Introduction
Cross-laminated timber (CLT) is an innovative engineered wood product that consists of at least three orthogonal laminated layers bonded with structural adhesives. Compared to traditional light-frame wood walls and floors, CLT exhibits relatively high in-plane and out-of-plane strength and stiffness properties, improved dimensional stability, adequate fire resistance, and satisfactory thermal and sound insulation performance when integrated with a holistic design solution (
Karacabeyli and Douglas 2013). Therefore, use of CLT has gained popularity in Europe, Japan, North America, Australia, and New Zealand for residential and commercial building applications (
van de Lindt et al. 2019). When CLT is used within a lateral force-resisting system (LFRS), the most common applications include its use as floor diaphragms and shear walls. CLT panels are often combined with posttensioning steel cables/rods to form a rocking wall system, which provides self-centering ability after a seismic event. A comprehensive review on posttensioning of timber members and structures, including CLT, was conducted by Granello et al. (
2020). In addition, several testing programs reported on the damage typically observed in rocking wall systems, which can include wall toe compression damage (
Blomgren et al. 2019;
Mugabo et al. 2021;
Hashemi et al. 2017). As indicated in Busch et al. (
2021), plastic deformations are permitted for maximum considered earthquake levels of design, which in the case of a CLT shear wall corresponds to wall toe crushing; acceptance limits of wall toe crushing can be found in Busch et al. (
2021) for CLT panels used in posttensioned rocking shear wall applications. In certain cases, to limit the wall toe damage in compression, especially in high seismic regions, wall toe reinforcement can be used. Applications of CLT LFRS are not limited to design of new buildings as several retrofit and rehabilitation solutions have been proposed for existing buildings (e.g.,
Engleson 2019). In all the solutions mentioned previously, it is crucial to understand the compressive behavior of the wall toe.
The compressive stress–strain relationship of CLT is required to design and model CLT walls, especially rocking walls. A typical stress–strain model for CLT in compression is the elastic-perfectly-plastic (EPP) material model, which does not include the strength degradation behavior of the material. Thus, in studies that adopt the EPP model, the force-displacement response and moment-rotation models are not able to capture post-peak loss in strength observed in experiments (
Akbas et al. 2017). To improve the accuracy in modeling, a fundamental understanding of the performance of CLT in parallel-to-major strength direction is imperative.
There have been several studies on the parallel-to-grain compression behavior of CLT panels (
Ganey et al. 2017;
Wei et al. 2019 He et al. 2020). Ganey et al. (
2017) tested three-layer and five-layer CLT specimens in three directions to obtain material properties for CLT panels. Ten three-layer and two five-layer specimens were tested in the primary strength direction. The three-layer specimens were 278 mm tall, and their gross cross-section area varied from
to
. Meanwhile, the five-layer specimens were 511 mm tall, and the cross-section area ranged from 28,650 to
. Based on the testing results, the modulus of elasticity and yield stress did not vary with the ratio of cross-sectional area to the height or number of layers. Wei et al. (
2019) investigated the compression behavior of Canadian western hemlock CLT panels by testing three-layer specimens with the cross-section dimensions of 105 mm by 105 mm and lengths of either 210 mm or 420 mm. The effects of specimen lengths on the compression properties of the specimens were analyzed by a one-way analysis of variance (ANOVA). Results show that the length of specimens did not significantly affect the compression strength and modulus of elasticity. Although the size effect did not considerably influence the results in these studies, the dimensions of specimens were small and thus do not represent the structural sizes of CLT walls as used in real world projects. Wei et al. (
2019) utilized experimental stress–strain curves to validate three axial compression stress–strain models. Wei et al. (
2019) defined 60% of the peak stress as the stress at the elastic limit (
), while the failure stress (
) was a post-peak stress at 90% of the peak. Most testing standards for wood-based materials [e.g., ASTM D4761 (
ASTM 2019b) and Eurocode EN 408:2003] provide guidance for determining the compression strength (i.e., the ultimate stress
), but not for yield stress and failure stress. In the standards ASTM D4761 (
ASTM 2019b), only the characteristic causing failure and its location are recorded. Therefore, there is a lack of definition of compressive yield stress and failure stress for wood-based materials in testing and no consensus on the definition of these stress points for modeling.
In terms of modeling the axial, unidirectional compression stress–strain behavior of CLT, Ganey et al. (
2017) utilized an elastic-perfectly plastic material model. The yield stress was taken as the maximum stress the specimen achieved before losing its load-carrying capacity. Wei et al. (
2019) developed three models for describing the wood axial compression stress–strain relationship (Fig.
1). A stress–strain curve can be divided into three segments, consisting of an elastic segment (or segment I), hardening segment (or segment II), and softening segment (or segment III). Segment I represent the elastic range of the stress–strain curve and it is limited by yield point
. The hardening segment indicates the range beyond the yield point up to the ultimate point
. The post-peak segment of the stress–strain curve is labeled as the softening segment and defined by ultimate point and failure point
. All models share the same linear function for segment I and use either a linear function or quadratic function to describe segments II and III. The parameters of quadratic functions are determined based on a system of three equations.
The main objective of this paper is to characterize the stress–strain relationship of CLT wall panels when loaded in compression and how different confinement mechanisms provide compression reinforcement for the CLT panels. The paper presents an experimental program developed to characterize the compressive behavior of a series of large-scale CLT panels. The large-scale physical tests were driven by the need for testing data on the structural performance of building subassemblies of an award-winning project (
Framework Portland Project 2017) and to increase the state of knowledge of mass-timber systems for building applications for broader use around the world. The panels were loaded under a displacement-controlled test beyond the peak load until the strength loss was greater than 20% of the peak load. In this study, the stresses derived from compression load were based on the gross cross-section area; therefore, they correspond to as gross stresses. The failure point in this study was defined as a post-peak point where its stress was 80% of the ultimate stress. In addition, the effect of reinforcements on the compression behavior is also analyzed with the integration of two reinforcement types that were subjected to the same testing protocols. The specimens with and without reinforcements were tested to characterize various damage points, which are represented by different stress–strain points, including: (1) point at the onset of damage (through visual observation and instrumental observation) (2) point of initiation of strength degradation, (3) point at which buckling occurs, (4) point at which 25.4 mm delamination was observed, and (5) failure point. Lastly, three axial compression stress–strain models are presented, calibrated, and compared.
Materials and Methods
Specimen Description
Twelve CLT panels were procured from a commercial manufacturer for this testing program. The panels were 457 mm wide, 175 mm thick, and 1,524 mm long (Fig.
2). The CLT panels used in the specimens are E1 grade per ANSI PRG 320-2019 (
ANSI/APA 2019), including machine stress rated (MSR) spruce-pine-fir (SPF) longitudinal layers and No. 3 SPF transverse layers. Before shipping, a water repellent was applied to the end grains in the fabrication shop. The average moisture content of the as-received panels was 15.5%, measured by a pin-type moisture meter. The moisture content and dimensions of each specimen are shown in Table
1.
A total of twelve panels were tested. Six (6) panels, designated as type A [Fig.
2(a)], were tested without any reinforcement to provide the baseline values. Two reinforcement types, designated as types B and C, as shown in Figs.
2(b and c), respectively, were tested with reinforcements installed to prevent buckling of the fiber ends and provide confinement at the ends. Three specimens were tested for each of the reinforcement panel types.
The intent of the reinforcement was to enhance the post peak performance of the panels in compression. The application of the reinforcements was performed at the laboratory by specialized crewmembers. The type B panels were reinforced using ASSY VG CSK
fully threaded screws (
MTC Solutions 2020) drilled on the surface of the panels penetrating through the thickness. The type C panel reinforcement consisted of welded steel plates forming a U-section installed at either ends of the panel. Each U-section consisted of two (2)
side plates and one (1)
end plate, which were connected using 6 mm fillet welds, as shown in section D-D in Fig.
2(d). The CLT panels were not rounded at the edges. The plates consisted of Grade 50 [ASTM A572 (
ASTM 2021)] steel. ASSY VG CYL
fully threaded screws (
MTC Solutions 2020) were used for segments with the U-section steel plates, while ASSY VG CSK
screws were installed at the other location. The screws were pre-drilled with 9.5mm diameter holes on the side plates of U-section steel plates. The screw pattern is shown in sections B-B and C-C in Fig.
2(d) and was similar for both type B and type C panels.
Testing Methods
The experimental set-up and the instrumentation plan are shown in Fig.
3. The tests were performed horizontally to ease the material handling and for safety reasons. A digital image of the testing set-up is shown in Fig.
4. The set-up consisted of a 2,224 kN (500-kip) hydraulic actuator with a 508 mm (20-in.) stroke. The hydraulic actuator was braced to prevent uplift that could, in turn, produce out-of-plane loading. A steel bulkhead was used at the other end of the panel, which was bolted to the strong floor using four anchor points. Each anchor point included a set of four 25.4 mm diameter bolts. A steel plate was attached to the steel bulkhead that was connected to the base of the CLT panel to provide an even surface for the panel to react against. The CLT panel was placed on a timber block. On top of the panel and in line with the timber block, a buckling restraint mechanism consisting of C-channels and 19 mm all-threaded rods was installed, which is labeled as
bracing system in Fig.
3. Frictionless surfaces were provided at the CLT panel to the timber block interface and the CLT panel to C-channels interface using two greased high-density polyethylene sheets. The alignment of the panels and the actuator load head were matched and checked using a laser level.
The instrumentation plan included a load cell, linear variable differential transducers (LVDTs), and digital image correlation (DIC) equipment. Two LVDTs were used on either side of the panel to measure the axial compression movement along the centerline of the panel, over a gauge length of 1,219 mm, shown as W-LVDT1 and E-LVDT1 in Fig.
3. To avoid destruction of the sensors these two LVDTs were removed after 80% of ultimate stress was reached post peak. A third LVDT was attached on the east side of a steel loading plate and measured the displacement between the plates at two ends of the specimen, shown as E-LVDT2 in Fig.
3. The range of the LVDTs was
. On either side of the steel plates, LVDTs were used to measure the vertical displacement of the steel plates on the east side (E-LVDT3 and E-LVDT4) as shown in Fig.
3, and W-LVDT2 and W-LVDT3 on the west side (not shown in Fig.
3 for clarity of the figure). These latter four LVDTs had a measurement range of 12.70 mm, and were installed to measure out-of-plane movement to confirm if the out-of-plane motion was small to negligible.
The top surface of the panels was tracked using DIC and the DIC data was analyzed to obtain displacements and strain maps of the top surface. In the DIC setup, two cameras take images of the specimen at a predetermined time interval. The images taken when the specimen is under load are then correlated back to a baseline (initial) image to form a strain and displacement profile of the surface. To ensure that this tracking was performed accurately, the surface was painted black and speckled with white dots to provide a unique pattern and contrast (Fig.
4). In addition, the DIC system was calibrated to capture local deformations.
The actuator load and deflection of the load head, as well as the LVDT deflections, were continuously monitored and logged using an automated data acquisition system. Based on the monitored load and deflection data, load-deflection curves were generated for further analysis.
A monotonic load was applied to the CLT panel using a steel plate between the head of the CLT panel and the actuator. The test was performed under a constant displacement rate of 1.27 mm per minute. The test was continued until the load dropped more than 20% with respect to the measured peak load.
Mechanical Parameter Determination
As mentioned in the “Introduction” section, in this study, the stress is calculated as an engineering stress based on the ratio of the actuator load and the gross cross-section area. For convenience, the word
stress is utilized in this study and refers to the gross stress or the engineering stress. The strains, which were determined from the measurement of W-LVDT1 and E-LVDT1 over a gauge length of 1,219 mm, are referred as west strain and east strain, respectively, and utilized to plot stress–strain curves. The strain derived from DIC data is labeled as DIC strain. Similar to Wei et al. (
2019), a stress–strain curve is divided into three segments, including elastic segment, hardening segment and softening segment.
Several material parameters, including yield point, ultimate point, failure point, gross modulus of elasticity, and stiffness, were estimated from both east and west stress–strain curves and the mean results are presented in this paper. The
yield point is a point on the stress–strain curve where the onset of local damage in terms of strain determined from DIC data, which will be discussed in the following section. The
ultimate point is a point on the stress–strain curve where the stress is maximum. A point on the softening segment when the stress drops to 80% of the peak stress was defined as the
failure point. The failure point was assumed to provide the compressive strain at which the panel suffered
complete failure. This failure point was used as a majority of studies in timber engineering has used this based on ASTM E2126 (
ASTM 2019a). The ultimate point and failure point were directly determined from the stress–strain curve. The
gross modulus of elasticity (MOE) was determined by applying linear regression to a linear segment on a stress–strain curve obtained from the LVDT1 measurements. The points defining the linear segment are 20% and 50% of the ultimate stress for the lower and upper values. The
instantaneous stiffness of each specimen was determined as the first derivative of a stress–strain curve, and the result was plotted as a stiffness-strain curve.
Failure Progression Observations
Failure progression was assessed during testing by tracking several damage indicators. The following damage points were determined, including: (1) point at the onset of damage (through visual observation and instrumental observation), (2) point of initiation of strength degradation (i.e., ultimate point where the stress reached its peak), (3) point at which buckling occurs, (4) point at which 25.4 mm delamination was observed, and (5) failure point.
The ultimate and failure points were determined from the stress–strain results, as indicated in the previous section. Three approaches were used to capture the other damage points (onset of damage, buckling, 25.4 mm delamination) in the failure progression. These approaches were stiffness observation, DIC estimation, and visual estimation. These approaches were utilized independently or combined to determine damage points. The description of these approaches is presented hereafter.
The stiffness observation approach is implemented by determining the tangent stiffness for the specimens. The tangent stiffness was plotted against the strains and two points identified, including the peak stiffness point and the 90% of the post peak stiffness point. The latter was defined as the onset of damage.
DIC observation is based on DIC data recorded during the test at an equal time interval of one second. These data were processed to produce DIC strain maps, which allowed to observe the change in surface strain during testing. These strain maps provide valuable information about the condition of the specimen surface. Quantitatively, the DIC technique can provide instantaneous strain at any point or an average strain over a selected area. Observing the DIC strain maps helps to determine when and where the first surface local strain reached to 3,000 microstrain over an area of 50.8 mm diameter circle, which indicated an onset of local damage in terms of strain and as discussed in the “Test Results and Discussion” section, correlated with average strain at peak load.
The visual estimation approach involves visually observing the damage visible by the naked eye during testing and correlate the observed damage with testing notes and the stress–strain curve. Correlation of visual observations and stress–strain curve was used to determine the onset of damage based on visual observations, the point at which buckling occurs, and the point at which 25.4 mm delamination occurs. In addition, DIC data were combined with visual assessment (photos and videos recorded during testing) and testing notes to report on the failure progression.
Engineering Stress–Strain Models
In this paper, the three models shown in Fig.
1 are adopted, which are all based on models proposed in Wei et al. (
2019) with minor modifications. For example, the hardening segment in Wei et al. (
2019) did not provide a smooth transition from the elastic segment, and therefore, a cubic function is proposed in this study. The details of the stress–strain curve models are discussed as follows.
In all three models shown in Fig.
1, the stress in the elastic segment (
) is expressed as a linear function given by:
where
= gross modulus of elasticity (MOE) determined from the experiment; and
= elastic compressive strain limit at yield point. The stress at the elastic compressive strain limit at the yield point is expressed as
.
For the hardening segments (
), a linear expression
and a cubic expression
are considered as shown in Fig.
1 and given by:
where
where,
are strain and stress coordinates corresponding to the ultimate point. For the softening segment (
), either a linear function
or a quadratic function
are used. These are given by:
where
where
= strain and stress coordinates corresponding to the failure point.
The goodness-of-fit of the numerical models to the mean experimental results of each treatment type is determined using the root mean square error (
) and R-squared (
) coefficient of determination, which are given by
where
is the number of data points in the mean experimental stress–strain curve.
is a mean experimental stress corresponding to the
th strain data point (
) and given by
where
is the sample size of a treatment type,
is an experimental stress of the
th sample corresponding to
.
is a predictive value obtained from a stress–strain model corresponding to .
is the mean stress value of all mean experimental data points given by:
Conclusion
A series of tests on large-scale specimens were performed in support of the performance-based seismic design project in the Pacific Northwest in the United States and to improve the understanding of the behavior of mass timber panels under compression. Six bare CLT specimens, three specimens with STS screw reinforcements, and three specimens with STS screws and U-section steel plate reinforcements were tested to characterize the stress–strain behavior of CLT panels under compression up to failure. The onsets of damage, strength degradation, buckling, and 25.4 mm delamination were observed visually and instrumentally by using DIC combined with other instrument results. Results indicate that the engineering values obtained from DIC observations tended to provide a lower bound for onset of damage strains and corresponding stresses, while values obtained from visual inspections provide the upper bound values. Moreover, the tests show that even though both reinforcement types exhibited an increase in the peak load as well as load at failure, the MOE only increased for the type C reinforcement, which is the one that included both STS and U-section steel plates. The treatments also changed the failure progression in the CLT panels, with the failure near the end or right below the buckling restraint, which was due to crushing and localized buckling of wood fibers.
Lastly, three axial compression stress–strain models were developed and calibrated for three specimen types. Model II, with a cubic function defining the hardening segment and a linear function defining the softening segment, fitted best to the experimental data as observed in the and validations.