Você está na página 1de 9

Reliability assessment of steel scaffold shoring structures for concrete formwork

Hao Zhang
a,
, Kim J.R. Rasmussen
a
, Bruce R. Ellingwood
b
a
School of Civil Engineering, University of Sydney, NSW 2006, Australia
b
School of Civil and Environmental Engineering, Georgia Institute of Technology, Atlanta, GA 30332, USA
a r t i c l e i n f o
Article history:
Received 9 March 2011
Revised 21 November 2011
Accepted 21 November 2011
Available online 26 December 2011
Keywords:
Construction formwork
Loads (forces)
Nonlinear frame analysis
Probability-based design
Scaffold
Shoring
Steel
Structural engineering
Structural reliability
a b s t r a c t
Many failures of concrete structures during construction are traceable to the collapse of formwork shoring
systems. A reliability analysis for typical steel scaffold shoring structures is presented herein utilizing
recent survey data on geometric and mechanical properties of steel scaffold members, and a second-order
inelastic structural analysis model. Published shore load surveys were analyzed to develop the probabilis-
tic models for loads acting on scaffolds during concrete placement. The paper investigates the mode of fail-
ure, the effects of different random variables on the variability of structural strength, and the reliability of
the analyzed scaffold structures. The reliability framework can be used to improve the current working
load limit basis for the design of steel scaffold structures and make scaffold design risk-consistent.
2011 Elsevier Ltd. All rights reserved.
1. Introduction
Steel scaffolds are commonly used in reinforced concrete con-
struction as shoring systems to support formwork. As compared
with other shoring systems such as wood or metal posts, steel scaf-
fold-type systems have the advantages of increased strength and
larger panel sizes. A steel scaffold system normally consists of up-
rights (tubular columns, referred to as standards in industry), led-
gers (horizontal members), braces and adjustable jacks. The
uprights are connected to the ledgers by means of various types of
connections such as cuplok joints or wedge-type joints. Note that
another type of scaffolds commonly used in construction is the ac-
cess scaffolds, which are mainly used to support light to moderate
loads fromworkers, small constructionmaterial and equipments for
safe working space. The access scaffold is not considered in this
study.
Many studies have shown that failures of reinforced concrete
structures that occur during their construction phase are often
traceable to the collapse of formwork shoring systems [1,2]. The
economic and legal consequences of such structural failures can
be disturbing. The Guangxi (China) Medical University library acci-
dent in 2007, in which seven construction workers were killed [3],
is a recent example of a catastrophic failure of steel scaffold
shoring system. Such failures have prompted numerous efforts to
investigate the structural performance of steel scaffolds both
experimentally and numerically [411]. More recently, second
-order inelastic structural analysis (advanced analysis) has been
used to predict the behavior and limit state load carrying capacity
of steel scaffold structures, including material and geometric non-
linearities, initial geometric imperfections, and semi-rigid joint
details [12]. Unfortunately, even with accurate structural models
that have been calibrated by experimental tests, the actual perfor-
mance of steel scaffolds cannot be predicted with certainty
because large uncertainties in structural properties and applied
loads will always be present. Recent advances in structural reliabil-
ity theory permit these uncertainties to be quantied using proba-
bilistic methods, and the safety of scaffold structures to be
rationally assessed in terms of their limit-state probabilities (reli-
abilities). Herein, structural reliability is dened as the probability
that the structure will meet its performance objectives for strength
or serviceability over a given period of time.
A limited number of structural reliability analyses have been
conducted for temporary structures during construction. Gromala
[13] reviewed the reliability requirements for wood-based access
scaffold planks in codes and standards, and related those require-
ments to allowable stresses published by lumber grading agencies.
The statistical information of load and resistance used in the reli-
ability calculation are based on assumptions rather than experi-
mental data. Charuvisit et al. [14] studied risk of access scaffolds
0141-0296/$ - see front matter 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engstruct.2011.11.027

Corresponding author. Tel.: +61 2 93513923; fax: +61 2 93513343.


E-mail addresses: hao.zhang@sydney.edu.au (H. Zhang), kim.rasmussen@syd
ney.edu.au (K.J.R. Rasmussen), bruce.ellingwood@ce.gatech.edu (B.R. Ellingwood).
Engineering Structures 36 (2012) 8189
Contents lists available at SciVerse ScienceDirect
Engineering Structures
j our nal homepage: www. el sevi er. com/ l ocat e/ engst r uct
under strong wind by wind tunnel test and eld measurement of
wind loads acting on actual scaffolds. The study focused on the
safety of the ties connecting the access scaffolds to external build-
ing walls. Epaarachchi et al. [15,16] calculated the probability of
structural failure during the construction of typical multistory
reinforced concrete buildings. It was assumed that system failure
occurs when the slab strength (in exure or shear) limit state is
violated. Failure of shores/reshores was not considered in their
study.
At present, industry practice for the design of steel scaffold
structures are based on the working load limit (WLL) philosophy.
The basic WLL design equation has the format
R
n
PF:S:D
n
L
n
1
in which R
n
= nominal system strength computed using the analysis
and design methods accepted by the design code under consider-
ation for the nominal values of the material and sectional proper-
ties, F.S. = factor of safety, D
n
and L
n
are nominal dead and live
loads due to construction. R
n
/F.S. represents the allowable working
load. The construction dead load includes the weight of concrete
and formwork. The construction live loads acting on scaffolds
include the weight of construction personnel, equipment, stacked
material, and an allowance for localized mounding during concrete
placing. The factor of safety, typically ranging from 2.0 to 3.0, is
judgmental in nature. The reliability associated with current steel
scaffold design practice is unknown, making it difcult to extend
scaffold design beyond current practices to meet the emerging
needs of the construction industry.
This paper presents a reliability assessment of a typical steel
scaffold structure designed in accordance with the WLL design for-
mat. The reliability implications for the scaffolds designed by
advanced nonlinear analysis and the customary elastic analysis
are compared. The probabilistic studies utilize the available litera-
ture and data on shore loads, the recent survey data on geometric
and mechanical properties of steel scaffold members, and a sec-
ond-order inelastic structural model. The effect on structural reli-
ability of human error is also discussed.
2. Structural reliability analysis
We consider the ultimate strength of a scaffold structure, in
which the system limit state is dened as failure by overload. Steel
scaffolds are designed to withstand vertical loads as well as the hor-
izontal loads due to lateral concrete pressure or wind, etc. The form-
work is usually restrained externally by the completed part of the
permanent structure, to which the horizontal loads are transmitted.
Thus the design of steel scaffolds is generally governed by the ver-
tical loads. This study only considers the combination of dead and
live load due to construction. The limit-state probability, P
f
, is de-
ned as:
P
f
PR D L 6 0 2
in which P() = probability of the event in the bracket, R = ultimate
structural strength (resistance), D = applied dead load effect, and
L = live load effect. Since the loads vary with time, D + L in Eq. (2)
is interpreted as the maximum total load effect during the construc-
tion process.
Eq. (2) can be evaluated using standard reliability analysis
methods such as Monte Carlo simulation or rst-order reliability-
method (FORM) if the probabilistic information of structural resis-
tance and loads is available. This study assumes that the loads act-
ing on each upright of a scaffold at the point in the process where
they reach their maximum values are perfectly correlated. Thus,
only two random variables are required to describe the dead and
live loads. This assumption makes it possible to analyze the reliabil-
ity efciently using FORM. If the structural resistance and loads are
mutually statistically independent normal variables, the limit-state
probability P
f
is given by
P
f
U
R D L

r
2
R
r
2
D
r
2
L
q
0
B
@
1
C
A Ub 3
where U() is the standard normal distribution function. In this
paper, X
n
, X, r
x
and V
X
denote the nominal, mean, standard devia-
tion and coefcient of variation (COV) for a random variable X,
respectively. The reliability index, b, has been customarily used as
an alternative to P
f
in probability-based limit-states design [17].
For correlated non-normal random variables, a Rosenblatt transfor-
mation may be applied to produce equivalent independent normal
variables for use with FORM. See Ref. [18] for details.
In this study, the probabilistic load models for scaffolds were
obtained from the load survey data published in the literature.
The statistical characteristics of the structural resistance were
derived using simulations and an advanced nonlinear nite ele-
ment analysis (FEA). The simulation studies also provide useful
information on the relative signicance of different sources of
uncertainty to the overall variability in structural strength. The
load and resistance data were then synthesized using FORM to
estimate the structural reliability.
3. Probabilistic load models for scaffolds
The vertical loads acting on a scaffold can be categorized as
dead and live loads. The dead load is due to the weight of concrete
and formwork. It increases as the concrete placement progresses,
and reaches its maximum at the end of the placement. The live
load consists of the weight of construction personnel, equipment,
stacked material, and the effects of any impact during concrete
placement. The magnitudes of the live loads depend on the stages
of construction, i.e., before, during, and after placement of con-
crete. Fig. 1 shows a typical history of a shore load (a combination
of the dead and live load effects) [19,20, c.f.]. The maximum shore
load typically occurs during the period of concrete placement. As
will be described in Sections 3.1 and 3.2, the dead load is the dom-
inant component of the shore load during concrete placement.
Therefore, the maximum shore load effect occurs when the dead
load is fully developed while the live load assumes its instanta-
neous value, which may be less than the maximum live load
[19,20]. At the end of the placement, the live load effects diminish
because construction activities lessen. The scaffold then supports
only the weight of the slab. The difference between the maximum
shore load and the load just after the casting provides an estima-
tion of the live load effect. The construction live load may increase
signicantly after the pouring due to increased construction
Fig. 1. Typical shore load history.
82 H. Zhang et al. / Engineering Structures 36 (2012) 8189
activity and storage of material. However, the shore loads tend to
decrease as the concrete slab gains partial strength and starts car-
rying part of the applied loads [20,21]. Therefore, the most critical
time for scaffold shoring system is typically during the process of
concrete placement. A survey of falsework collapses [1] has shown
that 74% of falsework failures occurred during concrete placement
operations. This stage of construction represents the critical condi-
tion for scaffold design, and it is this stage of construction that the
present paper addresses.
During the stage of concrete placement, the newly poured slab
has no rigidity, and all loads are supported by the scaffolds. The ef-
fect on structural reliability of human error is also discussed. In de-
sign practice, the vertical loads acting on the uprights of a scaffold
are most often calculated using the tributary area method, assum-
ing that the uprights support the weight of the newly poured slab
and a code-specied basic design live load (on the formwork).
Alternatively, the shore load can be obtained by calculating the
support reactions for the bearer beams loaded by concentrated
loads from the joists (referred to as distributed method in [19]
and beam method in [20]). It has been shown that the distributed
method or beam method is only slightly more accurate than the
tributary area method [19,20]. In the following discussion, unless
noted otherwise, the shore load was calculated using the tributary
area method.
As compared to the substantial work that has been done for
occupancy loads in buildings, relatively limited studies have been
conducted for construction loads. Several analytical models [22
25] have been proposed to study shoreslab interactions in multi-
story constructions. Those studies were focused on determining
the loads on partially cured slabs. In [26] and [27], construction
live loads on slab formworks before and after concrete placement
were surveyed. Other load surveys measured directly the load
effects (axial loads) on formwork supporting elements
[19,21,28,20,29]. A brief summary of pertinent shore load surveys
is given below.
3.1. Live loads
The live load of interest in this study is the companion live load
when the dead load reaches its maximum at the end of the pour.
The companion live load can be estimated by subtracting the max-
imum shore load by the load just after the pour. This companion
live load was examined by Fattal [19] and Ikheimonen [20]. From
the data reported in [20], it is found that: (1) the load due to the
weight of workmen and equipment is very small in comparison
with the dead load; (2) when concrete is placed by pumping, the
impact load is insignicant; (3) when the crane-and-bucket meth-
od is used for concreting, the COV V
L
equals 0.7, and the mean live
load L equals approximately 0.3 L
n
if the nominal live load L
n
is cal-
culated using a design formwork live load of 2.4 kPa (50 psf) as
specied in ACI 347 [30]. The investigation conducted by Fattal
[19], in which the crane-and-bucket method was used for concret-
ing, suggested the similar result: L=L
n
0:31 and V
L
= 0.71 for the
tributary area method, and L=L
n
0:27 and V
L
= 0.54 if the more
accurate distributed method was used to calculate the shore loads.
Based on this limited data, it is assumed in the present study that
the live load on scaffolds has a Type I extreme distribution, with
V
L
= 0.7 and L=L
n
0:3 for a design formwork live load of 2.4 kPa
(50 psf). It is worth emphasizing that L=L
n
depends on the design
formwork live load, and that the value may vary considerably from
standard to standard. For example, the American Concrete Institute
species a design live load of 2.4 kPa (50 psf) for all construction
stages [30], while in the Australian Standard for formwork for con-
crete AS3610 [31], the design live load varies for different con-
struction stages; it is 1.0 kPa for the phase of concrete placement.
3.2. Dead load
Theoretically, the dead load transmitted to the uprights of a scaf-
fold can be calculated with reasonable accuracy as the variability in
concrete weight is relatively low (a COV on the order of 0.060.09
[17]). However, many studies have shown that the measured-to-
calculated value for the dead load applied to scaffolds has a
relatively large variation [1921,29].
Fattal [19] studied a six-storey at plate ofce building with a
slab thickness of 203 mm (8 in.). One-bay steel scaffold portal
frames consisting of two uprights and three horizontal cross-bars
were used as shores. Eleven instrumented uprights were placed
within an interior bay under the formwork for the fourth oor slab.
It was observedthat the loadinupright No. 4 was signicantly lower
thanits theoretical share of the superimposedload, andthe adjacent
upright No. 10 picked up the additional load. The author suggested
this might be due to premature yielding or misalignment of upright
No. 4. If uprights Nos. 4 and 10 are excluded fromthe survey data, D/
D
n
has a mean of 0.92 and a COV of 0.35 for the tributary area meth-
od, and a mean of 0.93 and a COVof 0.32 for the distributed method.
Rosowsky et al. [21] andKothekar et al. [32] reporteda shore load
survey at a Federal prison site in Beckley, WV. Two identical small
pour areas (8.10 m
2
each) and one large pour area (51.28 m
2
) were
surveyed during both the casting and curing phases. Each of the
small pour areas was supported by a steel scaffold tower with 4 up-
rights. The formwork shoring system for the large pour area con-
sisted of 16 steel posts, eight of which were instrumented. COVs of
33.3% and 28.5% were observed among the dead loads for the small
pour areas and the large pour area, respectively. It was also found
that the tributary area load was close to the average of the actual
shore loads.
Ikheimonen [20] surveyed shore loads at four bridge and ve
residential building sites. The slab thickness varied from 950 mm
to 1310 mm for the bridges, and 150 mm to 350 mm for the resi-
dential buildings. A total of 66 shores were instrumented. Mea-
sured shore loads were subdivided into load due to formwork,
load due to concrete, and the maximumload during the placement.
The measured-to-calculated value for the dead load was found to
have a mean of 0.9 and a COV of 0.29 for the tributary area method,
and a mean of 0.99 and a COV of 0.30 for the beam method.
Puente et al. [29] investigated the construction of a concrete
building with seven stories and four levels of underground parking.
The thickness of the slab was 250 mm. A total of 34 instrumented
steel post shores were placed on a typical 7.2 5.65 m interior
bay of each parking level. The study suggested that the dead load
D has a mean value approximately equal to the nominal D
n
and a
COV around 0.25.
On the basis of the above survey results, it appears that D D
n
,
and V
D
ranges from 0.25 to 0.3. Note that the dead load and occu-
pancy live load for ordinary buildings have typical COVs of 0.1 and
0.25, respectively [17]. It is evident that the variability of dead load
onscaffolds is muchgreater thanthe variabilityassociatedwithcon-
crete weight. The reason for this additional uncertainty is not fully
understood but it might be related to differential settlement of the
uprights, or imperfections in the scaffold installations such as lack
of bearing between the uprights and the formwork/bearer beams
[19,20,33,34]. Also, the tributary area concept might be a source of
the discrepancy since strictly speaking, the tributary area concept
is valid only when the soft formwork is very rigid. It appears that
under current normal construction practice, a COV of 0.30 is repre-
sentative for the deadloadeffect onthe scaffolds. Thus, it is assumed
in this study that the dead load is normally distributed, with
D=D
n
1:0 and V
D
= 0.30. It should be emphasized that V
D
includes
the effect incurred by minor imperfections in construction within
the limits of practical tolerances; it does not account for the effect
of (gross) human error.
H. Zhang et al. / Engineering Structures 36 (2012) 8189 83
4. Strength of scaffold structure
4.1. Example structure
A typical steel scaffold tower as shown in Fig. 2 was considered.
The system features 3 stories in a 1 bay 1 bay arrangement, with
equal storey height of 1.5 m and a bay width of 1.829 m in both
directions. The uprights and ledgers are connected via cuplok
joints. A cuplok joint consists of a xed bottom cup that is welded
to the uprights and a sliding upper cup. The blade ends of the hor-
izontal members (ledgers) are placed into the bottom cup. To
engage the joint, the top cup is slid down over the blades and ro-
tated by striking its lugs with a hammer. This type of connection
requires no bolting or welding and allows for simple and speedy
erection. Experimental tests have shown that cuplok joints are
semi-rigid and generally show looseness with small stiffness at
the beginning of loading due to a lack of t [35]. The jacks at the
bottom and top of the frame are adjustable and are assumed to
be of the same length. Two extreme values for the jack extensions
are considered, i.e., 100 mm and 600 mm. The corresponding
frames are referred to as Scaffolds 1 and 2, respectively. The
uprights, ledgers and braces are steel circular hollow sections
(CHS). The jacks are solid steel rods. Table 1 summarizes the nom-
inal section dimensions and yield stresses for the members.
4.2. Random properties of steel scaffold members
Seven basic random parameters were identied in the present
study: (1) yield stress of the uprights, (2) yield stress of the jacks,
(3) initial out-of-straightness of the uprights, (4) load eccentricity,
(5) thickness of the CHS of the uprights, (6) diameter of the jacks,
and (7) joint stiffness of the semi-rigid cuplok joints. All random
variables are assumed to be mutually statistically independent.
The following paragraphs and Table 2 summarize the statistical
information for the basic random variables. Detailed information
may be obtained from the references cited.
The mean and COV in the member out-of-straightness and load
eccentricity are taken from the results of the eld measurements
[35]. The out-of-straightness for uprights with spigot joints (i.e.,
splices) is modelled by a lognormal distribution with a mean of
L/770 and a COV of 0.6. The nominal value for out-of-straightness
is assumed to be L/500, a typical value adopted in industry. The
magnitude of load eccentricity is modelled by a lognormal distri-
bution with a mean of b/4 and a COV of 0.55, in which b is the
width of the bearer beam; this mean value is equal to the nominal
design value specied in the Australian Standard for formwork for
concrete AS3610 [31]. It is assumed that the measured data for
out-of-straightness and load eccentricity are representative of nor-
mal quality construction.
The stability of a steel frame is affected not only by the magni-
tudes of load eccentricity and column out-of-straightness but also
their patterns. As specied in AISC 360-10 [37], the most unfavour-
able pattern should be assumed for the initial geometric imperfec-
tions such that they provide the greatest destabilizing effect. The
eld survey [35] suggested that the pattern of load eccentricity ap-
Fig. 2. A steel scaffold tower with cuplok joints.
Table 1
Nominal member dimension and yield stress.
Dimension (mm) F
y
(MPa)
Upright 48.3 4 450
Ledger 48.3 3.2 350
Brace 48.3 3.2 400
Jack 36 430
Table 2
Description of basic random variables.
Random variable Nominal value
mean
nominal
COV Distribution Refs.
F
y
(upright) 450 MPa 1.05 0.1 normal [36]
F
y
(jack) 430 MPa 1.05 0.1 normal [36]
Thickness (upright) 4 mm 1.0 0.08 lognormal [17]
Diameter (jack) 36 mm 1.0 0.08 lognormal [17]
Joint stiffness (k
2
) 77.6 kNm/rad 1.0 0.2 normal [35]
Load eccentricity b/4
a
1.0 0.55 lognormal [35]
Out-of-straightness L/500 0.65 0.6 lognormal [35]
a
b = width of the bearer beam.
84 H. Zhang et al. / Engineering Structures 36 (2012) 8189
pears to be random and that the load eccentricities of uprights thus
may be assumed to be uncorrelated. However, this conclusion is
difcult to generalize without additional eld data. In the present
study, then, only the magnitudes of the out-of-straightness and
load eccentricity are taken to be random variables. All load eccen-
tricities are oriented in the same direction to maximize the desta-
bilizing effect. The pattern of member out-of-straightness is
determined using an elastic buckling analysis. The rst elastic
buckling mode is scaled and added to the perfect geometry to
determine the form of member out-of-straightness [38].
For ordinary steel structures fabricated with average quality
control, it is reasonable to assume that the mean sectional proper-
ties are equal to the Handbook values, with a typical COV of 0.05
[17]. However, since steel scaffold members are reused from one
job to another and new members are mixed with old ones, many
believe that this gives rise to additional uncertainty in sectional
properties. Accordingly, it is assumed that the mean-to-nominal va-
lue is 1.0 and the COV is 0.08 for the thickness of the uprights and
the diameter of the jacks. Researchers at the School of Civil Engi-
neering at the University of Sydney measured the sectional proper-
ties for a sample of 54 uprights taken fromstocks of material in use.
That investigation showed that the statistics for sectional proper-
ties assumed above appear to be on the conservative side.
The relation between the moment and rotation (Mh) of the cu-
plok joints is semi-rigid. Based on joint tests [35], a simple tri-linear
model as shown in Fig. 3 is adopted in this study to idealize the Mh
relation for cuplok joints. The joint stiffness appears to be depen-
dent on the number of ledgers connected at the joint. Three cases
can be identied, i.e., 4-way (interior joint), 3-way (edge joint)
and 2-way joint (corner joint). Only the 2-way joint is applicable
to the scaffolds considered in this study since they are one bay by
one bay. Among the three stiffness parameters (k
1
, k
2
and k
3
), the
variations of k
1
and k
3
are relatively small. Thus they were treated
as deterministic and set equal to 40.9 and 4.6 kN m/rad, respec-
tively. k
2
was taken to be a normal variable with a mean of 77.6
kN m/rad and a COV of 0.2.
4.3. Statistical characteristics of system strengths
To derive the statistical information for the strength of the steel
scaffolds, simulation approach was employed. This approach re-
quires an accurate means to predict the actual system strength,
and knowledge of the probability distributions for each basic ran-
dom parameter. In this study, a three-dimensional second-order
inelastic nite element model [12] was used in simulation to obtain
accurate theoretical predictions of system strength. The cuplok
joints were modelled as semi-rigid using the tri-linear Mh model
as shown in Fig. 3. The stressstrain relation for the steel in all
CHS is assumed elastic perfectly-plastic. For simplicity, a pinned
connection is assumed for the base connection, i.e., exural rota-
tions are free to occur while twist rotations are prevented. Alterna-
tively, the base connection may be modelled as semi-rigid using an
elastic rotational spring k
rb
[38]. The top of each upright is re-
strained from translation in the horizontal directions. This assump-
tion is justied primarily because in construction practice, the
horizontal formwork is usually restrained externally by the (par-
tially completed) permanent structure. The bearer beam (see
Fig. 2) generally places a constraint on the rotation of the U-head,
which can be modelled by an elastic rotational spring k
rt
applied
at the top of each upright [12]. Values of k
rb
and k
rt
depend on many
factors such as the applied load and properties of the bearers, and
are difcult to quantify and generalize. In the present study, the ef-
fects of k
rb
and k
rt
are not included. It is believed that neglecting
their effects is conservative.
Concentrated vertical loads of equal magnitude are applied at
the top of each upright at an eccentricity with respect to the cen-
troid axis, as described in Section 4.2. The loads were increased
incrementally until system failed by instability. Details of the
structural modelling can be found elsewhere [12].
Statistics of system strengths were simulated using the Latin
Hypercube (LHC) sampling method and the basic random variables
summarized in Table 2. In each simulation, the ultimate system
strength was obtained using the second-order inelastic analysis.
The LHC method is a stratied sampling scheme to derive the statis-
tical properties of a response variable, and is highly efcient for a
complex system. Compared with the direct Monte Carlo simulation,
fewer samples are required in LHC method to cover the probability
space and to achieve a desired mean-square error. Several investi-
gations were performed for each scaffold. The overall uncertainty
in scaffold capacity is obtained by treating all parameters as ran-
dom. Treating one parameter as a random variable and the remain-
ing parameters as deterministic and equal to their respective
nominal values provides an indication of the relative signicance
of different random properties to the variability in systemstrength.
For simplicity, perfect correlation is assumed for the material and
geometric properties between upright-to-upright, jack-to-jack,
and for the stiffness between joint-to-joint. No correlation exists
between upright-to-jack, upright-to-joint, or jack-to-joint. Typi-
cally, 300 LHC simulations were performed for each uncertainty
analysis of each scaffold.
The nominal strengths of the two scaffold structures were rst
evaluated as reference points using the second-order inelastic FEA,
and the nominal values for the material, geometric and stiffness
properties, as summarized in Table 2. The nominal load eccentric-
ity is taken as b/4 = 20 mm, assuming that the width of the bearer
beam is 80 mm.
4.4. Scaffold 1100 mm jack extension
With the nominal properties, the ultimate load that can be car-
ried by Scaffold 1 is found to be 107.2 kN per upright. The uprights
buckled in an S curvature with negligible overall sway as the nal
failure shape, indicating the failure of the system is due to inelastic
exural buckling of the upright members. The maximum yield ra-
tio, dened as the percentage of cross-section that has yielded, is
about 40% for the uprights at the rst storey. This justies the use
of an inelastic analysis. The jacks, ledgers and bracings are all within
their elastic limits at failure.
Table 3 presents the simulated strength statistics for Scaffold 1
considering each random variable separately as described above.
The results show that R=R
n
is around unity for all cases. The Fig. 3. Idealized Mh curve for two-way cuplok joint.
H. Zhang et al. / Engineering Structures 36 (2012) 8189 85
variability in system strength mainly arises from the uncertainties
associated with load eccentricity, material (F
y
) and geometric prop-
erties (thickness) of the uprights. The strength histogramassociated
with random load eccentricity is shown in Fig. 4, with a COV of
12.9%. This is not surprising considering the large variability as-
sumed in the load eccentricity (a COV of 55%). Note that the histo-
gram is negatively skewed. Since Scaffold 1 with nominal
properties failed due to inelastic exural buckling of the uprights,
it is expected that the system strength will be sensitive to the vari-
ations of those parameters inuencing the inelastic buckling
strength of the uprights. For example, V
R
takes the values of 4.9%
and 5.1% for the random yield stress and random thickness of the
uprights, respectively. On the other hand, the joint stiffness and
out-of-straightness appear to have modest impacts on the variabil-
ity in system strength. Large changes in these two random proper-
ties are associated with relatively small changes in systemstrength.
The yield strength of the jacks has an inconsequential effect on V
R
since at failure the jacks were in elastic range and buckling of the
jacks did not occur.
Fig. 5 shows the strength histogram for Scaffold 1 when all the
random variables are included, assuming that they are statistically
independent. The mean strength R equals 1.02R
n
(108.8 kN) with a
COV of 17%. The histogramis noticeably skewed to the left, a similar
trend observed in the strength histogram associated with random
load eccentricity (see Fig. 4). Four probability distribution functions
were examined to represent the strength distribution of Scaffold 1:
the normal, lognormal, Beta and Weibull distributions. Two statis-
tical hypothesis tests, namely KolmogorovSmirnov test and
AndersonDarling test, were performed using the simulated data.
It was found that the strength of Scaffold 1 can be best tted by a
three-parameter Weibull distribution.
4.5. Scaffold 2600 mm jack extension
The nominal model of Scaffold 2 failed at 44.64 kN, indicating
an almost 60% reduction to the system nominal capacity when
the top and bottom jack extensions increased from 100 mm to
600 mm. A different failure mode, in which the lateral deforma-
tions were essentially conned to the jacks and the scaffolding in
between shows little deformation, was observed in this case. The
uprights and ledgers experienced only small curvature and
remained elastic. The system strength is governed by the buckling
of the jacks and the strength of other members and connection
stiffness of the cuplok joints are not being utilized.
Table 3 presents the results for Scaffold 2 obtained from simu-
lation studies. The diameter of the jack and the load eccentricity
are the most inuential factors for the variability in structural
strength. It is worth noting that the variability in jack diameter
(COV = 8%) led to the greatest variability in system strength
(V
R
= 12.7%). For all other basic random variables, however, the
strength variabilities are smaller than those associated with the
random properties. The variation of system strength with jack
diameter is given in Fig. 6 for Scaffolds 1 and 2. The slope of the
curve is a measure of the sensitivity of strength to the jack diame-
ter. From Fig. 6 it is evident that the strength of Scaffold 2 is very
sensitive to the jack diameter. In the case of Scaffold 1, the curve
is essentially at when the jack diameter is greater than approxi-
mately 33 mm since the failure mode, inelastic buckling of the up-
rights, is not affected by the jack diameter. However, a sudden
change in the slope of the curve is observed when the jack
diameter becomes less than 33 mm. This is because the failure
mode for Scaffold 1 changes to inelastic buckling of the jacks; thus,
the jack diameter becomes an inuential factor for system
strength.
Table 3
Effects of random variables on system strength R.
Random variable Scaffold 1 Scaffold 2
R=Rn
V
R R=Rn
V
R
F
y
(upright) 1.02 4.85% 1.0 0
F
y
(jack) 1.0 0 1.0 1.28%
Thickness (upright) 1.0 5.08% 1.0 2.70%
Diameter (jack) 0.99 2.48% 0.99 12.67%
Out-of-straightness 1.03 4.0% 1.0 0
Load eccentricity 0.99 12.91% 1.0 8.90%
Joint stiffness 1.0 2.10% 1.0 1.48%
All random variables 1.02 17% 1.0 16%
(a)
(b)
Fig. 4. Histogram for system strength under random load eccentricity. (a) Scaffold 1
and (b) Scaffold 2.
2.5
3
3.5
1
1.5
2
2.5
3
3.5
3
0
0.5
1
1.5
2
2.5
3
3.5
0.37 0.55 0.74 0.92 1.11 1.29
P
r
o
b
a
b
i
l
i
t
y

d
e
n
s
i
t
y
Normalized system strength (R/R
n
)
Fig. 5. Histogram for Scaffold 1 strength with all random variables (tted with a
three-parameter Weibull distribution).
86 H. Zhang et al. / Engineering Structures 36 (2012) 8189
Fig. 4 shows the strength histogramfor Scaffold 2 under random
load eccentricity. The system strength vs. load eccentricity relation
for Scaffolds 1 and 2 are presented in Fig. 7 for comparison. Fig. 7
suggests that the strength of Scaffold 1 is more sensitive to the var-
iation of load eccentricity than the strength of Scaffold 2.
The yield stress and initial out-of-straightness of the uprights
have no effects on V
R
for Scaffold 2. This is due to the fact that
the uprights remain elastic and essentially straight within each
storey when the jacks buckle. The joint stiffness does not inuence
the system strength signicantly as the rotation of the cuplok
joints was relatively small. The effect on V
R
of the variation in jack
yield stress is minor because only a small portion (less than 20%) of
the jack cross-section reached the yield stress.
The strength of Scaffold 2 has a mean-to-nominal ratio of 1.0
and a COV of 16% when all random variables are considered simul-
taneously. The strength histogram is shown in Fig. 8. The strength
is described by a three-parameter Weibull distribution in the sub-
sequent reliability analysis.
5. Scaffold reliability
5.1. Design with advanced analysis
Table 4 summarizes the statistical data of the loads (D and L)
and the system resistance (R). Note that the nominal system resis-
tance, R
n
, in Table 4 was computed by the advanced nonlinear anal-
ysis. With this information at hand, the structural reliability of
Scaffolds 1 and 2 can readily be evaluated using FORM. Using the
WLL design equation (Eq. 1) and a factor of safety F.S. = 2.0, the
allowable working load, R
n
/F.S., are 53.6 kN and 22.32 kN for Scaf-
folds 1 and 2, respectively. These values are half of the nominal
system strengths obtained using advanced analysis, as described
in Sections 4.4 and 4.5. Fig. 9 plots the reliability index, b, with
L
n
/D
n
for Scaffolds 1 and 2 when all the basic random variables
are taken into account. For a design live load on formwork of
2.4 kPa (50 psf), L
n
/D
n
for the supporting scaffolds has a practical
range of 0.30.7, since the slab thickness ranges from 150 to
300 mm (612 in.) in typical concrete residential and commercial
buildings. Representative values for b are 2.6 and 3.1 for Scaffolds
1 and 2, respectively, for the loading condition L
n
/D
n
= 0.5. These
two reliability indices correspond to probabilities of failure of
4.7 10
3
and 9.7 10
4
, respectively.
From Fig. 9, it can be seen that b for Scaffold 2 is somewhat
higher than that of Scaffold 1, although the two structures have
similar strength statistics. This result can be explained by the fact
that the strength distribution of Scaffold 1 is negatively skewed
with a long lower tail which governs the limit-state probability,
while for Scaffold 2, its strengths are relatively evenly distributed
on both sides of the mean. This highlights the importance of incor-
porating information about probability distributions in the reliabil-
ity analysis.
It should be recognized that although Scaffold 2 has a higher
reliability than Scaffold 1, its nominal strength is only about 40%
of the latter. Considering that Scaffolds 1 and 2 are identical except
for the length of jack extensions (100 mm vs. 600 mm), Scaffold 1
is a much more efcient design from the point of view of material
utilization.
Fig. 9 indicates that b increases as L
n
/D
n
increases. This observa-
tion is different from the trend observed for typical steel structural
members whose reliability generally decreases as L
n
/D
n
increases
[17]. The discrepancy is a consequence of two characteristics of
the probabilistic load models for scaffolds: (1) dead load is the dom-
inant component in the total loads; and (2) although the live load
has a larger COV, its mean-to-nominal ratio is much smaller in com-
Fig. 6. System strength vs. diameter of jack.
Fig. 7. System strength vs. load eccentricity.
Fig. 8. Histogram for Scaffold 2 strength with all random variables (tted with a
three-parameter Weibull distribution).
Table 4
Statistical information for the loads and resistance.
Variable Mean COV Distribution
D 1.0D
n
0.3 Normal
L 0.3L
n

0.7 Extreme type I


R (Scaffold 1) 1.02R
n
0.17 3-P Weibull
R (Scaffold 2) 1.0R
n
0.16 3-P Weibull

Based on a design formwork live load of 2.4 kPa (50 psf).


H. Zhang et al. / Engineering Structures 36 (2012) 8189 87
parison with the dead load, i.e., L=L
n
0:3 vs. D=D
n
1:0. It appears
that for scaffold structures, dead load has a more dominant inu-
ence on the structural limit-state probability than the live load.
5.2. Design with elastic analysis
The above reliability analysis is based on the nominal system
strengths R
n
that are computed using the advanced nonlinear FEA.
Although design by second-order inelastic analysis is permitted in
the Australian steel design standard AS 4100-1998 [39] and the
American specication for structural steel buildings AISC 360-10
[37], the current routine design procedure is to use a two-dimen-
sional elastic analysis and evaluate the limit states in accordance
with the Specication equations. Table 5 summarizes the nominal
strengths for Scaffold 1 computed by elastic analysis conforming
to the Australian steel specication [39]. The design was controlled
by the inelastic exural buckling of the uprights. Three elastic de-
sign methods, denoted EA1, EA2 and EA3, are presented. The cuplok
connections between uprights and ledgers are assumed to be rigid
in EA 1 and EA 2, and semi-rigid in EA 3. The effective length factor
for the uprights was conservatively taken as 1.0 in EA 1, but was
determined more rationally by a frame elastic buckling analysis
in EA 2 and EA 3. While method EA1 is the simplest and most likely
to be used by the design engineers, it is the least accurate and
underestimates the strength by 25% as compared with the ad-
vanced analysis. The last column of Table 5 compares the mean-
to-nominal ratio R=R
n
for different analysis and design methods.
Similar elastic designs were performed for Scaffold 2, and the
results are presented in Table 6. The strength of Scaffold 2 is gov-
erned by the buckling of the jacks. It can be seen that the results
from elastic analysis are very accurate in this case. This is to be ex-
pected since Scaffold 2 is largely in its elastic range when the very
slender jacks buckle.
Table 7 compares the reliability indices for Scaffolds 1 and 2
when their nominal strengths are computed by the advanced anal-
ysis and elastic analysis, assuming a typical loading condition L
n
/
D
n
= 0.5. For Scaffold 1, method EA 1 gives the most conservative
nominal strength, thus it has the highest reliability (b = 2.93). For
Scaffold 2, the reliability indices for the designs by advanced FEA
and by elastic analysis are all around 3.0.
5.3. Reliability in the presence of human error
The reliability assessment described above addresses structural
failures that are caused by stochastic variability in loads and struc-
tural strength. In addition to these uncertainties, the safety of scaf-
folds may be compromised by various types of human errors
[2,4043].
Considering the possibility of n human error scenarios, the
probability of structural failure due to human error can be derived
from the theorem of total probability [40]:
PFE
X
n
i1
PFjE
i
PE
i
4
in which FE = structural failure due to human error, E
i
= ith error
occurs (and is undetected), and FjE
i
= failure of the structure given
that ith error occurs.
It appears from Eq. (4) that the failure probability P(FE) can be
managed by controlling the consequence of an error if it does
occur, i.e., reducing the term P(Fj E
i
). In light of this, one might
attempt to use classic reliability theory to evaluate Eq. (4), and de-
rive a more conservative factor of safety (or load and resistance
factors) to compensate for the effect of human errors. This practice,
however, has proved to be ineffective and problematic [40]. It is
very difcult to evaluate Eq. (4) given the fact that there are
numerous sources of human errors, and available statistics of
human errors is very scarce and often biased. Even in the cases
where the variety of errors is limited or only one error is consid-
ered at a time, as suggested in [43], determining the magnitude
and frequency of human error is problematic since they may vary
greatly among individuals and organizations. Using a larger factor
of safety to account for human error (say, defective workmanship)
will unfairly penalize those who implement good construction
quality control. Instead, the risk of structural failure shall be miti-
gated by reducing the incidence of human error(s), i.e., reducing
the magnitude of P(E
i
). This may be achieved by quality assur-
ance/quality control strategies, vocational education and training,
and organization and management measures [2,4042]. Their
importance cannot be overemphasized.
Fig. 9. Reliability indices for Scaffolds 1 and 2 designed by the advanced analysis
with F.S. = 2.0.
Table 5
Nominal strength (R
n
) for Scaffold 1 computed using different methods.
Methods Upright-ledger joint k
e
R
n R=Rn
Advanced FEA semi-rigid 107.2 kN 1.02
EA 1 rigid 1.0 80 kN 1.36
EA 2 rigid 0.75
a
108 kN 1.0
EA 3 semi-rigid 0.83
a
91.2 kN 1.19
k
e
: effective length factor for uprights.
a
Determined by a frame elastic buckling analysis.
Table 6
Nominal strength (R
n
) for Scaffold 2 computed using different methods.
Methods Upright-ledger joint k
e
R
n R=Rn
Advanced FEA semi-rigid 44.6 kN 1.0
EA 1 rigid 2.35
a
47 kN 0.95
EA 2 rigid 2.15
b
48.0 kN 0.93
EA 3 semi-rigid 2.3
b
44.0 kN 1.01
k
e
: effective length factor for jacks.
a
Determined from the alignment charts for k
e
for sway members.
b
Determined by a frame elastic buckling analysis.
Table 7
Reliability indices, b, for Scaffolds 1and2designedwithdifferent methods (L
n
/D
n
= 0.5).
Methods Scaffold 1 Scaffold 2
Advanced FEA 2.6 3.12
EA 1 2.93 2.95
EA 2 2.6 2.88
EA 3 2.78 3.17
88 H. Zhang et al. / Engineering Structures 36 (2012) 8189
6. Conclusions
A reliability analysis of steel scaffold shoring systems was pre-
sented. Scaffold towers with steel cuplok connections and different
jack extensions were considered. The limit state was dened as
instability failure by overload. The study focused on the construc-
tion stage of concrete placement as this is the most critical time for
the safety of supporting scaffolds. Analyses of available literature
on shoring loads measured during concrete placement found that:
(1) Shoring load is mainly due to dead load effect. Live load, in gen-
eral, is small in this phase of construction. (2) The live load has a
COV of approximately 0.60.7. Its mean-to-nominal ratio equals
0.3 if the design live load (50 psf) given in ACI 347 [30] is used.
(3) The dead load has a mean-to-nominal ratio of 1.0 and a COV
around 0.3.
Simulation studies showed that for Scaffold 1, the variability in
structural strength mainly arises from the uncertainties in load
eccentricity, the yield strength and thickness of the uprights. For
Scaffold 2, the jack diameter and load eccentricity become domi-
nant inuences on the structural strength. These results indicate
the importance of using accurate structural models to capture
the mode of failure.
For a typical loading ratio L
n
/D
n
= 0.5, the reliability index for
Scaffold 1 is found to be 2.6 with a factor of safety of 2.0 if its nom-
inal ultimate strength is determined by an advanced nonlinear
structural analysis. If the nominal system strength is computed
using the customary elastic methods, the reliability index ranges
from 2.6 to 2.93, depending on the modelling of the cuplok joints
and the method for determining the effective length factor. For
Scaffold 2, the elastic analysis appears to be sufciently accurate
for evaluating the system strength, as the design is controlled by
the elastic buckling of the jacks. In this case, the reliability indices
for the designs by advanced FEA and by elastic analysis are similar,
all being approximately 3.0.
It appears that with a factor of safety of 2.0, the calculated reli-
ability indices for the scaffolds are lower than the reliability indices
for typical steel structural members under gravity loads, which are
about 4.04.2 on an annual basis [17]. While it is often argued that
the safety requirement for temporary structures need not be as
high as for permanent structures, risk acceptance and required
safety levels for temporary structures warrant further examina-
tion, as does the role of human error in control of scaffold safety.
Acknowledgements
This research is supported by Australian Research Council under
Linkage Grant LP0884156. This support is gratefully acknowledged.
The writers would also like to thank Dr. igo Puente (Universidad
de Navarra) for making the shore load survey data available.
References
[1] Hadipriono FC, Wang HK. Causes of falsework collapses during construction.
Struct Saf 1987;4:17995.
[2] Carper KL. Structural failures during construction. J Perform Constr Fac ASCE
1987;1(3):13244.
[3] Zhang ZY, Ke L. Safety accident of formwork engineering and its prevention in
China. Arch Technol 2008;39(7):4926 [in Chinese].
[4] Chan SL, Zhou ZH. Stability analysis of semirigid steel scaffolding. Eng Struct
1995;17(8):56874.
[5] Peng JL, Pan AD, Rosowsky DV, Chen WF, Yen T, Chan SL. High clearance
scaffold systems during construction I: structural modelling and modes of
failure. Eng Struct 1996;18(3):24757.
[6] Weesner LB, Jones HL. Experimental and analytical capacity of frame
scaffolding. Eng Struct 2001;23(6):5929.
[7] Yu WK, Chung KF, Chan SL. Structural instability of multi-storey door-type
modular steel scaffolds. Eng Struct 2004;26(7):86781.
[8] Prabhakaran U, Godley MHR, Beale RG. Three-dimensional second order
analysis of scaffolds with semi-rigid connections. Welding World
2006;50:18794.
[9] Peng JL, Chan SL, Wu CL. Effects of geometrical shape and incremental loads on
scaffold systems. J Constr Steel Res 2007;63(4):44859.
[10] Zhang H, Chandrangsu T, Rasmussen KJR. Probabilistic study of the strength of
steel scaffold systems. Struct Saf 2010;32:393401.
[11] Liu H, Zhao Q, Wang X, Zhou T, Wang D, Liu J, Chen Z. Experimental and
analytical studies on the stability of structural steel tube and coupler scaffolds
without X-bracing. Eng Struct 2010;32:100315.
[12] Chandrangsu T, Rasmussen KJR. Structural modelling of support scaffold
systems. J Constr Steel Res 2011;67:86675.
[13] Gromala DS. Calculating apparent reliability of wood scaffold planks. Struct Saf
1984;2:4757.
[14] Charuvisit S, Ohdo K, Hino Y, Takanashi S. Risk assessment for scaffolding work
in strong winds. In: Furuta H, Frangopol DM, Shnozuka M, editors. Proceedings
10th international conference on structural safety and reliability
(ICOSSAR09): safety, reliability and risk of structures, infrastructures and
engineering systems. CRC Press; 2010.
[15] Epaarachchi DC, Stewart MG, Rosowsky DV. Structural reliability of multistory
buildings during construction. J Struct Eng ASCE 2002;128(2):20513.
[16] Epaarachchi DC, Stewart MG. Human error and reliability of multistory
reinforced-concrete building construction. J Perform Constr Fac ASCE
2004;18(1):1220.
[17] Ellingwood BR, MacGregor JG, Galambos TV, Cornell CA. Probability based load
criteria: load factors and load combinations. J Struct Div ASCE
1982;108(5):97897.
[18] Melchers RE. Structural reliability analysis and prediction. 2nd ed. West
Sussex, England: John Wiley & Sons; 1999.
[19] Fattal S. Evaluation of construction loads in multistory concrete buildings, NBS
building science series 146. Washington, DC: National Bureau of Standards;
1983.
[20] Ikheimonen J. Construction loads on shores and stability of horizontal
formworks, Ph.D. thesis. Stockholm, Sweden: Royal Institute of Technology;
1997.
[21] Rosowsky DV, Philbrick Jr TW, Huston DR. Observations from shore load
measurements during concrete construction. J Perform Constr Fac ASCE
1997;11(1):1823.
[22] Nielsen KEC. Loads on reinforced concrete oor slabs and their deections
during construction, Bulletin No. 15. Stockholm: Swedish Cement and
Concrete Research Institute, Royal Institute of Technology; 1952.
[23] Grundy P, Kabaila A. Construction loads on slabs with shored formwork in
multistory buildings. ACI J 1963;60(12):172938.
[24] Liu XL, Chen WF, Bowman MD. Construction load analysis for concrete
structures. J Struct Eng ASCE 1985;111(5):101926.
[25] El-Shahhat M, Chen WF. Improved analysis of shoreslab interaction. ACI J
1992;89(5):52837.
[26] Karshenas S, Ayoub H. Construction live loads on slab formworks before
concrete placement. Struct Saf 1994;14:15572.
[27] Karshenas S, Ayoub H. Analysis of concrete construction live loads on newly
poured slabs. J Struct Eng ASCE 1994;120(5):152542.
[28] Agarwal RK, Gardner NJ. Form and shore requirements for multistory at slab
type buildings. ACI J 1974;71(11):55969.
[29] Puente I, Azkune M, Insausti A. Shoreslab interaction in multistory reinforced
concrete buildings during construction: an experimental approach. Eng Struct
2007;29:73141.
[30] ACI 347. Guide to formwork for concrete, ACI 347-02. Farmington Hills, MI,
USA: American Concrete Institute; 2004.
[31] Standards Australia. Australian Standard AS3610-1995 formwork for concrete.
Sydney, NSW 2001, Australia: Standards Australia; 1995.
[32] Kothekar AV, Rosowsky DV, Huston DR. Investigating the adequacy of vertical
design loads for shoring. J Perform Constr Fac ASCE 1998;12(1):417.
[33] Rosowsky DV, Huang YL, Chen WF, Yen T. Modeling concrete placement loads
during construction. Struct Eng Rev 1994;6(2):7184.
[34] Ferguson S. Formwork shore design loads: an evaluation of current practice.
Concrete 2001;35(5):2830.
[35] Chandrangsu T, Rasmussen KJR. Investigation of geometric imperfections and
joint stiffness of support scaffold systems. J Constr Steel Res 2011;67:57684.
[36] Galambos TV, Ravindra MK. Properties of steel for use in LRFD. J Struct Div
ASCE 1978;104(9):145968.
[37] AISC 360-10. Specication for structural steel buildings. Chicago, IL: American
Institute of Steel Construction; 2010.
[38] Chan SL, Huang HY, Fang LX. Advanced analysis of imperfect portal frames
with semirigid base connections. J Eng Mech ASCE 2005;131(6):63340.
[39] AS 4100-1998. Australian Standard AS4100-steel structures. Sydney, NSW
2001, Australia: Standards Australia; 1998.
[40] Ellingwood BR. Design and construction error effects on structural reliability. J
Struct Eng ASCE 1987;113(2):40922.
[41] Stewart MG. Structural reliability and error control in reinforced concrete
design and construction. Struct Saf 1993;12:27792.
[42] El-Shahhat AM, Rosowsky DV, Chen WF. Accounting for human error during
design and construction. J Archit Eng ASCE 1995;1(2):8492.
[43] Nowak AS. Effect of human error on structural safety. ACI J 1979:95972.
H. Zhang et al. / Engineering Structures 36 (2012) 8189 89

Você também pode gostar