Magazine of Concrete Research, 1997, 49, No. 181, Dec.

, 283-293

Drying shrinkage stresses in concrete patch

repair systems
M. Asad,* M. H. Baluch* and A. H. AI-Gadhib*

King Fahd University of Petroleum and Minerals

This paper proposes an approach to prediction of the build-up of stresses due to drying shrinkage in concrete patch
repair systems. The technique suggested consists of a twofold sequential approach, requiring the solution of the
moisture diffusion equation followed by execution of an appropriate stress analysis program. Solution of the non-
linear moisture diffusion problem is obtained conveniently through a finite element formulation, which outputs the
variation of moisture content in time and space. Inasmuch as drying shrinkage occurs by virtue of loss of moisture,
a phenomenological relationship between moisture loss and free shrinkage strain is established and used to convert
the finite element output of moisture loss to the free shrinkage strain. Time- and space-dependent values offree
shrinkage strain are then fed as input to an appropriate stress analysis program, conceptually similar to a thermal
stress analysis program, with the resulting output yielding the stress build-up in the patch repair system due to
drying shrinkage. The influence of repair layer variables including the diffusivity Kc. ultimate free shrinkage strain
fsh and modulus ofelasticity Er on the build-up of drying shrinkage stress is highlighted.

yield a continuous, monolithic system. In addition to

Introduction
The drying shrinkage of cementltlous material is
caused principally by the contraction of the calcium
silicate hydrate (C-S-H) gel in the hardened cement-
water paste, when the moisture of the gel is decreased.
Saturated cement paste does not remain dimensionally
stable when exposed to ambient humidities that are
below saturation, mainly because the loss of physically
adsorbed water from C-S-H results in shrinkage strain.!
It occurs initially at a high rate, and slows down with
time as the material ages.
Stresses due to drying shrinkage result when shrink-
age due to loss of moisture is not allowed to occur
freely, much in the same fashion as stresses result from
thermal changes in a constrained structural system.
Constraints may be of an internal form - as in non-
uniform drying in thick concrete pavements or deck
slabs - or may be of an external type as in a two-tier
slab system where a bottom precast unit is used in lieu
of formwork, over which fresh concrete is poured to
• Department of Civil Engineering, King Fahd University of Petro-
leum and Minerals, Dhahran 31261, Saudi Arabia.
(MCR 568) Paper received 17 April 1996; last revised 6 November
1996; accepted 14 January 1997
the prevalence of drying shrinkage stresses in new con-
struction, the repair of damaged bridge deck slabs or
industrial flooring by stripping and replacement of the
top layer almost inevitably leads to build-up of stresses
due to drying shrinkage.
Failure to account for stress build-up due to incom-
patible drying shrinkage characteristics often leads to
extensive cracking in the case of virgin construction
and delamination and cracking in the case of repaired
concrete construction - leading to the oft-repeated
cliche 'repair of the repair'.
Emrnons, Vaysburd and McDonald 2 - 5 emphasize
drying shrinkage of repair materials as the major pro-
blem of concrete repair. This is so because of the role
it plays in determining the durability of the repair
system. In patch repairs, the shrinkage of the repair
mortar is restrained by the existing substrate concrete
and this shrinkage of the repair mortar under restraint
is termed 'restrained shrinkage'. If there is no restraint
on the movement of the repair material then the shrink-
age of that repair material is referred to as 'free shrink-
age' and it does not induce any stress in the repair
material. Restrained shrinkage induces tensile and shear
stresses in the repair material, as illustrated in Fig. 1,
and these stresses increase with increase in drying
shrinkage. Once the stress build-up exceeds the materi-
283

0024-9831 © 1997 Thomas Telford Services LLd

Downloaded by [ Newcastle University] on [24/09/16]. Copyright © ICE Publishing, all rights reserved.
 Asad et at.
Moisture diffusion
Zone of high shear
. .<J. .
v.
'[>'
[>'
Fig. 1. Stresses in patch repair due to drying shrinkage
al tensile and shear strengths, crack initiation and pro-
pagation in the repair material results, together with
debonding and distress along the interface between the
repair and the substrate.
Selection of the repair materials for dimensional com-
patibility should be on the basis of minimum shrinkage,
such that the stresses produced are within acceptable
limits. However, this appears to be the issue most ig-
nored by the manufacturers of the repair materials as
well as the designers of repair systems. Studies con-
2
ducted by The Alberta Transportation and Utilities for
evaluation of concrete patching materials indicated that
the shrinkage of the majority of the repair materials far
exceeded the shrinkage strain value of a normal concrete
of 500 microstrain at 30 days. The ASTM C 157 shrink-
age test was used to determine the shrinkage values.
Only seven repair materials out of the 46 tested exhib-
ited a shrinkage value less than that of plain concrete.
This indicated that repair products are not being de-
signed to minimize shrinkage, despite the claim that the
materials available commercially are non-expansive,
non-shrinking or shrinkage-compensating. Moreover, it
was found that only limited information on the proper-
ties was available from manufacturers' data sheets. In-
formation regarding shrinkage was not even listed on
some of them.
2
It has been reported that the industry cannot limit
the manufacturers of repair materials to a certain
maximum shrinkage value because the basis for an
acceptable shrinkage value has not been established.
ASTM C 928, which specifies physical requirements
for packaged cementitious concrete repair materials,
recommends that shrinkage strain should not exceed
1500 microstrain. This is three times the shrinkage of
normal concrete. Repair materials which satisfy the
above criterion will be highly susceptible to excessive
drying shrinkage stress, cracking, delamination and
failure.
This paper addresses the methodology to be used in
the determination of drying shrinkage stresses occur-
284
Downloaded by [ Newcastle University] on [24/09/16]. Copyright © ICE Publishing, all rights reserved.
.[>.
r;, .[>
ring either in concrete patch repair systems or in two-
tier new construction. The problem is formulated using
finite elements to predict the non-linear moisture diffu-
sion through concrete. A phenomenological relationship
is established between moisture loss and free shrinkage
strain, unique to a specific cementitious product. The
free shrinkage strains are then fed into an appropriate
stress analysis program, yielding an estimate of drying
shrinkage stresses.
The model does not incorporate the beneficial effect
associated with tensile creep of the top-layer concrete,
which effectively reduces the tensile stress in the repair
layer with the passage of time. A properly designed
repair would guarantee a material tensile strength
development that was always greater than the evolution
of drying shrinkage stress, and this process would be
further aided by eventual reduction of shrinkage stress
in the post-creep phase.
Prediction of moisture distribution in
cementitious repair materials
Mathematical model for moisture diffusion
The mathematical model for moisture diffusion is
based on the assumption that moisture flow within a
cementitious material obeys the moisture diffusion
equation. 6 The diffusion equation, which is the govern-
ing differential equation for moisture flow in porous
materials, is given by Fick's second law as
(1)
where C(x, y, z, t) is the moisture content, Kc is the
moisture diffusivity and t is the time from the start of
the diffusion process.
The moisture diffusivity Kc is a material property
Magazine a/Concrete Research, 1997,49, No. 181

and can be defined as the rate at which the moisture
flows through the material when there exists a unit
moisture gradient. The diffusivity Kc is highly mois-
ture-dependent, which renders the moisture diffusion
problem through any porous medium non-linear.
Initial and boundary conditions
The initial condition is
C(x, y, Z, 0) = Co(x, y, Z)
where Co is the initial distribution of moisture content
in the moisture flow region.
The boundary conditions at the boundary between
the moisture flow region and the surrounding environ-
ment may be a combination of the following:
(a) no-flow boundary
8C =0
8n
(b) surface evaporation boundary
8C
Kc(C) 8n = f(C e - Cs)
where 8C/8n is the moisture gradient at the drying
surface with a unit normal n, Cs is the prescribed
moisture content at the boundary, f is the surface
moisture transfer coefficient or surface factor and Ce is
the ultimate or equilibrium moisture content that an
element would reach in steady-state equilibrium with
the environmental moisture conditions. It is both a
material property depending on mix proportions and a
forcing function depending on the relative humidity of
the environment. Cs is the moisture content at the dry-
ing surface.
Finite element formulation
The diffusion equation in one-dimensional form, in-
cluding domain evaporation, is given by
where G = jP/A, Q = JPCe / A, A is the cross-sec-
tional area and P is the perimeter of the cross-section.
The two-dimensional moisture diffusion equation, in-
cluding domain evaporation as shown in Fig. 2, is
8C 8 ( BC) 8 ( 8C)
at = 8x Kc(C) 8x + 8y Kc(C) 8y - GC + Q
where G = 2f/ t, Q = 2fCe / t and t is the thickness of
the body.
The finite element formulation of a system governed
by either of equations (5) or (6) results in the following
matrix equation:
[KHC} + [LHC} = {F}
MagaZine of Concrete Research, 1997,49, No. 181
Downloaded by [ Newcastle University] on [24/09/16]. Copyright © ICE Publishing, all rights reserved.
Drying shrinkage in patch repair systems
Boundary r
(2) Moisture diffusion
Fig. 2. Two-dimensional body with domain evaporation
where [K] represents the moisture diffusivity matrix,
[L] the moisture velocity matrix, {F} the external
moisture flow vector, {C} the nodal moisture content
and {C} the rate of change of nodal moisture content.
(3)
The matrix equation (7) is integrated in the time
domain using the finite difference method. By adopting
a backward difference scheme and discrete time steps
l:i.t, we obtain the following equation:
(4)
([L] + I:i.t[K]){ C}f+~f = [LH C}f + I:i.t{F} (8)
The matrix equation (8) can be solved for {C} at any
time t+l:i.t if the matrices [K], [L], {Cl and {F} are
known at the time step t. Since the diffusivity Kc is a
function of moisture content {C}, it changes during the
solution process and therefore [K] is re-evaluated for
each time step. Also, for each time step the solution
has to be obtained iteratively because the problem is
non-linear; equation (8) will not be satisfied and a
residual vector {lJI} will exist, such that
{lJI} = ([L] + I:i.t[K]){ C}'+M - [LH C}f -1:i.t{F} ~ 0
(9)
The iterations are continued until convergence is at-
tained, that is, when a selected norm of the residual
vector {lJI} becomes less than a specified level of
(5)
tolerance.
Computer program implementation
For the prediction of moisture distribution using this
moisture diffusion theory and finite element formula-
tion, two separate codes MSTDIFF I and MSTDIFF2,
for one- and two-dimensional moisture flow respec-
tively, were developed. A flow chart of the code
MSTDIFF2 is shown in Fig. 3.
(6)
Shrinkage stress analysis
A computer program was developed to compute the
stresses induced in a repair overlay/concrete substrate
system due to drying shrinkage of the repair material.
The code STRSRSYS (stress in repair system) is based
(7) on a finite element formulation of the two-dimensional
285
 Asad et at.

I START I
+
DATA NODEXY I
Input description of geometry, material properties,
boundary conditions and environmental moisture content
GAUSSQ I
t SFR2
I
I
LOADPS
Calculates the nodal force vector {to} and adds it to vector {F} I JACOB2 I
t
STIFFGT
Calculates the matrix [KG] and moisture velocity matrix [L)
SFR2
I
and adds [KG] to diffusivity matrix [K] JACOB2 I
+
EDGCON
Calculates the matrix [Kml and nodal force vector {ts} and adds
[Km] to diffusivity matrix [K] and {ts} to vector {F}
~ SFR2 I

•
TIME = 0
I Initializes time
I
STIFFD
• SFR2 I
Calculates the matrix [KD] and adds it JACOB2 I
to diffusivity matrix [K]
BMATPS I
t
ASSEMB
Assembles all element matrices
to form global matrices

I
-t
TIME = TIME + L'> T
Incrementing time with a constant
time step of L'> T

t
COMBIN
Combines the matrices [K], [L] and {F} according
to a specified finite difference scheme

+
IITER = 0
Initializes the iteration counter
I
A
6 (a)

Fig. 3. Flow chart of two-dimensional moisture diffusion program MSTDlFF2

plane stress problem. It allows for specifying in-plane
free shrinkage strains in the elements, representing re-
pair material in the repaired system.?
Experimental programme
An experimental programme was necessary to deter-
mine the moisture content C at different depths from
the drying surface in the repair material and thereby
286
determine the moisture diffusivity Kc as a function of
moisture content C, which is required as input data to
the codes MSTDIFF I and MSTDIFF2 for the predic-
tion of moisture loss from the repair material.
Experimental determination of free shrinkage data
of the repair material was necessary to relate moisture
loss to free shrinkage, and the free shrinkage data
is also required as the input to the code STRSRSYS for
the prediction of shrinkage stresses in the repaired
system.
Magazine of Concrete Research, 1997, 49, No. 181

Downloaded by [ Newcastle University] on [24/09/16]. Copyright © ICE Publishing, all rights reserved.
 Drying shrinkage in patch repair systems

0"
GREDUC, BAKSUB
Solves the equations for unknown moisture content {Cl

t
~j
I
IITER = IITER + 1
Incrementing iterations

t
I
STIFFD ~ SFR2 I
Updates the matrix [Ko] for the values of moisture
content {Cl obtained in the previous iteration and also +----1 JACOB2 I
updates the diffusivity matrix [K]
+----1 BMATPS I
t
ASSEMB
Assembles the new element diffusivity
matrices [K] to form global matrix

t
COMBIN
Combines the matrices [K], [L] and {F} according
TIME to a specified finite difference scheme
ITERATION
INCREMENT
LOOP
LOOP
t
REFOR1
Calculates the residual values {1J'}
at each iteration

~
CONUND
Checks if solution converge for a
particular time increment

t
RESULT
Prints moisture content {Cl for a
particular time

...
I END
I
(b)

The repair material used was a commercially avail-
able cementitious mortar, specially formulated for use
in the hot conditions of the Middle East. The ingredi-
ents include special pre-blended non-shrink cements,
graded sands, fillers and chemical additives. The mor-
tar can be trowel-applied to vertical and overhead sur-
faces. The manufacturer's data includes a compressive
strength of 33 MPa, flexural strength of 7 MPa, slant
shear bond strength of 23 MPa and coefficient of ther-
mal expansion a = 7·3 X 10-6 re.
The moisture con-
tent was determined at various depths from the drying
surface of the repair material by measuring the weight
changes of specimens of various length. Fig. 4 shows
Magazine of Concrete Research, 1997, 49, No. 181
the shapes and dimensions of the specimens. Moisture
was allowed to diffuse from the two ends while the
other four sides of the specimen were sealed with
paraffin wax so that the diffusion of moisture was one-
dimensional.
Free shrinkage strains in the repair material were
measured using a thin specimen of 10 mm thickness, as
shown in Fig. 4. Drying was allowed from the top
surface, and all the other surfaces were sealed. The test
was representative of a 20 mm thick specimen, diffus-
ing from both sides. The thickness is considered small
enough to not exhibit any effects of internal constraint.
The strains were measured using a demec gauge.
287

Downloaded by [ Newcastle University] on [24/09/16]. Copyright © ICE Publishing, all rights reserved.
 Asad et al.

~ Direction of moisture loss

40 - - , - - , - - - - - , - - - - - - - - - - - . , - - - , - - - - - - - - ,

1 cm 35
cm---+t~1 -J ~
I L.--::.:~----"';;:::"-':';:""'~bU !:: tEl~~7i-""··-~· ·.- ·T~.·- j~
1+-----7-'- 20

.•-..· .- ..··..
··· ·..,·· ·····-;·····..·..···,.. ·..· ·1
b = 4, 8, 12 cm

(a)
I ::t~7'" <: + + + -,- -.. L f-.._ _.j ; --cl

1 cm

I- 20cm -------+l~1 -J ~ i

U
5+/---..+---·:-:,...--T=~·-- ..---:---+--·+--.i-.._,-_·_1

5cmI I
_
•
_
0
o 5 10
I
15 20 25 30 35 40 45 50
Drying time: days

1 cm
.l- I
!- ~ 10cm

..
yealed

I Fig. 5. Moisture loss versus drying time at different depths

T
(b)

Fig. 4. Specimens for (a) measurement of moisture loss and Determination of moisture diffusivity
(b) measurement offree shrinkage strains The equation for one-dimensional moisture diffusion
when the moisture diffusivity Kc is a function of moist-
ure content C is

The tests were performed under controlled ambient oC = ox

&t 0 ( Kc(C) OC)
ox (11)
conditions, with a constant temperature of 35°C and
55 ± 5% RH. Using the Boltzmann transformation,9 C can be ex-
pressed in terms of a single variable 1] = 1](x, t) which
is a function of both space and time, defined as
Results and discussion 1] = x/ J t. Therefore, equation (11) is reduced to the
following ordinary differential equation by the intro-
Experimental determination of moisture content duction of the new variable 1]:
If it is assumed that the distribution of moisture at a
distance x from the drying surface is the same regard- --I JC' 1]dC = ( dC)C'
Kc(C)- (12)
less of the length of the specimen, then the moisture 2 I d1] I

content C(x, t) at time t and at distance x(x = 1, 3,

5 cm) is given by the following equation: 8 At C = I, rJ = 00 we have the condition dC/d1] = 0
and the moisture diffusivity Kc at any moisture content
C(x t) = (1 - (Wi - Wi-I) ) X lOO (10) C I is given as
, WO(li - li_I)S

[_~/~CI
C
where li is the length of the specimen, Wi is the
absolute moisture loss of the specimen (of length li) at
Kc\c=c, =
1] C=C,
] J'1]dC I
(13)
time t, Wo is the diffusible moisture in a unit volume
and S is the area of drying surface. The diffusible Therefore, the moisture diffusivity Kc can be deter-
moisture is given by the difference between the weight mined using equation (13) and the changes in moisture
of the specimen at the start of drying and its absolute content C with distance x and time t obtained from the
weight on oven drying after the experiment. experiments.
The relationship between moisture loss and drying The variation in moisture content C with depth x
time for the various specimens of different lengths was from the drying surface and time t is plotted against
obtained directly from the experiments. Time-depen- the variable 1] in Fig. 6. The relationship between the
dent changes in moisture content at distance x from the variable 1] and the moisture content C is expressed in a
drying surface (in the direction of width b, see Fig. form anal~~ous to that suggested by Penev and
4(a)) were calculated from equation (10). Fig. 5 shows Kawamura,
the moisture loss at different distances for different
drying times. According to Fig. 5, the change in moist- C(rJ) = (I - _P-)
P+1]
X lOO (14)
ure content near the surface of the concrete is rapid,
but at points slightly removed from the surface, the where p is the value of 1] at 50% moisture loss and is
changes are slow. equal to 0·065.
288 Magazine of Concrete Research, 1997, 49, No. 181

Downloaded by [ Newcastle University] on [24/09/16]. Copyright © ICE Publishing, all rights reserved.
 Drying shrinkage in patch repair systems

The moisture diffusivity Kc for concrete varies from

100 0·76 cm 2 /day in the saturated condition (C = 100%) to
0·047 cm 2 /day in the oven-dry condition (C = 0%).8
90 For lean concrete Kc varies from 90 cm 2 /day (satu-
~
c:
rated) to 0·31 cm 2 /day (oven-dry) and for soil-cement
J!l 80
c: Kc is in the range of 6 cm 2 /day (saturated) to
0
()
0'15cm 2 /day (oven-dry).lO The constant value of
~
~ 70 moisture diffusivity Kc used for prediction of moisture
'0
::2: loss by linear moisture diffusion theory is 0.3 cm2 / day
8
for concrete and soil-cement,IO and for lean concrete
it is 0·9 cm 2 / day. 10 The moisture diffusivity Kc is
found to be much higher than the chloride diffusivity D
3 4 5 6 7 in concrete, which is 8 X 10- 3 cm 2 / day. 11
n = x/J
Determination ofsurface factor
Fig. 6. Moisture content C versus variable rt
A repair mortar specimen of dimensions 10 cm X
2 cm X 2 cm, in which diffusion of moisture was from
The moisture diffusivity Kc is obtained by substitut- only one side of cross-section 2 cm X 2 cm, was ana-
ing for 17 and dC/d17 from equation (14) into equation lysed. The surface factor f was determined by compar-
(13) and integrating the result. The relationship be- ing the experimental moisture losses with the predicted
tween the calculated values of diffusivity Kc and moist- values using the one-dimensional finite element code
ure content C is shown in Fig. 7. It can be inferred MSTDIFF I. The values of moisture loss predicted
from Fig. 7 that the moisture diffusion from the interior using the diffusivity Kc as a function of moisture con-
of the repair material is closely related to the moisture tent C were found to be in good agreement with the
content of the repair material. More precisely, in the experimental values, for a surface factor f = o· 3
range of high moisture content, the diffusivity falls cm/day, as shown in Fig. 8. The surface factor f of
sharply with a decrease in moisture content, but for a this repair material is low compared to that of normal
moisture content below 60-70%, it may be considered concrete, 0·55 cm/day.8 Also, the moisture loss values
to attain a more or less constant value. predicted using the linear moisture diffusion theory
On the basis of the results shown in Fig. 7, a rela- with Kc = 0·3 cm 2 /day are shown in Fig. 8. The pre-
tionship between the moisture diffusivity Kc and moist- diction of moisture loss using a moisture-dependent
ure content C was obtained using non-linear regression diffusivity is found to be more accurate than the predic-
as follows: tion using a constant diffusivity, especially after about
15% of the moisture has been lost.
(15) The high rate of moisture loss, especially in the
initial period of drying, is attributed to the high value
of diffusivity in that period. Hence, it can be said that
where Ko is the diffusivity in the oven-dry condition the moisture diffusivity Kc is one of the major para-
(C=O%) and is equal to 0'1175cm 2 /day, and the meters affecting the moisture loss from a cementitious
regression parameters a and bare 0·05 and 1·878 repair material.
respectively.
100-,---------------------~
I

~~ 1Jl
§ !

~ 1~
:~ I
6 I -:;:;:::::::::::l'r
0'1~ x Experimental values
l
..... Experimental values - Non-linear theory
- - Linear theory
,1--"
0.01 .....
i
-....,,----,---r-....,--..,~._R,e_9 r_es....,s,ed_v_a,lue_s--f
i I j I
0+--,--.------,----,---,-_'---,--...,--,---1,
o 10 20 30 40 50 60 70 80 90 100 o 5 10 15 20 25 30 35 40 45 50
Moisture content C: % Drying time: days

Fig. 7. Relationship between moisture content C and difJusiv- Fig. 8. Experimental and predicted values of moisture loss at
ityKc x = 1 cm from the drying surface

Magazine of Concrete Research, 1997, 49, No. 181 289

Downloaded by [ Newcastle University] on [24/09/16]. Copyright © ICE Publishing, all rights reserved.
 Asad et al.

Relationship between moisture loss and free shrinkage
strains
The free shrinkage strains obtained from the experi-
ments are plotted against drying time in Fig. 9. It can
be inferred from Fig. 9 that the strains increase rapidly
during the initial drying period,' but, as drying pro-
gresses, the increase in strain becomes slow and the
strain reaches a constant or ultimate value at the end of
28 days.
To obtain a relationship between free shrinkage
strains and moisture loss, the free shrinkage strains
were plotted against the experimentally determined
moisture loss values at a depth x = I cm from the
drying surface, as shown in Fig. 10. According to
Fig. 10, the relationship between moisture loss and
free shrinkage strain is not linear. Thus, it is not
possible to predict directly the free shrinkage strain
from the moisture loss for this repair material, as has
been the trend for shrinkage modelling in ordinary
concrete. 6,12 The actual relationship between free
shrinkage strains and moisture loss, which is non-
2000,,----------------------,
~----------------~----------------
i!! ---
'(ij 1500-1 . content Ce as 50%. The thickness t used was 1 cm.
~
linear, is given by the following expression based on
a regression model:
M 25
(Csh)M = 195 + M25 (csh)ult (16)
This is a form similar to the ACI Committee 209
13
equation for the variation of shrinkage with time. In
the above equation, Esh represents the free shrinkage
strain, M the percentage loss of moisture starting
from 0% and (Esh)ult the ultimate free shrinkage
strain, which is taken as 1600 microstrain. Equation
(16) enables the prediction of free shrinkage strains
for different values of moisture loss for this particular
repair material.
Application of two-dimensional finite element model
In order to study the application of the two-dimen-
sional finite element model, a practical case which
simulates field conditions of applying a patch repair to
substrate concrete was analysed. The problem of two-
dimensional moisture diffusion from the repair material
was analysed using the experimentally obtained curve
for moisture diffusivity Kc shown in Fig. 7 and the
material was assumed to be isotropic so that
.r
K ex = Key. The surface moisture transfer coefficient
used was 0·3 cm/day, with the equilibrium moisture

i
I A repair mortar layer cast over a concrete sub-
strate, having a cross-section 50 cm X 2 cm, in which
the diffusion of moisture was from the top surface
'0001 and through the sides, was analysed. The repair sys-
tem was assumed to be very long in the third direc-
lE 500 tion as, for example, in the case of industrial flooring
Lt
or a bridge deck. Moisture cannot diffuse into the
O+-----,---.---,--J -,-'Ultimat,e shrink~ge strain! substrate concrete as it is assumed to be sealed at the
o 10 20 30 40 50 60 70 interface between the concrete and the repair materi-
Drying time: days
al. One half of the cross-section was analysed using a
two-dimensional finite element grid with 27 nodal
Fig. 9. Free shrinkage strain versus drying time
points and 16 four-noded elements, as shown in Fig.
11. The time-dependent changes in moisture content
along the section B-B (Fig. 12, inset) are shown in
2 0 0 0 - , - , - - - - - - - - - - - - - - - - -__ Fig. 12.
I

1800~, The results of the analysis indicate that the moisture

1600~ content at a particular time of drying is high in the
~ 1400~ interior and there is a significant variation in moisture
content at the ends. The variation in moisture content
e I
.~ 12001~ at the ends is due to the diffusion of moisture in two
~ 1000 directions, that is, through the top surface and through
~
'"
.S 800 the end, showing the effect of two-dimensional moist-
.c ure diffusion. The diffusion of moisture in the central
~
l.L
600
400l
l portion is predominantly through the top surface and
hence the moisture content in this region is almost
200l constant, at a particular time of drying. Therefore, it
O-¥---,----r--,-------,----,----,---,-----1 can be concluded that in the case of patch repairs the
o 5 10 15 20 25 30 35 40
Moisture loss: %
moisture diffusion is predominantly one-dimensional,
and even when the ends are exposed to drying, a major
Fig. 10. Relationship between free shrinkage strain and portion of the repair has one-dimensional moisture dif-
moisture loss fusion.
290 Magazine of Concrete Research, 1997,49, No. 181

Downloaded by [ Newcastle University] on [24/09/16]. Copyright © ICE Publishing, all rights reserved.
 Drying shrinkage in patch repair systems

y concrete substrate of size 200 X 50 X 50 mm, as shown

6 912 15 18 21 24 27
in Fig. 13, was analysed for shrinkage stresses.
3 A 24-element mesh with eight-noded isoparametric
2 4 6 8 10 12 14 16
T
20rm
plane-stress elements, shown in Fig. 13, was used to
model the repair materialjconcrete substrate system.
2 1 3 5 7 9 11 13 15
The effect of shrinkage along the width and thickness
1,.4 7 10 13 19 22 25 x of the repair layer was considered to be negligible, as
16 250 m
the width and thickness are very small compared to the
length, and is ignored by this model. The analysis
carried out in this study was linear elastic. The bond
(a) between the repair material and the substrate concrete
was assumed to be perfect. The laboratory-measured
y data relating the elastic and mechanical properties of
et the substrate concrete and the repair material are as
follows:
12 18 24 30 36 42 48
6 (a) concrete: compressive strength = 40 MPa, direct

I
5 10 15 20 25 30 35
5 tensile strength = 4 MPa, modulus of elasticity
4 9 14 19 24 29 34
4 = 30000 MPa, Poisson's ratio = 0·2
3
3 8 13 18 23 28 33 50mm (b) repair material: compressive strength = 25 MPa,
2 7 12 17 22 27 32 direct tensile strength = 3 MPa, modulus of elas-
2
1 6 11 16 21 26 31 ticity = 14000 MPa, Poisson's ratio = 0·2.
1 .7 13 19 25 250 mm 31 37 3 x
1

(b)
20 33 40 53 60 73 80 93
13 T
Fig. 11. Finite element discretization for two-dimensional 12 6 12 18 24 Repair material 10mm
11

i
moisture diffusion problem 10 5 11 17 23
9
8 4 10 16 22
7
6 3 9 15 21 Concrete 50mm
Shrinkage stresses in repair material/substrate 5
4 2 8 14 20
concrete system 3
2 1 7 13 19
The free shrinkage strains Esh were used as the input
to the finite element code STRSRSYS to find the A 14 21 34 41 54 61 74 81 x
1...~--------200mmt----------.l.1
stresses due to restrained shrinkage of the repair mater-
ial in the repair material/concrete substrate system. A Fig. 13. Finite element discretization for stress analysis in
thin repair layer of size 200 X 50 X 10 mm cast over a restrained concrete system

100..,----------------------r-----,
- 1= 50 days

-1= 40day$
90
.' 1= 30 days
_ 1= 20 days

__ 1= 10 days
.................................................................................1------1
..... 1= 5 days

B- ~-------""""B .l
,.'" ---
------------------------------

,'
- Repair material -"'" 20 mm

~
:.

//
T
60 ~.-._._._._._-_._._._._._.- Concrete

50
L
;.~ :;.:;.: .;;.~:;~.~~~~.~~.;-~- ;.;;. ~.;.;..~:.:.
o 2
I i
4
i
6
I
8
i
10 12 14
i I
16 18
Distance along length: cm
I I I
20
I
22
I
24 26 t - - - - - 500mm
""I
..

Fig. 12. Time-dependent changes in moisture content C along section B-B (inset)

Magazine of Concrete Research, 1997,49, No. 181 291

Downloaded by [ Newcastle University] on [24/09/16]. Copyright © ICE Publishing, all rights reserved.
 Asad et al.

Figure 14 shows the predicted values of tensile stres- 16 I

ses due to restrained shrinkage at the top surface of the
repair layer in the repaired system. The tensile stress
14~
I
values calculated using the experimentally measured '"
CL.
:::; 12
1
10~
Vi
restrained shrinkage strains in a corresponding re- en
~
strained overlay/ substrate sy~tem are also shown in <ii
2
Fig. 14. The predicted values of restrained shrinkage
stresses are in good agreement with the stress values
.;;;
2
c: 81
E
:J
6---j
calculated using the experimentally measured shrinkage E
I
I

'" 4~
'x
strains. :::;
The tensile and shear stresses in the repair material 2 I
close to the interface are shown in Fig. 14. It is clear
from Fig. 14 that the maximum tensile stress occurs at J 0
I
2 4 6 8
i
10
I I
12
I
14 16
the centre and the maximum shear stress occurs at the Ultimate shrinkage strain: 10- 4
ends.
Fig. 15. Variation of maximum tensile stress near interface
These results show very high tensile and shear stress
with ultimate shrinkage strain
values in the repair material, possibly leading to crack-
ing of the repair material at the centre and debonding
at the ends. The high stress in the repair material/ 3·0 MPa, which is equal to the direct tensile strength of
concrete substrate system is attributed to the high free the repair material. Therefore, for the repair
shrinkage strains in the repair material. material/concrete substrate system considered, the ulti-
mate shrinkage strain should not exceed 300 micro-
strain to avoid cracking of the repair material due to
Determination of allowable ultimate drying shrinkage of the repair material. By comparison,
the ultimate shrinkage strain for concrete at a relative
shrinkage strain
humidity of 40% ranges from 400 to 1100 microstrain,
2
The allowable ultimate shrinkage strain in the repair with an average value of about 500 microstrain.
material to avoid cracking, that is, the strain at which It must be noted that in the presence of tensile creep,
tensile stresses due to restrained shrinkage are within somewhat higher ultimate shrinkage strains (> 300
the permissible limits, is determined by using different microstrain) could be tolerated prior to damage due to
values of ultimate shrinkage strains, ranging from 200 cracking and/or delamination.
to 1500 microstrain. Fig. 15 shows the maximum ten-
sile stress values for different ultimate shrinkage strains
in a repaired system with a 10 mm thick repair layer
Conclusions
over a 50 mm thick concrete substrate, and in the con-
text of a linear constitutive model, the variation is It has been noted that the use of a moisture-depen-
linear. The allowable shrinkage strain was found to be dent diffusivity Kc in the study of moisture diffusion
300 microstrain, causing a maximum tensile stress of through concrete results in predicted values of moisture

Moisture diffusion

~~~~~~~ ..t.
1-_---:.R"'e"'D,a"'i:...:rm::..:=at::::;e:.:;ria::;.I_---I1 0 mm

Concrete T
50mm

~ Tensile stress at top (experimental)

1
1-l1-1----200 mm -----1~~1
-+- Tensile stress at top (predicted)

~ Tensile stress at interface (predicted)

6
~ Shear stress at interface (predicted)
4

25 50 75 100
Length of specimen: mm

Fig. 14. Shrinkage stresses in repair material due to free shrinkage strain of 1500 microstrain

292 Magazine of Concrete Research, 1997,49, No. 181

Downloaded by [ Newcastle University] on [24/09/16]. Copyright © ICE Publishing, all rights reserved.

content in closer agreement with those obtained experi-
mentally, in contrast to those obtained by the use of a
constant diffusivity. The moisture diffusivity Kc is high
in the initial period of drying, that is, at high levels of
moisture content, decreasing rapidly as the moisture
content decreases.
The phenomenological relationship established be-
tween moisture loss and free shrinkage strain is not
linear, with a rapid increase in strain in the initial
drying period followed by a gradual reduction prior to
reaching an asymptotic value. This behaviour indicates
that direct solution of a shrinkage diffusion equation in
which the moisture content is simply replaced by a
shrinkage strain will not yield results of the desired
accuracy. The regressed form chosen to represent the
experimentally determined relationship between free
shrinkage strain and moisture loss is observed to be
similar to the ACI e~uation for the variation of shrink-
age strain with time. I
The higher the ultimate free shrinkage strain (csh) of
the repair material, itself a function of environmental
humidity, the higher are the induced tensile and shear
stresses in the repair layer and the interface. An accept-
able bound for the ultimate free shrinkage strain was
300 microstrain for the patch repair system investi-
gated, and this bound can be relaxed upwards in view
of the beneficial effect of tensile creep that would oc-
cur in the system.
In addition to the diffusivity Kc and the ultimate free
shrinkage strain Csh, the other parameters found to in-
fluence the magnitude of stresses include the modulus
of elasticity Er of the repair material and the degree of
restraint exerted by the concrete substrate. Stiffer repair
materials, that is, those with high elastic moduli, devel-
op higher stresses for the same level of free shrinkage
strain in comparison to softer or more flexible repair
materials.
Acknowledgements
The authors acknowledge the support and coopera-
tion of the Department of Civil Engineering at King
Magazine of Concrete Research, 1997,49, No. 181
Downloaded by [ Newcastle University] on [24/09/16]. Copyright © ICE Publishing, all rights reserved.
Drying shrinkage in patch repair systems
Fahd University of Petroleum and Minerals in the pur-
suit of this work, in terms of release time and use of
laboratory and computer facilities.
References
I. MEHTA K. P. Concrete - structure, properties and materials.
Prentice-Hall, Englewood Cliffs, 1986.
2. EMMO s P. H. and VAYSBURD A. M. Factors affecting the dur-
ability of concrete repair: the contractor's viewpoint. Construct.
Bldg Matel:, 1994,8, o. 1,5-16.
3. EMMONS P. H., VAYSBURD A. M. and McDoNALD 1. E. A rational
approach to durable concrete repairs. Concr. Int., 1993, IS,
No. 9, 40-45.
4. EMMONS P. H., VAYSBURD A. M. and McDoNALD 1. E. Concrete
repair in the future turn of the century - any problems? Concr.
Int., 1994, 16, No. 3,42-49.
5. EMMONS P. H. and VAYSBURD A. M. The total system concept -
necessary for improving the performance of repaired structures.
Concr. Int., 1995, 17, No. 3,31-36.
6. PICKETT G. Shrinkage stresses in concrete. JAm. Concr. Inst.,
1946,17, No. 3, 165-204; 0.4,361-398.
7. ASAD M. Computational modelling of shrinkage in repaired
concrete. Master's thesis, King Fahd University of Petroleum
and Minerals, Dhahran, Saudi Arabia, 1995.
8. SAKATA K. A study on moisture diffusion in drying and drying
shrinkage of concrete. Cem. Concr. Res., 1983, 13, No. 2,
216-224.
9. CRANK 1. The mathematics of diffusion. Clarendon Press, Ox-
ford, 1975.
10. PENEV D. and KAWAMURA M. Moisture diffusion in soil-cement
mixtures and compacted lean concrete. Cem. Concr. Res., 1991,
21, No. I, 137-146.
11. NAvAz C. M. Chloride difjilsion in concrete/Prediction of the
onset of corrosion in reinforced concrete structures. Master's
thesis, King Fahd University of Petroleum and Minerals, Dhah-
ran, Saudi Arabia, 1994.
12. IDING R. and BRESLER B. Prediction of shrinkage stresses and
deformations in concrete. In Fundamental research on creep
and shrinkage of concrete (ed. F. H. Wittmann). Martinus Nijh-
off, The Hague, 1982, pp. 341-352.
13. ACI COMMITTEE 209. Prediction of creep, shrinkage and temp-
erature effects in concrete structures. American Concrete Insti-
tute, Detroit, 1994, Technical Report, ACI Manual of Concrete
Practice.
Discussion contributions on this paper should reach the editor by
26 June 1998
293