SP-246—8

Creep and Shrinkage and the Design of

Supertall Buildings—A Case Study:
The Burj Dubai Tower

by W.F. Baker, D.S. Korista, L.C. Novak, J. Pawlikowski, and B. Young

Synopsis: The Burj Dubai tower, which is currently under construction, will be the world’s tallest
structure. This paper addresses the structural system utilized for the Burj Dubai tower and the
structural design implications of creep and shrinkage of the high performance reinforced concrete
vertical load carrying elements.

Keywords: Burj Dubai; column shortening; creep; gravity sidesway; high-rise;

shrinkage; vertical shortening

133
134 Baker et al.
William F. Baker, CE, SE, PE, FASCE is a Partner with Skidmore, Owings & Merrill LLP, Chicago, IL

D. Stanton Korista, PE, SE, FASCE, FIStrucE, MICE, CEng is the Director of Structural Engineering with
Skidmore, Owings & Merrill LLP, Chicago, IL

Lawrence C. Novak, SE, SECB, LEED® AP is an Associate Partner with Skidmore, Owings & Merrill LLP,
Chicago, IL

James Pawlikowski, SE, LEED® AP and Bradley Young, SE, PE are Associates with Skidmore, Owings &
Merrill LLP, Chicago, IL

STRUCTURAL SYSTEM DESCRIPTION

The superstructure is currently under construction and as of spring 2007 has reached over 120 stories. The final
height of the building is a “well-guarded secret”. The height of the multi-use skyscraper will “comfortably” exceed
the current record holder of 509 meter (1671 ft) tall Taipei 101. The 280,000 m 2 (3,000,000 ft2) reinforced concrete
multi-use Tower is utilized for Retail, a Giorgio Armani Hotel, Residential and Office. The goal of the Burj Dubai
Tower is not simply to be the world’s highest building; it’s to embody the world’s highest aspirations.

Designers purposely shaped the structural concrete Burj Dubai – “Y” shape in plan – to reduce the wind forces on
the tower, as well as to keep the structure simple and foster constructability. The structural system can be described
as a “buttressed” core (Figs. 1 and 3). Each wing, with its own high performance concrete core and perimeter
columns, buttresses the others via a six-sided central core, or hexagonal hub. The result is a tower that is extremely
stiff torsionally. SOM (Skidmore, Owings & Merrill LLP) applied a rigorous geometry to the tower that aligned all
the common central core and column elements to form a building.

Each tier of the building steps back in a spiral stepping pattern up the building. The setbacks are organized with the
Tower’s grid, such that the building stepping is accomplished by aligning columns above with walls below to
provide a smooth load path. This allows the construction to proceed without the normal delays associated with
column transfers.

The setbacks are organized such that the Tower’s width changes at each setback. The advantage of the stepping and
shaping is to “confuse the wind”. The wind vortexes never get organized because at each new tier the wind
encounters a different building shape.

The Tower and Podium structures are currently under construction (Fig. 2) and the project is scheduled for topping
out in 2008.

STRUCTURAL ANALYSIS AND DESIGN

The center hex reinforced concrete core walls provide the torsional resistance of the structure similar to a closed
pipe or axle. The center hex walls are buttressed by the wing walls and hammer head walls which behave as the
webs and flanges of a beam to resist the wind shears and moments. Outriggers at the mechanical floors allow the
columns to participate in the lateral load resistance of the structure; hence, all of the vertical concrete is utilized to
support both gravity and lateral loads. The wall concrete specified strengths ranged from C80 to C60 cube strength
and utilized portland cement and fly ash. Local aggregates were utilized for the concrete mix design. The C80
concrete for the lower portion of the structure had a specified Young’s Elastic Modulus of 43,800 N/mm2 (6,350ksi)
at 90 days. The wall and column sizes were optimized using virtual work / LaGrange multiplier methodology which
results in a very efficient structure. The reinforced concrete structure was designed in accordance with the
requirements of ACI 318-02 Building Code Requirements for Structural Concrete.

The wall thicknesses and column sizes were fine-tuned to reduce the effects of creep and shrinkage on the individual
elements which compose the structure. To reduce the effects of differential column shortening, due to creep,
between the perimeter columns and interior walls, the perimeter columns were sized such that the self-weight
 Structural Implications of Shrinkage and Creep of Concrete 135
gravity stress on the perimeter columns matched the stress on the interior corridor walls. The five (5) sets of
outriggers, distributed up the building, tie all the vertical load carrying elements together, further ensuring uniform
gravity stresses; hence, reducing differential creep movements. Since the shrinkage in concrete occurs more quickly
in thinner (smaller V/S ratio) walls or columns, the perimeter column thickness of 600mm (24”) matched the typical
corridor wall thickness (similar volume to surface ratios – V/S) (Fig. 4b) to ensure the columns and walls will
generally shorten at the same rate due to concrete shrinkage.

The top section of the Tower consists of a structural steel spire utilizing a diagonally braced lateral system. The
structural steel spire was designed for gravity, wind, seismic and fatigue in accordance with the requirements of
AISC Load and Resistance Factor Design Specification for Structural Steel Buildings (1999).

The structure was analyzed for gravity (including P-Delta analysis), wind, and seismic by ETABS version 8.4. The
three-dimensional analysis model consisted of the reinforced concrete walls, link beams, slabs, raft, piles, and the
spire structural steel system (Fig. 4). The full model consisted of over 73,500 shells and 75,000 nodes.

The Tower foundations consist of a pile supported raft. The solid reinforced concrete raft is 3.7 meters (12 ft) thick
and was poured utilizing C50 (cube strength) self consolidating concrete (SCC). The raft was constructed in four (4)
separate pours (three wings and the center core). Each raft pour occurred over at least a 24 hour period.

The Tower raft is supported by 194 bored cast-in-place piles. The piles are 1.5 meters in diameter and
approximately 43 meters long, with a capacity of 3,000 tonnes each (pile load tested to 6,000 tonnes). The C60
(cube strength) SCC concrete was placed by the tremie method utilizing polymer slurry. The friction piles are
supported in the naturally cemented calcisiltite / conglomeritic calcisiltite formations developing an ultimate pile
skin friction of 250 to 350 kPa (2.6 to 3.6 tons / ft2).

CONSTRUCTION

The Burj Dubai utilizes the latest advancements in construction techniques and material technology. The walls are
formed using Doka’s SKE 100 automatic self-climbing formwork system. The circular nose columns are formed
with steel forms, and the floor slabs are poured on MevaDec formwork. Wall reinforcement is prefabricated on the
ground in 8m sections to allow for fast placement for the 3-day cycle on the typical floors. The construction
sequence for the structure has the central core and slabs being cast first, in three sections; the wing walls and slabs
follow around 10 stories behind; and the wing nose columns and slabs follow behind these (Fig. 2). Concrete is
pumped via specially developed Putzmeister pumps, able to pump to heights of 600m (1,970 ft) in a single stage and
generate 350 bar pressure. Due to the limitations of conventional surveying techniques, a special GPS monitoring
system has been developed to monitor the verticality of the structure. The construction survey work is being
supervised by Mr. Doug Hayes, Chief Surveyor for the Burj Dubai Tower, with the Samsung BeSix Arabtech JV.

CREEP AND SHRKINAGE ANALYSIS

Historically, engineers have typically determined the behavior of concrete structures using linear-elastic finite
element analysis and/or summations of vertical column loads. As building height increases, the results of such
conventional analysis may increasingly diverge from actual behavior. Long-term, time-dependant deformations in
response to creep and shrinkage can cause redistribution of forces and gravity induced sidesway that would not be
detected by conventional methods. When the time-dependant effects of creep, shrinkage, variation of concrete
stiffness with time, sequential loading and foundation settlements are not considered, the predicted forces and
deflections may be inaccurate. To account for these time-dependant concrete effects in the Burj Dubai Tower
structure, a comprehensive construction sequence analysis incorporating the effects of creep and shrinkage was
utilized, based on the procedures and equations presented in Appendix A, to study the time-dependant behavior of
the structure.
136 Baker et al.
Construction Sequence Analysis Procedures
The time-dependant effects of creep, shrinkage, the variation of concrete stiffness with time, sequential loading and
foundation settlement were accounted for by analyzing 15 separate three-dimensional finite-element analysis
models, each representing a discrete time during construction (Fig. 5). At each point in time, for each model, only
the incremental loads occurring in that particular time-step were applied. Additional time steps, after construction,
were analyzed up to 50 years. The structural responses occurring at each time-step were stored and combined in a
database to allow studying the predicted time-dependant response of the structure. The results for the predicted
movement of the center of the tower are reported in Figs. 8, 9, 11 to 14.

Long-term creep and shrinkage testing, over one year in duration, have been performed, by the CTL Group (located
in Skokie, IL) under contract with Samsung, on concrete specimens to better understand the actual behavior of the
concrete utilized for the project.

The typical creep and shrinkage specimen is a 6” diameter cylinder with a volume to surface ratio (V/S) of 1.5”
(38mm); however, the typical concrete elements used for columns and walls in highrise buildings have volume to
surface ratios ranging from 12” to 20” (300mm to 500mm). The rate of creep and shrinkage as well as the total
value is important in the design of structures. Figs. 6 and 7 indicate the GL2000 model predicted creep and
shrinkage with time for various volume to surface ratios. For V/S ratios between 1.5” and 20” (38mm and 500mm),
at 50 years, the predicted creep is within a range of 15%; however, the predicted range of shrinkage is 40%. Hence,
one area which stills needs research is to better understand the time-dependant inelastic creep and shrinkage
movements of concrete with high volume to surface ratios.

Compensation Methodology
The tower is being constructed utilizing both a vertical and horizontal compensation program. For vertical
compensation, each story is being constructed to the theoretical elevation and incorporating a typical floor-to-floor
height increase of 4mm starting after completion of the lower typical floors. This vertical compensation was
selected to ensure the actual height of the structure, after the time-dependant shortening effects of creep and
shrinkage, will be greater than the as-designed final height.

For horizontal compensation, the building is being “re-centered” with each successive center hex core jump. The re-
centering compensation will correct for all gravity induced sidesway effects (elastic, differential foundation
settlement, creep and shrinkage) which occur up to the casting of each story.

Vertical Shortening
Based on the procedures presented above and in Appendix A, the predicted time dependant vertical shortening of the
center of the core can be determined at each floor of the Burj Dubai tower (Figs. 8 and 9), not accounting for
foundation settlements. The total predicted vertical shortening of 300mm (12”) at the top of the concrete core,
subsequent to casting, is more than offset by the additional 550mm (22”) height added by the increased floor to floor
height compensation program.

Due to the compatibility requirements of strain between the rebar and the concrete in a reinforced concrete column,
as the concrete creeps and shrinks, i.e. shortens, the rebar must attract additional compressive stress and forces to
maintain the same strain as the concrete. Since the total load is the same, over time, part of the load in a reinforced
concrete column is transferred from the concrete to the rebar. This un-loading of the concrete, therefore, also
reduces the creep in the concrete (less load results in less creep). As per Fig. 10, the rebar in the columns and walls
(with a steel rho of about 1%) at Level 135 support about 15% of the load at the completion of construction and the
concrete supports 85%. However, after 30 years, the rebar supports 30% of the total load and the concrete supports
70%. This percent increase in force carried by the rebar increases as the steel rho is increased and/or as the total
load decreases.
 Structural Implications of Shrinkage and Creep of Concrete 137
Gravity Induced Horizontal Sidesway
Prediction of the gravity induced horizontal sidesway is more difficult than predicting the vertical shortening.
Gravity induced horizontal sidesway is extremely sensitive to the following:
• Differential Foundation Settlements
• Construction Sequence
• Differential Gravity Loading
• Variations in the Concrete Material Properties

The gravity sidesway can be thought of as the difference between the vertical shortening at the extreme ends of the
building causing curvature which is integrated along the height of the structure. Concrete creep and shrinkage
properties are variable. Taking the difference between two variable numbers results in a value which has an even
greater variability; hence, prediction of gravity induced horizontal sidesway is more of an estimate than the
prediction of vertical shortening alone.

Based on the construction sequence, time step, elastic, creep, shrinkage, and foundation settlement analysis,
predictions of the Burj Dubai tower gravity-induced horizontal sidesway have been made (Fig. 11). These
predictions, which are based on the outlined procedures and equations in Appendix A, compare favorably to the
survey data in both magnitude and direction (Figs. 12, 13 and 14).

CONCLUSION

The latest advances in creep & shrinkage modeling allows the designer to better understand the effects of the time-
dependant behavior of reinforced concrete and to design and detail the structure accordingly. When completed, the
Burj Dubai Tower will be the world’s tallest structure. It represents a significant achievement in terms of utilizing
the latest design, materials, and construction technology and methods, in order to provide an efficient, rational
structure to rise to heights never before seen.

PROJECT TEAM

Owner: Emaar Properties PJSC

Project Manager:
Architect, Structural Engineers,
& MEP Engineers:
Adopting Architect/Engineer
& Field Supervision:
Independent Verification
& Testing Agency:
General Contractor:
Turner Construction International
Skidmore, Owings & Merrill LLP
Hyder Consulting Ltd.
GHD Global Pty. Ltd.
Samsung / BeSix / Arabtec

APPENDIX A: CREEP AND SHRINKAGE EQUATIONS AND METHODOLGY

The creep and shrinkage prediction approach is based on the Gardner-Lockman GL20001 model with additional
equations to incorporate the effects of reinforcement and complex loading history. The governing equations and
procedures of which are summarized on the following pages.

1
N. J. Gardner, “Comparison of prediction provisions for drying shrinkage and creep of normal strength
concretes”, Canadian Journal of Civil Engineering, Vol.30, No.5, October 2004. pp 767-775.
138 Baker et al.
General Definitions

′
fc (MPa ) = specified 28-day concrete compressive strength
′ (MPa ) = mean (actual) 28-day concrete compressive strength
f cm28
′ 28 , may be estimated using:
when test data is not available, f cm
′ 28 = 1.1 f c′ + 5.0 (MPa )
f cm
t (days ) = age of concrete at time considered
t o (days ) = age of concrete at time of loading
t c (days ) = concrete curing period
h = relative humidity (example: use 0.50 for 50% RH)
V (mm ) = concrete volume to surface ratio
S

Factors depending on cement type selected:

Cement Type A b K s
(shrinkage)
I 2.8 0.77 0.335 1.0
II 3.4 0.72 0.4 0.75
III 1.0 0.92 0.13 1.15
Note: a, b, s and K should be determined from test data when possible.

General Equations

⎛ t 34 ⎞
f cmt (MPa ) = f cm 28 ⎜ ⎟ = concrete compressive strength at time t
⎜ a + bt 3 4 ⎟
⎝ ⎠

Alternately, fcmt can be determined from the modified CEB (Comite European du Beton) strength-development
equation:

2
f cmt (MPa ) = (β e ) f cm 28 = concrete compressive strength at time t

⎡s⎛ 28 ⎞ ⎤
⎢ ⎜⎜ 1− ⎟⎥
2
⎣⎢ ⎝ t ⎟⎠ ⎥⎦
where β e = e

1
E cmto (MPa ) = 3,500 + 4,300( f cmto
′ ) 2 =
concrete modulus at time of loading
1 = steel reinforcement correction factor
Rcfi =
ρni
1+
1− ρ
where As = steel reinforcement ratio
ρ=
Ag
E s (MPa ) = steel modulus of elasticity
Eci (MPa ) = concrete modulus of elasticity at time i
Es = modular ratio at time i
ni =
E ci
 Structural Implications of Shrinkage and Creep of Concrete 139
Derivation of the Rebar Correction Factor:

Consider the following basic equations for shrinkage strain for the condition in which the steel reinforcement does
not yield:

P
ε steel =
As E s
P
ε concrete = ε sh −
Ac E c
ε steel = ε concrete
where:
ε sh = shrinkage strain assuming a concrete-only section
ε steel = shrinkage strain in steel
ε concrete = shrinkage strain in concrete
P= compression in steel due to shrinkage
As = area of steel reinforcement = ρAgross
Ac = net area of concrete (1 − ρ )Ag
Es = steel modulus of elasticity
Ec = concrete modulus of elasticity
n= modular ratio = E s
Ec

Equating steel strain to concrete strain and solving for P:

ε sh
P=
1 + 1
As E s Ac E c
Substituting for As and Ac : P=
ε sh
1 1
+
ρAg Es (1 − ρ )Ag Ec
Multiplying by As Es :
As Es
⎡ ⎤
ε sh (As E s ) ε sh (As E s ) ⎢ 1 ⎥
P= = = ε sh As Es ⎢ ⎥
As E s As E s ρn ⎢1 + ρn ⎥
+ 1+
ρ Ag E s Ag E c (1− ρ ) (1 − ρ ) ⎣⎢ (1 − ρ )⎦⎥

R cf

ε AER
∴ P = ε sh As E s Rcf and ε conc = ε s = Ps = sh s s cf = ε sh Rcf
As Es As Es

Elastic Strain Equations

At time i: Pg i
ε ei = + ε ei−1
Ati Eci

where ε e = total elastic strain at time i

i
140 Baker et al.
Pg i = incremental gravity load applied at time i (kN)
Ati = transformed section area at time i (mm 2 )
Eci = concrete modulus of elasticity at time i (MPa)
ε ei−1 = total elastic strain at the previous time interval

Shrinkage Strain Equations

At time i: ( )
ε si = ε si w − ε si −1w Rcf i + ε si −1

where ε si = total shrinkage strain at time i

ε si w = baseline shrinkage strain at time i
ε si −1w = baseline shrinkage strain at the previous time
Rcf i = steel reinforcement correction factor at time i
ε si −1 = total shrinkage strain at the previous time interval

Note: equation applies for condition in which the steel reinforcement does not yield
where ε =ε β β siw shu h ti
1
2
⎛ 30 ⎞
-6
ε shu = ultimate shrinkage = 900 × K × 10 ⎜⎜ ⎟⎟
⎝ f cm 28 (MPa ) ⎠
β h = relative humidity factor = 1 − 1.18h 4
0.5
⎡ ⎤
t − tc
β ti = time factor = ⎢ ⎥
⎢ 2 ⎥
⎢⎣ t − t c + 0.12 S
V ( ) ⎥⎦

Creep Strain Equations

Incremental creep strain at time i due to load at time j (when steel does not yield):

( )
ε c (i −1)→ij = ε ci w j − ε c(i −1)w j Rcfi

where:
εc (i −1)→ij = incremental creep strain at time i due to load at time j
ε ci w j = creep strain at time i due to load at time j without
rebar effect
ε c( i−1)w j = creep strain at the previous time interval due to load at time j without rebar effect

Rcf i = steel reinforcement correction factor at time i

Pcj
ε ci w j = (CCi ) j
Ac Ecm 28
Pcj = load in concrete at time j
[ ]
Pcj = Pgj − As E s (ε ej + ε sj + ε cj ) − (ε e ( j −1) + ε s ( j −1) + ε c ( j −1) )

where:
 Structural Implications of Shrinkage and Creep of Concrete 141
Pgj = incremental gravity load applied at time j (kN)
2
As = area of steel reinforcement mm ( )
Es = steel modulus of elasticity (MPa)
ε ej = elastic strain in concrete at time j
ε sj = shrinkage strain in concrete at time j
ε cj = creep strain in concrete at time j
2
Ac = net area of concrete (mm )
Ecm 28 = conc. modulus of elasticity (MPa) at 28 days
(CCi ) j = creep coefficient at time i due to load at time j
0.5
⎡ 0.3 0.5 0.5 ⎛ ⎞ ⎤
⎢ 2(ti − t j ) ⎛ 7 ⎞ ⎛ ti − t j ⎞ ⎜ ti − t j ⎟ ⎥
(CCi ) j = Φ(tc )⎢ + ⎜ ⎟ ⎜ ⎟ + 2.5 1 − 1.086h 2
( )⎜ 2 ⎟ ⎥
( − )0.3
+ ⎜t ⎟ ⎜t −t +7 ⎟
⎢
⎣
t i t j 14 ⎝ ⎠ ⎝
j i j ⎠ ⎜ ti − t j + 0.12 V
⎝ S
( ) ⎟
⎠ ⎥
⎦

0. 5 0.5
⎡ ⎛ ⎞ ⎤
⎢ ⎜ t j − tc ⎟ ⎥
Φ (t c ) = Drying before Loading Factor = ⎢1 − ⎜ 2 ⎟ ⎥ , if (t j > t c )
⎢ ⎜⎝ t j − t c + 0.12 V S
⎣
( ) ⎟
⎠ ⎥
⎦

To calculate relaxation, Φ(t c ) remains constant at the initial value throughout the relaxation period. For creep
recovery Φ(t c ) remains constant at the value at the age of loading.

The value of ε c (i −1)→ij must be determined for each and every load time j that has occurred prior to the current time
i.
i −1
Total incremental creep strain at time i: ε c (i −1)→i = ∑ ε c (i −1)→ij where:
j =1

εc (i −1)→i = total incremental creep strain at time i

i −1
Total creep strain at time i: ε c = ∑ε c( n −1 ) → n
i
n =1
where:
ε ci = total creep strain at current time i
ε c( n −1) →n = total incremental creep strain at previous time n.
142 Baker et al.

Fig. 1—Rendering.

Fig. 2—Tower construction photo.

 Structural Implications of Shrinkage and Creep of Concrete 143

Fig. 3—Typical plan.

Fig. 4—Three-dimensional analysis model: (a) 3D view of analysis model; (b) 3D view of single story.

Fig. 5—Construction sequence 3D analysis models.

144 Baker et al.

Fig. 6—Predicted creep coefficient (CC) versus time for various V/S.

Fig. 7—Predicted shrinkage versus time for various V/S.

 Structural Implications of Shrinkage and Creep of Concrete 145

Fig. 8—Predicted vertical shortening versus story at 30 years (subsequent to casting).

Fig. 9—Predicted vertical shortening at top of concrete versus time (subsequent to casting).
146 Baker et al.

Fig. 10—Exchange of gravity axial force between concrete and reinforcing bar versus time.

Fig. 11—Predicted gravity lateral sidesway versus time.

 Structural Implications of Shrinkage and Creep of Concrete 147

Fig. 12—Surveyed versus predicted gravity sidesway deflection (X and Y).

Fig. 13—Surveyed versus predicted gravity sidesway X - deflection.

148 Baker et al.

Fig. 14—Surveyed versus predicted gravity sidesway Y - deflection.