1 0
K k , μk , μv , and μms are model constants; Bω is a constant associated
(0.0284–0.0336 in.2 · day−1 ) for the RH range between 70% with the time required to achieve moisture equilibrium and is de-
and 90%. These values were about 4–5 times greater than the pendent on the size of the sample; jΔwj ¼ ωe − ωi ; the eventual
value reported in a study by Tong (1987) on spruce, which was equilibrium MC in the new environment is ωe ; and ωi is the initial
0.045 cm2 · day−1 (0:007 in.2 · day−1 ) in the transverse direction MC in the original environment. In Fridley et al. (1992a), the two
(a step change in RH from 65% to 80%). Note that CLT panels are hygrothermal variables ω and θ were introduced as
made of several layers of small lumber, and the gaps between two
adjacent pieces of lumber in the same layer are not glued, which M − M0
ω¼ and ð12Þ
allows moisture to migrate easily into deeper layers. Therefore, the M0
larger absorption coefficients of CLT in this study compared to
solid timber in the Tong (1987) study can be considered reason- T − T0
able. Also, from Fig. 6(d), the average absorption coefficient θ¼ ð13Þ
T0
measured with a RH of 70%–90% was about 2–3 times larger than
that measured with a RH of 50%–70%. This means that the ab- where ω is a relative MC; M = actual MC; M o = reference MC; θ is
sorption coefficient was not constant as the MC changed in the a relative temperature; T = actual temperature; and T o = reference
CLT. Therefore, in order to apply Fick’s law in the prediction of temperature. The quadratic functions of ω and θ were assumed for
Table 4. Moisture diffusion coefficients and moisture contents at the boundary of CLT panels for different relative humidity intervals
Rx (cm2 · day−1 ) MC (%)
Dimensions (cm) Specimen 50%–70% RH 70%–90% RH 90%–70% RH 70% RH 90% RH
Group 1 L-1 0.0742 0.171 0.226 14.5 18.5
9.91 × 60.9 × 60.9 L-2 0.0774 0.232 0.284 14.8 18.8
(3 layers) L-3 0.0581 0.158 0.232 14.4 18.4
L-4 0.0581 0.158 0.240 13.8 17.5
L-5 0.0684 0.148 0.208 13.6 17.7
L-6 0.0600 0.229 0.294 14.4 18.0
Group 2 L-7 0.0548 0.197 0.313 13.5 16.9
9.91 × 30.4 × 60.9 L-8 0.0613 0.229 0.288 13.9 17.3
(3 layers) L-9 0.0568 0.203 0.277 13.9 17.5
L-10 0.0613 0.210 0.284 13.6 17.2
L-11 0.0600 0.216 0.255 13.7 17.3
L-12 0.0626 0.242 0.306 13.8 17.5
Group 3 S-1 0.106 0.194 0.245 13.4 17.4
16.9 × 30.4 × 30.4 S-2 0.103 0.187 0.224 13.5 17.4
(5 layers) S-3 0.113 0.206 0.245 13.5 17.5
S-4 0.100 0.223 0.271 13.8 17.6
S-5 0.0968 0.239 0.299 13.7 17.2
S-6 0.110 0.223 0.292 13.6 17.5
S-7 0.116 0.201 0.258 13.3 17.3
S-8 0.119 0.221 0.239 13.6 17.4
S-9 0.116 0.231 0.265 13.5 17.3
S-10 0.132 0.252 0.277 13.7 17.8
S-11 0.106 0.212 0.226 13.3 17.3
Fig. 5. Cumulative distribution plot of Rx for different groups (1 in: ¼ 2.54 cm): (a) Rxa2 for Specimens L1–L6 at 70%–90% RH;
(b) Rxd for Specimens L1–L6 at 90%–70% RH; (c) Rxa2 for Specimens S11–S23 at 70%–90% RH; and (d) Rxd for Specimens S11–S23 at
90%–70% RH.
Table 5. Moisture content diffusion coefficients and equilibrium moisture content for different relative humidities and specimens
Groups Number of specimens Parameters RH (%) Mean Variance Distribution
2 −1
(1) 3 layers 6 Rx (cm · day ) 50–70 0.066 0.000072 Lognormal
70–90 0.183 0.001348
9.91 × 60.9 × 60.9 90–70 0.248 0.001132
MC (%) 70 14.2 0.202
90 18.1 0.251
(2) 3 layers 6 Rx (cm2 · day−1 ) 50–70 0.059 0.000009
70–90 0.216 0.000277
9.91 × 30.4 × 60.9 90–70 0.287 0.000450
MC (%) 70 13.7 0.0234
90 17.3 0.0428
(3) 5 layers 11 Rx (cm2 · day−1 ) 50–70 0.110 0.000069
70–90 0.217 0.000343
16.9 × 30.4 × 30.4 90–70 0.259 0.000694
MC (%) 70 13.5 0.0311
90 17.4 0.0392
All specimens 23 Rx (cm2 · day−1 ) 50–70 0.086 0.000699
70–90 0.208 0.000886
90–70 0.264 0.000948
MC (%) 70 13.8 0.146
90 17.5 0.193
Fig. 6. Distribution of Rx for all groups at different relative humidity levels (1 in: ¼ 2.54 cm): (a) Rxa1 for all groups at 50%–70% RH; (b) Rxa2 for all
groups at 70%–90% RH; (c) Rxd for all groups at 90%–70% RH; and (d) Rx for all groups for different RH changes.
Fig. 7. Distribution of equilibrium moisture content for all groups: (a) at 70% RH; and (b) at 90% RH.
Fig. 9. Creep test setup for a five-layer, 15% fc0 specimen: (a) schematic location; and (b) ready specimen.
Fig. 10. Fitted model constants for a three-layer, 10% f c0 specimen (1 in: ¼ 2.54 cm): (a) data from DEMEC strain gauges; and (b) data from strain
gauges.
variance than five-layer CLT panels over three months. At this clearer patterns, as can be seen in Fig. 12 and in the Appendix.
stress level, creep behavior takes a longer time to reach the creep Even though the MOE has consistent mean values between the
deformation limit, especially for five-layer CLT panels. For this three-layer and five-layer CLT panels at different stress levels,
reason, fitted parameters do not seem to be reliable for the long- the creep modulus K k and the viscous damping creep parameter
term prediction of creep deformation, especially in five-layer CLT μk have a large variance within each test group and between dif-
panels—at least for the time interval that was examined. To have ferent CLT thicknesses. This finding illustrates that although
better fitted parameters, longer-duration creep tests are recom- CLT is a good quality-controlled engineering material, prediction
mended for future studies. Some creep tests at this stress level of its creep behavior has high uncertainty. Therefore, reliability
are ongoing in the UA labs in order to achieve a more robust data analysis is needed for proposed design equations related to creep
set. Creep deformation at stress levels of 10% and 15% fc0 showed behaviors.
Fig. 11. Lognormal distribution of elastic constant K e for different groups (1 psi ¼ 6.89 kPa): (a) elastic constant for Group 2; and (b) elastic
constant for Group 4.
Fig. 12. Lognormal distribution of constants for Group 5 (1 psi ¼ 6.89 kPa): (a) elastic constant for K k ; and (b) viscous constant for μk.
Fig. 13. Lognormal distribution for all specimens (1 psi ¼ 6.89 kPa): (a) elastic constant for K k ; and (b) viscous constant for μk.
Kk
μk 1.36 (1012 ) kPa · min 20.2 (1012 ) kPa · min
Fig. 14. Moisture content and axial strain history in a three-layer specimen (1 in: ¼ 2.54 cm): (a) moisture content profile; and (b) axial strain.
Fig. 15. Axial strain history for different CLT configurations (1 in: ¼ 2.54 cm): (a) moisture content profile; and (b) axial strain.
50% and 70% within a four-day period. During this period the tem- 17.1 cm) (3.9 and 6.7 in.), the variance in axial strain was negligible
perature was maintained at 22.8°C (73°F). The axial stress level of under a constant surrounding RH of 50%. When the ambient RH
0.10 f c0 was applied for all tested specimens, which means that the changed from 50% to 70%, the differences in axial strains were sig-
axial forces applied were equal to 145 kN (32.7 kip) and 252 kN nificant, at around 4% (40.6 × 10−6 cm=cm) (16 × 10−6 in:=in:)
(56.7 kip) on three-layer and five-layer CLT panels, respectively. The effect of changing the surrounding RH on axial strain in
Fig. 14 shows the MC profile and changes in the axial strain at three-layer CLT panels was larger than the effect on five-layer
different locations in a CLT specimen. CLT panels [as shown in Fig. 15(b)]. This can be explained by
The effects of panel size on axial strain in CLT panels were also the fact that the five-layer CLT panels are thicker than the three-layer
studied by numerically examining four different CLT panel con- ones; thus, the five-layer CLT panels required more time to absorb
figurations. The results are presented in Fig. 15. moisture from the ambient environment.
As can be seen in Fig. 15, under identical ambient RH profiles
[Fig. 15(a)], the axial strain in various types of CLT specimens with
different configurations showed similar strain patterns [Fig. 15(b)]. Conclusions
The variation in axial strain due to changing ambient RH between
50% and 70% was approximately 1.8% (18.3 × 10−6 cm=cm) In this paper, a numerical model for the prediction of MC migration
(7.2 × 10−6 in:=in:). For CLT specimens with the same size in CLT panels was introduced. A series of tests on MC migration in
(60.9 × 60.9 cm) (24 × 24 in:) but different thicknesses (9.9 and CLT panels were conducted, and the collected data were used to
Ho, T. X., T. N. Dao, S. Aaleti, J. W. van de Lindt, and D. R. Rammer. 2016. Skrzypek, J. J., and R. B. Hetnarski, eds. 1993. Plasticity and creep:
“Hybrid system of unbonded post-tensioned CLT panels and light- Theory, examples, and problems. Boca Raton, FL: CRC Press.
frame wood shear walls.” J. Struct. Eng. 143 (2): 04016171. https://doi Thompson, E. G. 2005. Introduction to the finite element method: Theory,
.org/10.1061/(ASCE)ST.1943-541X.0001665. programming, and applications. Hoboken, NJ: Wiley.
Hoyle, R. I., M. C. Griffith, and R. Y. Hani. 1985. “Primary creep of Tong, L. 1987. Moisture transport in wood and wood-based panels. A pre-
Douglas-fir beams of commercial size and quality.” Wood Fiber Sci. study of sorption methods. Rapport I8712078. Stockholm, Sweden:
17 (3): 300–314. TrateknikCentrum.
ICC (International Code Council). 2009. International building code. USDA, Forest Service, Forest Products Laboratory. 2010. Wood handbook:
Washington, DC: ICC. Wood as an engineering material. Madison, WI: USDA, Forest Service,
Karacabeyli, E., and B. Douglas, eds. 2013. CLT handbook: Cross lami- Forest Products Laboratory.
nated timber—US edition. Pointe-Claire, QC, Canada: FPInnovations. Wacker, P. J. 2009. “Performance of stress-laminated timber highway
Mohammad, M., S. Gagnon, B. K. Douglas, and L. Podesto. 2012. “Intro- bridges in cold climates.” In Proc., 14th Conf. on Cold Regions Engi-
duction to cross laminated timber.” Wood Des. Focus 22 (2): 3–12. neering. Reston, VA: ASCE.
Nakai, T., and P. Grossman. 1983. “Deflection of wood under intermittent Yazdani, N., E. Johnson, and S. Duwadi. 2004. “Creep effect in structural
loading.” Wood Sci. Technol. 17 (1): 55–67. https://doi.org/10.1007 composite lumber for bridge applications.” J. Bridge Eng. 9 (1): 87–94.
/BF00351832. https://doi.org/10.1061/(ASCE)1084-0702(2004)9:1(87).