Content uploaded by Alicja Pace
Author content
All content in this area was uploaded by Alicja Pace on Dec 06, 2023
Content may be subject to copyright.
Experimental and numerical analyses of full-
span floors and component level subassem-
blies for robust design of CLT floors
Alicja Przystup, PhD candidate, School of Engineering, University of Edinburgh, United
Kingdom
Thomas Reynolds, Lecturer, School of Engineering, United Kingdom
Thomas Tannert, Professor, School of Engineering, University of Northern British Co-
lumbia, Canada
Keywords: Robustness, progressive collapse, catenary action, combined loading
1 Introduction
1.1 Robustness of tall timber structures
Tall mass timber structures are becoming increasingly prevalent and, with some now
rising as tall as 81m (Abrahamsen, 2017) it is vital to consider design for structural ro-
bustness and disproportionate and progressive collapse prevention under accidental
actions (Starossek & Haberland, 2012). The Eurocode 1-7 approach (European Com-
mittee for Standardization, 2006) focuses on material independent objective-based
design. An inherent robustness through alternative load paths (ALPs), primarily cate-
nary action, is targeted through introduction of vertical and horizontal ties. This is a
prescriptive approach without the necessary physical basis when introducing novel
construction methods.
Previous experimental work on catenary action in mass timber floors, specifically floors
made of cross-laminated timber (CLT) (Lyu et al., 2020; Mpidi Bita & Tannert, 2019)
suggested that common connections designed to current standard requirements
might not be appropriate for sufficient load redistribution to develop catenary action
due to their limited deformation capacity. Performance-based design methods such as
the robustness index method (Voulpiotis et al., 2021) have been proposed to capture
the physical behaviour of the system, but such methods require experimental research
to provide the necessary input parameters.
1.2 Connection properties under catenary action
After the loss of a load-bearing member, a crucial mechanism called catenary action
comes into play, allowing for the redistribution of load within the structure. A common
scenario, where this mechanism occurs is, when two floor panels are joined over a
support, and that support is lost, as depicted in Figure 1. Under catenary action, the
floors, which typically endure bending and shear forces, experience a combination of
bending and tension. Understanding the effect of this combined loading on mass tim-
ber connections is essential for effective modelling and performance-based design.
Figure 1: The progression of catenary action in a floor with a joint at mid-span
The required parameters to accurately predict the second-order deformation behav-
iour of such floor-to-floor subassembly are: i) rotational and axial stiffnesses of con-
nection and supports, ii) maximum rotation of central connection, and iii) maximum
axial displacement of central connection. The main challenges to develop catenary be-
haviour in timber structural systems are the lack of reliable plastic deformations within
the timber and connections, and the orthotropic nature of the material. The interac-
tion of these parameters with one another influences the loads within this system (pic-
tured in Figure 2) and can only be predicted and with empirical evidence.
Initially, the magnitude of the vertical load defines the angle of rotation and the tension
demand, as shown in Figure 1a. The tension demand decreases with increasing angle
of rotation (Figure 1b). The latter affects the rotational stiffness, the value of which
affects the ability to develop large deformations beyond the elastic limit. Larger axial
stiffnesses of connection and support increase the tension demand and ultimately
might cause connection failure before catenary equilibrium is achieved (Figure 1c). The
angle of rotation and axial stiffness jointly influence the maximum tension developed
in the system. This results in the total horizontal elongation for a given rotation to be
achieved. Additionally, the connection tension capacity directly restricts the tension
level the system can achieve.
Figure 2: Catenary action parameter interaction diagram
1.3 Objectives
The aim of the research presented in this paper was to conduct an experimental inves-
tigation into the mechanical properties of typical floor panel-to-panel Cross Laminated
Timber (CLT) connections and floor systems when subjected to combined bending and
tension. A secondary objective was to numerically evaluate the potential to predict the
full-span test results based on significantly faster and more cost-effective component-
level tests. Such experiments are necessary for verification of the objective-based de-
sign assumptions. Distilling moment rotation behaviour from simple assemblies can be
utilised in performance-based design for defining connector stiffness, strength, and
deformation capacity.
2 Experimental investigations
2.1 Materials
CLT panels used in Setup A were produced by the Construction Scotland Innovation
Centre to order and had two different panel layups. The 5-ply setup consisted of 20-
20-20-20-20mm panels, and the 3-ply setup consisted of 33-34-33mm panels, both
with the total thickness of 100mm, with moisture content averaging 9.8%. The CLT
panels used for tests in Setup B were 5-ply 100 mm thick Binderholz BBS 125 of 20-20-
20-20-20 mm layer thickness and 10% moisture content.
The tested connections are illustrated in Figure 3. Setup A was used for two series of
half lap connections, using a single self-tapping screw (STS) installed at 90-degree with-
out predrilling. The difference between the two series was the panel layup, with the
first series using the 5-ply and the second using 3-ply. Setup B was used to test two
types of joints using the same CLT panels: spline joints and butt joints. Spline joints
used 8x100 mm STS installed through the plywood perpendicular to the surface.
For the butt joints, 8x140 mm STS were installed at 45° at symmetric spacing. The spec-
imens had a width of d = 400 mm and screw spacings of a = 110 mm, b = 100 mm, and
c = 90 mm. This butt joint connection was also tested in full span where test specimens
had a width d = 600 mm and either 4 or 6 screws with spacings of a = 130 mm, b = 260
mm, and c = 110 mm.
In total 74 component-level specimen were tested with 3 connection types: i) spline
joints; ii) half laps with screws in shear, and iii) butt joints with screws in withdrawal.
Each connection type was tested axially to failure, bending only and under combined
loads of 25%, 50% and 75% of the maximum tested axial strength. Subsequently, 20 of
the full-span specimens were tested under constant tension of 15 kN and 30 kN, and
under fixed horizontal displacement with tension increasing as catenary action pro-
gressed.
Figure 3: Joint types: a) half lap joint; b) single surface spline; c) butt joint
2.2 Experimental setup
Two types of tests were performed, component and full-span floors, as illustrated in
Figure 4a) and b), respectively. Both setup A and B were used for the isolated compo-
nent-level experiments, with setup B span L = 0.8m. The full-span tests were performed
in setup B, where the span L = 3m. The component test was developed for span reduc-
tion in combined axial and bending loads, allowing for increased number of tests
thanks to the reduction in the volume of timber used per test.
To adequately represent the combined loading condition under catenary action and to
be able to be enable comparing the results to the full-span tests, several features were
implemented. These features were: i) active application of a secondary horizontal load
allowing for any desired level of horizontal tensile load irrespective of setup geometry
and stiffness; and ii) the elimination of compressive arching. Previous research on ca-
tenary action utilised some level of physical horizontal restraint at the supports fitted
with load cells to impose and measure tension in the system. This only allows for one
tension-moment combination at failure to be investigated, governed by the arbitrary
stiffness of supports and geometry of the experiment. By introducing an active hori-
zontal loading, any combinations of loads can be implemented on both test scales.
Moreover, compressive arching becomes more significant with an increase in depth-
to-span ratio. Using the active load application combined with roller supports which
do not restrain the outward movement, allowed for this effect to be minimised to neg-
ligible system friction. Therefore, the investigation could focus on extreme rotations
and deformations and their effect on the component performance.
Figure 4: Component test experimental setup A (a) and setup B (b)
2.3 Interaction of applied and internal forces
To determine bending moment capacity, a common approach is 4-point bending,
which creates a shear force free zone between the loading points and enables behav-
iour of the connection to be isolated without additional interaction with the above
wall/column element. In building systems however, point loads are often present due
to the wall/column acting as the primary load path for vertical loads and therefore a 3-
point bending test is also a useful experimental tool for investigation of connection
behaviour with this additional interaction. Both approaches are visualised in Figure
5Figure 4.
Figure 5: Free body diagrams for 3-point bending (force shown as solid line) and 4-point bending
(force shown as dashed line)
In a floor assembly with pin supports and a joint at the centre, when the panel's bend-
ing stiffness is significantly higher than the connections', panel deflection is minimal
and rotational rigid body movement can be assumed. Equation (1) allows computing
the moment at the connection for 4- and 3-point bending (Setup A and B, respectively):
(1)
Where is the central pushdown load, is the location of load (equal to span for
the three-point bending value of force lever arm), is the gravity loading of the floors,
is the angle or rotation, represents the tension applied to the connection, and is
vertical displacement at midspan.
The tension utilisation ratio of the connection is the ratio of applied tension load
and the connection axial tension capacity the connection (not under combined load-
ing), see Equation (2). It is crucial to highlight that refers to the tension force be-
tween the ends of the component and does not represent the fastener utilisation.
(2)
A 2D-beam model pictured in Figure 5 shows an equal tension propagated through-
out the entire subassembly. One of the important concepts to understand under the
combined loading is that in connections (such as butt joints) where a substantial lever
arm generates rotational stiffness, there will be an important differentiation between
the applied tension load () and internal connection tension ().
It is possible to predict the value of through analysing a component as a 3D element,
as shown in Figure 6 resulting in Equation (3). In this joint, under combined loading
and given significant fastener axial stiffness, the internal resultant tension at the con-
nection will quickly become larger than the applied load , and therefore it will be
the value dictating the ultimate strength of the subassembly.
Figure 6: Free body diagram of a butt-joint connection illustrating resultant tension
(3)
As this relationship is not as straightforward for other joints, an objective measure such
as tension utilisation is extremely useful to allow for adequate comparison of mo-
ment and rotation capacity under combined loading.
2.4 Results
2.4.1 Influence of tension utilisation on joint performance
The influence of the tension utilisation ratio on the moment and rotation capacity
of the connections was evaluated. Figure 7 shows the interaction diagram of the mo-
ment tension for all tested connections, and Figures 8 shows the maximum rotation
and tension utilisation interaction diagrams. Each connection was tested in 5 test se-
ries – bending, axial and 3 combined load levels at tension utilisation levels equal to
25%, 50% and 75%. For each of the series 95% confidence intervals (CI) were calculated
based on the Extreme Value Distribution Type I (Gumbel, 1948), to account for asymp-
totic behaviour of the upper outliers of the datasets.
The half-lap and the butt joints showed an inversely proportional relationship of the
moment capacity and tension utilisation due to the direct increase of the resultant
tension through increased bending. The 5-ply half-lap joint (Figure 7a) and butt joint
(Figure 7c) showed an almost perfectly linear relationship, while half-laps in 3-ply test
(Figure 7b) showed a decrease of about 25% in its moment capacity until 60% utilisa-
tion, with sharp drop off after that point, attributed to splitting failure occurring in the
transverse layer of the 3-ply panels. In contrast, the spline connection (Figure 7d) ex-
hibited an increase in moment capacity up until about 50% , with steep drop off after
that point. The interaction in this connection cannot be simplified down to two result-
ant point loads but is directly caused by the bending and splitting of the plywood. The
capacity increase may be attributed to the direct interaction between the wall-stub
point load and the spline. Catenary action enables greater overall force and rotation,
facilitating sufficient friction between the wall stub and the connection to transfer
compressive loads. Looking at the rotation capacity interaction curves, 5-ply half lap
(Figure 8b), butt joint (Figure 8c) and spline joint (Figure 8d) show significant increase
until 25% and subsequent drop off, with spline having by far the largest increase. The
3-ply (Figure 8a) is an outlier here and it is also thought to be due to the difference in
the failure modes.
Figure 7: Force interaction curves from Setup A: 3-ply (a), 5-ply (b) half lap joint and Setup B: butt
joint (c) and spline joint (d)
a)
b)
c)
d)
Figure 8: Rotation failure envelope curves from Setup A: 3-ply (a), 5-ply (b) half lap joint and Setup B:
butt joint (c) and spline joint (d)
2.4.2 Full span test results
The component tests used active tensioning to a load level and subsequent vertical
load pushdown. In true catenary action, this tension will gradually increase throughout
the movement progression. The full-span tests aimed to evaluate the impact of com-
bination of loading, progressively increasing the tension versus sustaining the same
level of load throughout.
The load profiles of the full-span load hold samples and passive tension samples are
compared in Figure 9. Both load hold and passive tension yielded comparable combi-
nation of vertical and horizontal forces at failure. However, it is noteworthy that load
hold exhibited significantly higher variability and, in outlier cases, facilitated substan-
tially larger deformations and loads. Therefore, when interpreting load hold results for
the absolute characterisation of individual connection performance, it is important to
exercise caution. However, given the general replicability of load combinations, the
a)
b)
c)
d)
load hold scenario can be effectively employed as a tool for introducing controlled pa-
rameters to investigate their influence on the behaviour of the connections, as was
done in this study. This is crucial for advancing the more general understanding of con-
nection behaviour and its underlying mechanisms.
Figure 9: Load profiles of passive (a) and active (b) load induced catenary activation
The subsequent relative correlation between the tests at the two scales shown in Fig-
ure 10: Correlation between the component and full-span test. The 95% CI were de-
veloped by utilising the method of least squares on the CI values for each separate test
type calculated through EV1, as shown before in Figure 7c. The full-span tests for the
butt joint showed a decline in moment capacity proportional to increase in tension.
Moreover, mapping both test types onto one another shows a good match, and alt-
hough some of the full-span tests showcase a higher moment resistance, the lower
confidence interval does a good job of capturing the bottom boundary of large-scale
test values, which is the more crucial for design.
Figure 10: Correlation between the component and full-span test
a)
b)
2.4.3 Resultant butt joint tension within the connections
The internal resultant tension forces in the connector were calculated according to
Equation (3) for the butt joint component tests. The tension in the connection is a
combination of the moment couple from the opening of the joint and the external
tension from catenary forces. For most of the tension utilisation levels, the maximum
tension force as well as the relating deflection corresponds to the uniaxial tension
tests, as shown in Figure 11.
The uniaxial tension test results can therefore be used as a reliable parameter for fail-
ure checks in a variety of different bi-axial load combinations. The shape of the curves
also remains unchanged when compared to uniaxial tension tests. One notable obser-
vation is the ability to progress through the post peak plateau within the connection,
as seen in the 25% series and to limited extend in the 50%, which resulted in much
higher overall force and deformation capacity (seen in more detail in Figure 12: in Sec-
tion 3.2).
Figure 11: Resultant total tension at the screws in butt joint component tests as compared to the
uniaxial tension test results
3 Numerical investigations
3.1 Overview
Numerical investigations were used to evaluate whether utilising experimental findings
from the small-scale series (component combined tests and uniaxial tension tests) al-
low to predict the behaviours of the larger subassemblies. The model relies on empir-
ically derived stiffness values of connections and was validated with the experimental
results from the large-scale tests. The validated model allows for parametric explora-
tions concerning a wide range of subassembly geometries and other influential factors,
such as the effects of support conditions on failure modes.
The model was built in the ABAQUS software utilising 2D beam elements and assigning
them the cross-sectional properties of the CLT as used in the large-scale tests: 600 mm
wide and 100 mm thick. The model focused on the butt joint, as it was investigated
with most replicates in large scale testing and therefore the modelling results can be
compared to the largest experimental database.
3.2 Spring model
The main challenge of accurate modelling is reconciling the interactions that the pa-
rameters have on one another, mainly the rotational spring stiffness. This stiffness un-
der combined loading is nonlinear; its value along with maximum moment capacity
varies with change in the tension utilisation . This change is illustrated in Figure 12,
which shows a comparison between force displacement and moment rotation curves
of tests representative of their series.
Figure 12: Vertical force displacement curve (a) and moment rotation curve (b) of the component
tests at of 0, 25, 50 and 75%
This parametric change required developing the two spring models shown in Figure 13.
The axial stiffness of the fastener and the maximum tension force remain unchanged
regardless of the overall maximum tension imposed on the connection (c.f. section
2.4.3). Thus, modelling the axial spring along with compressive point at a lever arm
seen in the connection (Figure 13a) inherently accounts for rotational stiffness changes
and could be used in all tension loading history cases. Figure 13b shows the more clas-
sic representation of connection behaviour, where the rotation and axial behaviour are
treated as separate parameters. Its benefit is the simplicity, not requiring modelling a
contact point; however, it is only representative of a scenario under specified tension,
which is uncommon in real catenary activation. Both spring models were used in the
following analysis to investigate their applicability.
Figure 13: Connection spring models: a) axial spring with lever arm; b) series of rotational and axial
non-dimensional spring
3.3 Results
Modelling the geometry and loading conditions of the larger setups can be predicted
using the stiffness specification of the connector directly from the processed compo-
nent test data, as shown in Figure 14. Both models closely represent the test data, with
model a) – the lever armed axial spring – showing higher stiffness and model b) – series
of rotational and axial spring – showing lower stiffness. The governing failure criterion
for model a) was reaching the ultimate tension force of the axial spring and for model
b) it was the maximum rotation based on the values presented in Figure 8c).
Both models closely predict the displacement under gravity loading. Model b) initially
provides better results, with the stiffness accuracy dropping off with the increase of
mid-span displacement. The failure points in model b) better align with the experi-
mental data. For model a), the idealised lever arm results in a higher stiffness which
results in underestimation of the strength. In reality, the top contact point formation
in failed specimen has consistently shown a contact slip or approx. 10 mm. Therefore,
an empirical factor for lever arm of 0.8 was introduced, resulting in adjusted model
a*). This adjustment allowed for a much better, however still conservative, prediction
of the failure point and system stiffness.
The rotational spring model could also be improved for use under variable tension,
through introducing parametric variables. This however introduces another level of
complexity and might not be replicable for many design scenarios. The discrepancies
between both models and the test results are attributed to the non-linearity of stiff-
nesses; however, the linear approximations used along with the outlined failure crite-
ria are adequate predictors of the behaviour.
Figure 14: Comparison of ABAQUS and test results for a 15kN load hold
4 Conclusions
Moment-tension as well as maximum rotation-tension interaction curves were derived
for 3 types of common CLT joints. For joints with significant rotational stiffness, a neg-
ative correlation was shown between the tension level and moment capacity. Supple-
mentary rotation curves indicated that the best catenary can be achieved with around
25% of tension utilisation (normalised to maximum uniaxial tension). The spline con-
nection exhibited the greatest benefit from catenary action; however, it did not offer
significant bending stiffness in normal conditions.
A positive correlation was found between the full span test results of two horizontal
load application methods: active load hold and passive horizontal restraint. As these
represent two extremes of boundary stiffness resulting in very different tension his-
tory, it can be asserted that given the same final force combinations, similar failure is
to be expected irrespective of the loading history. This allows for a higher flexibility in
methods of testing. Moreover, using component-level bending and tensile test results
as input for numerical modelling has proven to be a dependable technique for predict-
ing the behaviour of CLT floor subassemblies subjected to catenary action.
The emphasis on performance-based progressive collapse prevention holds significant
importance for the future generation of timber design standards. The research pre-
sented herein provides an empirical foundation and presents an economical test ap-
proach, known as the component test method, for effectively measuring these param-
eters in connections. The recommendation for streamlining the data gathering process
for the performance of mass timber subcomponents is to standardise the component
test method and data presentation in the form of interaction curves as presented
above.
References
Abrahamsen, R. (2017). Mjøstårnet-Construction of an 81 m tall timber building. Inter-
nationales Holzbau-Forum IHF, 12.
European Committee for Standardization. (2006). Eurocode 1 - Actions on structures -
Part 1-7: General actions - Accidental actions BS EN 1991-1-7:2006 (Vol. 1, p. 66).
Gumbel, E. J. (1948). Statistical theory of extreme values and some practical applica-
tions: a series of lectures.: Vol. Vol. 33. US Government Printing Office.
Lyu, C. H., Gilbert, B. P., Guan, H., Underhill, I. D., Gunalan, S., Karampour, H., & Ma-
saeli, M. (2020). Experimental collapse response of post-and-beam mass timber
frames under a quasi-static column removal scenario. Engineering Structures, 213,
110562. https://doi.org/10.1016/j.engstruct.2020.110562
Mpidi Bita, H., & Tannert, T. (2019). Experimental Study of Disproportionate Collapse
Prevention Mechanisms for Mass-Timber Floor Systems. Journal of Structural En-
gineering, 146(2), 04019199. https://doi.org/10.1061/(ASCE)ST.1943-
541X.0002485
Starossek, U., & Haberland, M. (2012). Robustness of structures. Int. J. Lifecycle Per-
formance Engineering, 1(1), 3–21. https://doi.org/10.1504/IJLCPE.2012.051279
Voulpiotis, K., Köhler, J., Jockwer, R., & Frangi, A. (2021). A holistic framework for de-
signing for structural robustness in tall timber buildings. Engineering Structures,
227, 111432. https://doi.org/10.1016/j.engstruct.2020.111432
