Materials and Structures, 1993, 26, 532-548

Cover cracking as a function of rebar corrosion:

Part 2 Numerical model
F. J. M O L I N A
Universidad de Oviedo, Departmento de Construcci6n, Campus de Viesques, 33204 Gijon, Spain
C. A L O N S O , C. A N D R A D E
lnstituto de Ciencias de la Construcci6n Eduardo Torroja, Serrano Galvache s/n, 28033 Madrid, Spain

A numerical model based on standard finite-element techniques is proposed for the simulation
of crackin9 in concrete specimens when subjected to corrosion of their reinforcement. A
smeared-crack approach is used to model the behaviour of the concrete finite elements, while
the corrosion, which is understood as the applied load on the structure, is modelled by a
combination of initial strains and chanye of elastic properties, which are respectively
equivalent to the expansion and softenin9 of the steel elements at the rebar surface when they
rust. The model is applied to four examples, which were simultaneously tested experimentally
and reported in Part 1 of this paper, and the influence of the main parameters of the model in
the response is studied separately. In particular, the effect on the crack-width rate of the
specific volume of the rust bein9 formed is quantitatively assessed.

1. I N T R O D U C T I O N incomplete because they do not include the rebar as a

part of the model and, moreover, before the problem is
The degradation of a concrete structure by the corrosion
solved experimentally or by another model, the applied
of its reinforcement is clearly a very complex phenomenon
load in each case (pressure or dilatation) cannot be easily
which involves multiple causes and effects. From the
related to the time the structure has been exposed to the
purely mechanical point of view, it consists of a loss of
corrosive conditions.
the rebar cross-section which is transformed into rust.
In our approach, the applied load, say x (independent
These products of corrosion, depending on several
variable), will be the radial loss, or attack penetration,
environmental factors, can have different properties, but
in the rebar. This penetration will be assumed uniform
normally their specific volume is greater than that of the
along the perimeter of the bar. The results of the analysis
virgin steel; the result is a dilatation of the bar (steel + rust)
will be the evolution of the induced cracking in the
and consequently the cracking of the surrounding
concrete as a function of that increasing parameter.
concrete. To apply the model to a real situation, the speed of
To study this phenomenon, some experimental studies
corrosion must be known as a function of both the
have been carried out which, in general, show some
environmental conditions and the type of structure, in
interesting features, but suffer two important limitations:
order to substitute x by time and then obtain useful
(i) the difficulty of isolating factors, due to the information concerning the service life of that structure.
complexity of the phenomenon, and However, at this stage, at least two factors of possible
(ii) the cost and duration of the tests which can take significance are not yet covered by the model. These
years or at least months, in the case of artificially are, on the one hand, the dependence of the said speed
accelerated corrosion [1]. of corrosion on the self-induced cracking and, on the
other hand, the possible partial expansion effectiveness
Obviously, the interest in the development of a numerical of the formed rust due to its diffusion through the
model for the study of this problem lies in the possibility concrete.
of overcoming these drawbacks, after the necessary
agreement between experiments and computations is
achieved.
2. N U M E R I C A L M O D E L
Finite-element (FE) modelling of cracking in concrete
is currently still under development, but some of the The proposed model tries to cover the mechanical aspects
available models are probably accurate enough for the of the phenomenon under study by the use of a
object of our study and have, in fact, been applied to it. combination of several types of load and material beha-
For example, Sygula and Ryz I-2] and McLeish I-3] use viour which are currently available in FE techniques.
FE modelling for the concrete part of the specimen, while In fact, this model was previously presented by us [4],
the corrosion is modelled by a uniform pressure or but without any experimental data with which to
dilatation. However, both models could be considered as compare.
0025-5432/93 ~c; RILEM
Materials and Structures 533

2.1 Simulation of corrosion However, we now propose to introduce a time lapse

between both effects, so that an element is not expanded
Because corrosion is not a standard type of load in the
before being fully softened, since in this way some
FE codes, its implementaton is made in a special way.
unrealistic tensions inside the rebar are eliminated.
The corrosion of a steel element is modelled by the
superposition of two effects: a decrease in its stiffness,
and an increase in its specific volume. 2.2 Simulation of cracking
The first effect is achieved by a linear variation of the Within both basic categories (discrete and smeared) of
properties of the material from those of the steel to those crack models for concrete developed in FE analysis, the
of the rust. The latter properties, considering the present smeared-fixed-crack model with linear softening proposed
lack of information, are assumed nearly equal to those by Rots e t al. i-63 has been chosen by us because of its
of the liquid water (which is one of the main components relatively easy implementation and its clear presentation.
of the rust under these conditions), i.e. The fundamentals of this model are included in Appendix
vr = 0.5 (1) B for completeness.
However, hopefully, the type of crack model imple-
for the Poisson's ratio and
mented for this problem should not be the primary
K r = 2.0 G P a (2) determinant with respect to the results, since mode I
for the bulk modulus. It is not possible to use Equation 1 fracture is the type of crack mainly observed in the
directly within the FE displacement method, but it can examples and both discrete and smeared models usually
be replaced by a value slightly lower than 0.5 1-5] and give coincident results for such a mode of fracture E73.
then used to compute a Young's modulus
Er = 3(I - 2vr)K ~ (3) 3. EXAMPLES
which is small, but not zero. Obviously, these values from 3.1 Materials and experimental results
Equations 1-3 of the properties of the rust should be The first part of this work [1] contains the results of
replaced by others which are more realistic if a way can four corrosion tests which were carried out to obtain
be found to obtain them. useful experimental data in the validation of the numerical
Both virgin steel and rust are considered within the model proposed here. That information is summarized
model as only one material with properties varying as follows.
linearly. The specimens were made of a 30 M Pa resistant
The second effect (increase in volume) is achieved concrete with a measured splitting tensile strength of
by the imposition of an initial strain on the element being 3.55 MPa. In order to induce uniform accelerated
corroded. Calling corrosion in the rebar, a fixed electric current was
t'~/~ = u~/v~ (4) applied between the bar and a counterelectrode. Assum-
ing that the entire current I was consumed in the
the ratio between the specific volumes of rust and steel,
oxidation, the attack-penetration increment was estim-
the isotropic initial strain to impose on the diagonal
ated as
components of the strain tensor is
Ax = 0.032I At (gm) (7)
exo = eyo = eo _ (vr/~ - 1) (5) where I was 100 laA cm -2 (or 10 pA cm -2 for specimen
3
IV) and At was the time increment in days. The initial
when infinitesimal strain theory is applicable. Unfortun- value of radius loss was also estimated by measurement
ately, the said theory is not strictly applicable because of the polarization resistance [1].
the value of Vr/~ is about 2, as has been observed in the The histories of the strains, measured by the strain
examples of this work. Nevertheless, the infinitesimal gauges at the surface, are plotted in Fig. 1, where the
strain model is recommended by us due to its simplicity abscissa represents the time, as well as the attack
and because, surprisingly, even in the case of an important penetration x estimated by Equation 7, and the ordinate
change in volume, considering certain present conditions represents the strain, as well as the crack width estimated by
of symmetry, the initial strains (Equation 5) can still give
w = LE (8)
a good approximation for this problem, as justified in
Appendix A. e being the strain at the gauge and L its length (30 mm
As we proposed [4], both corrosion effects can be in this case). For compactness, Fig. 1 also includes the
obtained simultaneously by a single thermal load if the analytical results obtained with the numerical model
FE code allows temperature-dependent properties as well (dashed line), but the comparison is postponed until the
as thermal expansion. In that case it must be verified, discussion.
where ~ is the coefficient of thermal expansion and AT
the increment of temperature, that 3.2 Applied model and analytical results
~AT- (vr/~ - - 1) (6) Within the numerical model, the values adopted for its
3 parameters were the following:
 534 Molina, Alonso and Andrade

STRAIN-GAGE

,crock
width(mm)
-0.15
4xlO 3
g- ..'=" X@

I0

u) 3xlO 3 -0.1
Z
0
I--
<
'~" ." q / I /
~ 2xlO 3- .x.- , / i /
hi -../- I ff ~ /
X:7": i / I X
121
::k -0.05
; I I

10 3 . / I/ i x~

0 .... U ~- ~ -0
I I

8 lO 12 14 16 18 20 2'2 2'4 2'6 2'8 S() 32 S'4 TIME (doys)

1Z9
1.9> 81s 'zi.1 33.9 46.7 5;5 7:~.3 8;.1 9-Z9 l./.m(x)
(a)

@|174174
~ _,:_.,____,_;
__,_(_/~
_: o ...... |162 ,crack
width [mm)
-0.15
4xlO3
,., ..'"

,@:..~/""
I0
x
v
m 3 xlO ~
Z
-0.1
0
I....-

a: 2xlO 3
ou_ -,-,
uJ

::k
._~..-~ .0.05
I0 ~- ~.." ,, /,,,F//
9= i

9. ~ 0----~@
~- ~| -o
6 =
; 4 ~ 8 ,0 1~ 11~ 16 18 20 z'2 24 26 28 3o 32 34 T,ME(doy,)
14.4
2.1 >
81s ' 2i.s s;,, 4~.9 s;.~ ~,~.5 s~.3 9M 1~6.9 ~m(,)
(b)

Fig. 1 Experimental results from Part 1 rl]: (a) specimen I, (b) specimen II. (continued)
 Materials and Structures 535

5xlO 3
/r </ @@|

9174174
crack
width (mm)
-0.15
4xi03

3xi03
-0.1
0

o 2x103.

::L
/ -0.05
1(33 .

I
0

6
24.5
:~ 4 s 8 lb 12 lb, 1~ 1'8 2b z'2 "~,MECdoy,)
21.3>
-~m(x)
i i

277 ~;3.s s~.s 6;1 z;9 9i~

(c)

9174174

5xlO 3 ,crack
| width(ram)
-0.15
4x103.
g- ,. , . , . "

b
v
x ...... , .....,......'""'"'"""" "'"'"" ..-'~
r 3x103
Z
-0.1
o
I--

oc 2 xlO 3-
0
u.
W C)
::L
10 3 .. ........ -0.05

0
0

b ,i + 1~ 1~ 2'0 z'4 2'0 iz g6 43 4'4 4'8 dz d6 6b 6'4 6'8 ~,'z ~JME(doy,)

10.I > , i 17.4
i i i i = i ! I

13 15.2 17.8 20.3 22.8 25.5 28 :30.6 33.1 ~m(x)

(d)

Fig. 1 continued (c) specimen 1II, (d) specimen IV.

536 Molina. A l o n s o a n d A n d r a d e

(i) For the virgin steel (elastic)

E~ = 210 GPa v~ = 0.3 (9)
(ii) For the rust (elastic)
K r = 2.0 G P a v, = 0.499 vr/~ = 2.0 (10)
(iii) For the concrete (smeared-crack)
/" t \ "-\
E~ = 36 GPa vc = 0.2
(11)
Jlt = 3.55 M Pa Gf = 200 J m - 2 fl = 0.2
where the properties of the rust were estimated as
explained above and the properties of the concrete were
either measured (in the case of the tensile strength fct)
or taken from the literature. It must be observed that the
measurement of properties like the fracture energy Gf or
the shear retention factor/7 would have required very
specialized equipment [8] which was not available to us.
A mesh of eight-node plane-strain elements was created
to model the cross-section of each of the specimens (Fig.
2) using three layers of very thin elements (20 p.m, not
visible in the figure) of rustable steel at the surface of the (b)
rebar in order to properly graduate the simulation of
uniform corrosion, layer by layer. In the case of the mesh
for the first specimen (Fig. 2a), the ideal symmetry was
altered by introducing on the left a cover 1 mm greater
than the nominal, just to force the appearanc ~. of the
main crack in the ascending direction (12 h in a clock
reference) as was observed in the test.
The corrosion was simulated by the application of
three cumulative load combinations, each representing
an advance of 20 p.m in the attack penetration. The
integration was performed by the incremental version of
the modified Newton-Raphson algorithm 1-5] with simul-
taneous checking of force and displacement errors and
by using a secant-stiffness approach.

"x, \ / .,."
x,x ...-~....</

(c)
,/ ,/ "\, "\ Fig. 2 (continued) (b) specimens II and IV, (c) specimen III.

i i
Figs 3a, 4a and 5a show the final deformed meshes
I
with the resultant cracking for x = 60 p.m. At each
I integration point, a segment represents the direction of
cracking (if any), its width being proportional to the
crack width using the same magnification as in the
displacements (30 times). Consequently, even though
the cracking is present in all radial directions, it is only
I noted in those directions where significant crack widths
are developed, that is to say, using a clock reference:
(a)
(i) Specimen I: main crack at 12 h, secondary crack at
Fig. 2 Finite-element meshes: (a) specimen I. (continued) 4.30 h.
 Materials and Structures 537

,,~ J STRESS Nlmrm2

4.8fl8 I

.g"

! 2.8~8~ ~,'' . , ',,

2' , .

I B88 :';' " '"s

(a) (b) Penetration x mm

STRESS N/mm2 STRAIN rr~/~

4.888
.. 8. ..............
8.889 ..............
! ..............
,~ ..............
i ..............
,~ ..............
i..............
:..............
i .............
i ........... 7,'-I
?
~ ~ - ~ . : : : . ........

8.888 ............... *, . . . . . . . . . . . . . . ~, . . . . . . . . . . . . . . ',. . . . . . . . . . . . . . i-'";'- ........ ~,

3..11. il; "- ~'.~ 8.887 .............. ; ............ " .............. " ........... ~-" ....... ,.:2 .... '
: ' ' ~ " J/ i
8.~86-' i ! i .- ~ . t . , / :
.............. T. . . . . . . . . . . . . . ,~. . . . . . . . . . . . . . , " - ' - . . . . 5 7 -7 . . . . . . . . . . . . . . 1
i." ! "/
2.088 .' 8.~s4- ............ ;............. ~. . . . . . . . . . . -- -~-/-<- . . . . . . . . ~. . . . . . . . . . . . . . . '
: L. / ' , , :
11.~e4 . . . . . . . . . . . . . . . i.............. i---.---;- ~. . . . . . :. . . . . . . . . . . . . . i . . . . . . . . . . . . . . !

/
B.083 .............. : ........... .-~'- ............. i.............. i.............. i

o. 1182 ............... 1 ...... --'-_ ..... ',-.............. i .............. ~' .............. ,'

8.fl81-, ........... ,": i

-,-a.-, ........................................... ~- ............. 3"" ~.-,.r:- ~ - ' = : ......

_~,' ::::::::::::::::::::::::::: ..' .....

0.8 8 -;-'-''- ~ -: ....
..~4. e.ks.
(c) pENFTRATION X (d) peNetrarIO. •

PRESSURE Nlmm2

IS.888-

,/ i

"'"~%oe ..... 8.~10 ,.~2, R.~3o ,.~*o ,.~s,

PENETRATION X mm

(e)

Fig. 3 Results for specimen I: ( a ) d e f o r m e d mesh with cracks, (b) normal s t r e s s a t m a i n c r a c k , (c) n o r m a l stress at secondary
crack, (d) average normal s t r a i n , (e) p r e s s u r e at rebar.
538 Molina, Alonso and Andrade

STRESS N/mm2

4.886

3.008
%..

/ \ r ". .,.,,,,

2.000

1.888
, x, ,.

1 ',

8"8~%88 a.,18 0.~88 6.~38 ' . . . . . 8.<148 8.os~

(a) (b) PENETRATION X r~n

STRESS N/mm2 STRAIN mm/mm

"888t r--~~ .....

fl. 818-

"'- ............... i .............. i .............. i ............. ! ............. :i

8.888 .............. ~ .............. ~ .............. ',. . . . . . . . . . . . . . . ~ .......... ~--'.

8 88, .............. 7 .............. F.............. i.............. ~r: . . . . . . . . . . . .

6.886 ............... ; .............. " ............. -- .......... -~-~ ...........

"'--. .

2.088 8.088............... i.............. ,

',~.............. i:......
/.
: .......
/ ' ,
~-~---- .......

1. 888

e 881~ .......... =.~.~:'- ........... ~ ............................ ~,---:.'~----':~

8"8~8!i:i" 8.418 8.~20 0.438 8.~48 _8.~80 8"8~".~0~ . . . . . . . ~:~la ' 8.;:8 ' 8.~3, " 8.~,8 8.~$8
(c) PENETRATION X inn (d) PENETRATION X rn~

PRESSURE N/Inm2

15.868

18.888

s.ooo //

8"88%88 8.4a0 8.428 0.438 8.~48 8.458

PENETRATION X rr~

(e)

Fig. 4 Results for specimens II and IV: (a) deformed mesh with cracks, (b) normal stress at main crack, (c) normal stress at
secondary crack, (d) average normal strain, (e) pressure at rebar.
 Materials and Structures 539

....7 7 - ' I
STRESS N/mm2
4.888

! ~{'\,
C- -
'-, ,xx~'-"'- - 2 . BeB /:" ...........i i i i ~

1,888

\,i'\
". ~h \
~,...... )_.?a. .....................................
B.,~a. ' 8.~1, e.~2, o.g32 e.~4e e.~se
(a) (b) PENETRATION X mf~

STRESS N/mm2 STRAIN ,lfil/~lm

~"~'[ .............T..............[..............::
..............i.............. !
'.','r ............. i .............. i ............. + ............. ! .............
! [
T l
~.,aTJ- ............. ~. . . . . . . . . . . . . . ~. . . . . . . . . . . . . + ............ i" . . . . . . . . . . . . .
'. ', IE .-" I ',
/ hQ~ "-,

2.'88- ii / '
" -,~...
: : / : : i
'"e"i .............. t. . . . . . . . . . . . . . - " ~ ?
! ..... ........... !
i i~ ~ : : i
I / ' : : :

] .,'/1/1/ : 1 : :

, / / ~ - : z .. ', .... .................... F'" I

BB" ~ .e1 8 8. ~ 8.B1B
~ .............. e.B2e
~........ :........ g.1~38
' 8.a48
' ,.1~'$8
(C) PENETRATION X mm (d) PENETRATION X m~

PRESSURE N/mm2
15.868

18.888
/ ...........

S'BSOI ]/

e.~28 e.~3e ~.~4~ ~.es8

PENEIR~TION X ~n

(e)

Fig. 5 Results for specimen III: (a) deformed mesh with cracks, (b) normal stress at main crack, (c) normal stress at
secondary crack, (d) average normal strain, (e) pressure at rebar.
540 Molina, A l o n s o and A n d r a d e

(ii) Specimens II, lII and IV: main crack at 12 h, 1

secondary cracks at 4.30 and 7.30 h.
It is observed that the crack widths are not fully localized,
/ -/
which is a common occurrence in the smeared-crack
model [6] and, probably, is also an analytical interpreta-
tion of the lack of uniformity in the cracking pattern
along the depth of the test specimens. The asymmetric
crack distribution obtained for specimen III (Fig. 5a) is
possibly due to the intrinsic instabilty of the softening
model.
Figs 3b, 4b and 5b show the evolution of the
main-crack normal stress registered in the three elements
of the concrete cover. According to these results, the crack
is initiated (microcracking at the deepest element (solid
curve in the graphs) and then is propagated to the
external ones which reach the tensile limit (fct = 3.55 MPa)
a bit later. However, the crack first becomes totally Fig. 6 Location of simulated strain gauges.
opened, consuming the whole fracture energy (macro-
cracking), at the most external element. For all the
specimens, the main crack is initiated and propagated
rates w/x, obtained by multiplying the strain rates by the
between penetrations of about 2 and 8 ~tm, the normal
assumed common gauge length of 30 mm. The sum of
stress becoming zero for penetrations which range from
width rates is also necessarily observed as rather stable
20 to 50 lam.
among the specimens. In fact, if the fracture energy is
In the same way, Figs 3c, 4c and 5c show the evolution
completely released in all the cracks and an ideal
of the secondary-crack normal stress registered in the
stress-free state is assumed, the problem becomes a purely
first three element layers that surround the rebar. It is
geometric one and the sum of width rates in all the cracks
observed that the secondary crack is initiated (micro-
cracking) at the same time as the main crack (x - 2 tam) is almost exactly
at the steel-concrete interface, but its propagation to
the surface is a bit slower and its full opening can wi = 2rt(Vr/s - 1) (12)
take much longer. Physically, for these examples, the said X
full opening of the secondary crack would mean the
detachment of a piece of cover, since no stirrups were as shown in Appendix C, which, by Equation 10, for these
present. examples becomes
In order to obtain results to compare with those of the
tests, the measurement at the gauges was simulated by
~ w i = 6.3 (13)
computation of the relative displacement between two X
nodes (one at each Side of the main crack in the surface)
divided by 30 mm. The evolution of this average strain This value is greater than those recorded in Table 1,
is represented as a solid line labelled 1 in Figs 3d, 4d and probably because the latter include only three cracks
5d. As in the experimental case, the conversion from strain instead of all, and the fracture energy has not yet been
to crack width is made by use of Equation 8. These same completely consumed in the model. Anyway, what is
three curves have been copied in Fig. 1 as dashed lines significant is that Equation 12 could be used as a practical
in order to establish a comparison between experimental
conservative bound for the asymptotic crack growth with
and analytical results in the discussion.
respect to the radius loss, at least within the assumptions
The same Figs 3d, 4d and 5d also include the strain
of the numerical model.
registered at other simulated gauges (not existing in the
Finally, Figs 3e, 4e and 5e contain the evolution of
test), labelled 2 and 3, and the summation labelled Z, of
the three strains. The exact situation of all the gauges is
shown in Fig. 6. It can be observed that, although the
strain at the main crack presents important variations Table 1 Analytical crack-width rates for x = 0.05 mm
from specimen to specimen, the summation of strains at Specimen Specimens Specimen
the three principle cracks is very similar among all the I II and IV III
specimens. In quantitative terms, considering that the
curves are more or less straight lines which intersect I~l/X (mm- 1) 0.154 0.120 0.137
the origin, Table 1 displays the secant rate of strain wx/x 4.6 3.6 4.1
(obtained by dividing the final value in the graph by a ~3 (e,/x) (mm -1) 0.186 0.168 0.178
penetration of x = 0.050 mm). Z~(wi/x) 5.6 5.0 5.3
The same table includes the corresponding crack-width
Materials and Structures 541

pressure, computed by could be due to a migration of the rust through the

opened crack [1]. Finally, for specimen IV (Fig. lb), the
P= ax+ay (14) slope of the experimental data is only well reproduced
2 initially, particularly before x ~ 23 gm, since afterwards
the experimental slopes grow considerably.
where a x and ay are the in-plane normal stress components
registered at the six elements in the centre of the rebar.
This result is included here for completeness, because in 4.2 Interpretation of results
other tests the corrosion is simulated by an applied
pressure on a fluid which fills the bar [9]. It is observed Although the proposed model needs more comparison
that the pressure presents a maximum (between 11 and with experimental data for its validation and for the
14 MPa, depending on the specimen), which can constitute adjustment of its parameters, the results of the examples
a serious impediment for a pressure-control test since, at that have been treated could supply interesting informa-
that point, the test would finish abruptly. tion about the phenomenon (see with respect to this the
study of parameter sensitivity contained in Appendix E).
For example, if the growth of the main crack is
4. DISCUSSION considered as the principal determinant for service life of
4.1 Validity of the numerical model the reinforced concrete element, the most important
datum to investigate is probably the specific volume of the
The proposed numerical model is mainly based on rust being formed, once the speed of corrosion has been
established numerical approximation techniques. Hence, estimated. It has been shown by the results of Appendix
its consistency has been only partially tested (Appendix E that, within the numerical model, only a change of that
D). However, the generality of the said consistency specific volume produces a strong effect on the crack-
follows from the proper use of such techniques and the width rate.
checking of the assumptions introduced. In relation to the effect of the mechanical properties
On the other hand, prior to the comparison of the of the concrete it can be said that, for a constant geometry,
analytical and experimental results (Fig. t), it must be they may only affect the evolution of cracking during the
kept in mind that some of the parameters which can initial phases of crack creation at the interface with the
strongly affect the results were arbitrarily chosen in the rebar and its propagation to the surface, but not during
numerical model. This is especially the case for the specific the subsequent crack growth. With the aid of other
volume of the rust which is a determinant variable for studies, we have confirmed that a ductility number, such
the strain rate, as shown in Appendix E. Another as the one indirectly proposed by Hillerborg et al. [10],
observation to be made is that the relation between days and whose definition could be
and penetration (Equation 7) as well as the initial
value of the penetration were obtained under simple ~)- GfEe (15)
assumptions which could be strengthened in future work. Lsf ~t
There also exists an unknown magnitude in these figures
which is the offset in the initial value of the experimental L s being the size of the specimen section, is the one that
strain. controls the initial delay in the crack growth which was
For the first specimen (Fig. la) the analytical strain observed when the Young's modulus was diminished.
rate is perhaps too high, but the existing scatter in the However, with respect to the maximum pressure developed
experimental values for this specimen does not allow a in the rebar, the tensile strength is determinant if the
definitive conclusion. However, for the second specimen geometry is not altered.
(Fig. lb), the experimental dispersion is low and the
analytical strain rate is of the order of the experimental
5. CONCLUSIONS
one. Nonetheless, the experimental curves present a
growing slope which could not be easily reproduced by From the comparison of the experimental and analytical
the present numerical model. We believe that the initial results, a definitive validation of the proposed numerical
value of the experimental strain rate is comparatively model is not yet possible, but it can be said that it is
small due to a diffusion of the earlier formed rust into capable of reproducing some of the main characteristics
the surrounding porous concrete. When the latter is of the experimental behaviour with success. Further
saturated, the expansion effect is consequently larger. In comparison with new experimental data and improvement
fact, with other tests of controlled expansion not of the model are needed as well as the development of
involving corrosion (not yet reported), we have obtained testing procedures for adjustment of the most significant
the uniform type of response predicted by the numerical parameters, such as the specific volume of the rust.
model. For the type of problem that has been studied, a
For specimen Ill (Fig. lc), the dispersion is low and determinant influence of the said specific volume on the
the experimental strain rate is relatively well reproduced crack growth has been shown by the numerical results
by the model. The experimental decrease of the rate after of Appendix E and also by the theoretical development
x = 40 gm is not observed in the analytical results and made in Appendix C based upon simple geometrical
542 Molina, A l o n s o and A n d r a d e

considerations. As is clear in Equation 12, the asymptotic Thus, the imposition of an initial volumetric strain of
global crack-width rate with respect to the penetration value
of corrosion is proportional to the relative increase in e ~ = (Vrls - - 1) (A7)
volume of the steel when rusted, 2re being the factor of
proportionality. Such an asymptotic global rate could be as made in Equation 5, is equivalent to imposing
used as a conservative limit for the width rate of the main ~o = ( v , i s _ 1) (a8)
crack; it does not depend on other properties of the
materials or on the geometry of the specimen (including as was to be shown.
the depth of the cover), which only affects the way the
global width rate is distributed among the cracks.
However, the experimental results allow us to see the APPENDIX B: Fundamentals of the adopted smeared-
relevance in the phenomenon of some factors not yet crack model
included in either numerical or geometrical models. One The adopted smeared-crack model [6-] is based on
of these factors could be the diffusion of the rust within continuum mechanics and in its formulation is very
the concrete and through the cracks, which could indeed similar to an elastoplastic model. When the cracking
be affected, for example, by the geometry of the specimen. criterion (maximum principal stress criterion) is verified,
These preliminary conclusions must be confirmed once a crack is created in the direction normal to the principal
the model has been better elaborated and extensive stress and the strain tensor is decomposed into
parametric studies have been carried out, possibly leading
to new and interesting conclusions. = ~e + ~r (B1)

where the first term on the right is the elastic strain and
APPENDIX A:. Justification of the applicability of the second one is the crack strain. While the relationship
small-strain theory between the stress and the elastic strain obeys the linear
isotropic elastic law, the relationship between the said
Using here for convenience cylindrical coordinates, stress and the crack strain is defined as an orthotropic
within the theory of infinitesimal strain, the volumetric softening law which depends on the history of the crack
strain is defined as and is given in the differential form
e = ~, + eo + ~- (A1) ds Cr = D cr de c' (B2)
while, if large strain is considered, the unit volume change where s Cr are the components of the stress in local
is better approximated by coordinates of the crack, e ~ are the components of the
crack strain in the same reference and D r is a diagonal
~ = (I + e,)(1 + ~o)(1 + E.) - 1 (A2) matrix. The non-zero coefficients of this matrix are
In order to show how e can give a good estimation of ev obtained from the diagrams of behaviour in normal and
for this problem, it will be considered firstly that a plane tangent directions to the crack reproduced in Fig. B1,
strain model is applicable to the cross-section of the where linear softening is adopted.
The area under the line of Fig. Bla is G f / h , where Gf
specimen, as is reasonable in the examples here considered,
so that is the fracture energy of the concrete and h is the crack
band width for the smeared strain within the FE mesh.
~. = 0 (A3)

and secondly that, taking the origin of coordinates in the

centre of the rebar, axisymmetric dilatation of the rust
is credible, so that

IAr ErX
e0 --- -- ~ - - (A4) Gt
r r

where r is the radial coordinate, u, is its associated

displacement and the strain of the steel is assumed much
smaller than that of the rust. Under these assumptions
and limiting the analysis to a maximum penetration of
the order of r/100, it is clear that

~r
e0 < - - (A5)
100 (a)

and then Fig. B1 Stress-strain behaviour in the crack (after Rots e t al.
e ~ ~v ~- e, (A6) [6]): (a) normal traction versus normal strain. ( c o n t i n u e d )
Materials and S t r u c t u r e s 543

~'-nt volume of the rusted bar is

dV, et = (Ur/s -- 1)2rcrH dx (C1)

/ which corresponds to an increment of its external radius

dR --
dV"e'
2~RH
-- (/)r/s -- 1) r dx
R
(C2)

and to an increment of its external perimeter

2re dR = 2rC(Vr/s- 1) r dx (C3)

R
r being the current radius of the virgin steel and H the
depth of the plane model. Now, assuming that the
(b)
elongation (Equation C3) is absorbed in the concrete by
Fig. B1 (continued) (b) shear stress versus shear strain. merely incrementing the crack widths, it follows that

dwi= 27c(v,/s -- 1) r
Z d~-~ ~ (C4)

Improved recommendations for the selection of h as a which results in Equation 12 if r/R is approximated by
function of the type of element used are also given by unity and a backward integration is performed to zero.
Rots [11]. The slope of the line in Fig. Blb is

APPENDIX D: Study of model consistency

G~ =flEc (B3)
In order to test the consistency of the proposed numerical
where fl is the shear retention factor and E c is the Young's model with respect to some of its parameters which are
modulus of the concrete. partially arbitrary, specimen I was re-analysed with
different data and the results were compared with those
contained in Fig. 3.
APPENDIX C: Derivation of the formula for an asymp- First, a refined mesh model with twelve layers of
totic crack-width rate rustable steel elements and five layers of concrete
elements in the cover was analysed. Not one of the
Assuming that a finite number of cracks are completely
material properties (Equations 9-11) was altered. The
opened in the concrete that surrounds the rebar and the
results fitted quite well with those of the original mesh,
state is free of stress, the differential increments in the
except perhaps for the final part of the stress at the
crack widths will be labelled dwl, dw2, dw 3. . . . and
secondary crack which opened more slowly with this
the associated uniform penetration of corrosion, dx, as
refined mesh. The crack patterns were also comparable.
shown in Fig. C1. In this state, the net increment of
In the second case, only the Poisson's ratio of the rust
was altered from the original data by taking vr = 0.4999,
which is nearer to 0.5 than in Equation 10. The results
were almost exactly the same as the original ones.
In the third case, the shear retention factor was
diminished to fl = 0.I which produced only minor
changes in the results, except for the secondary crack
which probably presents a mixed-mode fracture.
The consistency of the numerical model with respect
to the size of the load increments and the tolerance to
error for the solution was also tested separately for
specimen I with success.

APPENDIX E: Study of parameter sensitivity

This study was limited to specimen I which was
re-analysed by changing each of its main parameters
separately, but maintaining the original geometry and
mesh. The results are collected in Figs E l - 4 and must
be compared with the original ones in Fig. 3.
The effect of an increase in the ratio between the specific
Fig. C1 Geometrical model. volumes of rust and steel, which was augmented to
544 Molina, Alonso and Andrade

STRESS N/nan2

4"6891

3.880

2.808-

1 .BBB

8,B
(a) (b) PENETRATION X

STRESS N/r~m2 STRAIN ram/ram

4.800~ 8. 018 7 ............. T ...................... 7---'-,-'''2~ ........ ~ ..............

o.oo~+- ............. ~,. . . . . . . . . . . . . . . . . . . . . . . . . . . t .............. ~,. . . . . . . . . . . . . . .

I , "'N 0. 083.- . . . . . . . . . . . . . ~". . . . . . . . . . . . . . . . ~". . . . . .; . . . . . . ',. . . . . . . . . . . . . . ", . . . . . . . . . . . . . . .

3,~HH i" ............. .............. /_!2"._i ....... . . . . . . . . . . . . . . ...............

8.~0~2 ............. ; .......... : - 2 .............. [.............. i . . . . . . . . . . . . . . .

+ : - ' /

%:
,(.

..,
""21- ....... ;~;-"-i .............. i ..... i---::5:: .......... ::----~ ............... :
..j ; , j . J " ; ,'

B.030 8.040 8.858 o-~o '~---' . . . . ; 9 - ' ' ,

9~w ~00 0,91o e. B2e e. ~3o e. 94.e o. 05e
(c) (d) PENETRATION X ha.

PRESSURE N/nlm2

IS.800

18.000

$,08B.

8,9

PENETRATION X m~

(e)

Fig. E1 Results for specimen I (specific volume of rust augmented): (a) deformed mesh with cracks ( x = 0 . 0 4 m m ) ,
(b) normal stress at main crack, (c) normal stress at secondary crack, (d) average normal strain, (e) pressure at rebar.
 Materials and Structures 545

STRESS N/mm2

4.888-

3.888

2.888

! -.\.
L.888

8"83? ~8 8.418 8.4~8 8.~38 8.~48 .... e..~sa

(a) (b) PENETRATION X rqn

STRESS N/mm2 STRAIN itml/r~n

4.888 8.818 ............... ,. . . . . . . . . . . . . . ] .............. T .............. : ..............

8.889 ............................... ".............. '............... ; ..............,

3.888

8.88, ................ ~.............. i .............. i ....... -~--:-?:--:;7~--:

8.886. .5.............. ].............. i........:: .... ~<-'- ...........

~---~. . . . ~..,,.,++.,_._~: ,......... ~ .._- . ~ . . - . - ~ , .

2.8e8 ~ 8.885 ................ ',. . . . . . . . . . . . .

'
~ ............
', -"
~'t--yl
2"
........ ;" . . . . . . . . . . . . . .
:

8.884 ................ ,. . . . . . . . . . . . . . ": . . . . . . . . . . / ................. k ............

: "9 / i i
8.883 ................ :............ T,~-/s .......... r .............. ~ ...............
1.988
8.882 ................ ~ :--~::-~ .............. ~ .............. i .............. i
8.88,.
...........
/7. i ...... ~:::-.:----...-.47:::::=:::~:::::::;;:.i
........ i .............. i ............. i ............ ,:!

8.8~0. 80 8.'h.o 8.428 0.~38 8.,;48 8.~s8 8'8~".i7g . . . . . . . ;-:41~ ........ 8.~18 8.438 8.I,R ' 8.1SR
(C) PENETRATION x mn (d) PENETRATION X mm

PRESSURE N/m12

15.888-

18. 888 ~"-"~

S. 888 /

8'8~P.i,8 8.~18 8.428 R.438 8.448 8.,$8

PENETRATION X mm

(e)
Fig. E2 Results for specimen I (tensile strength diminished): (a) deformed mesh with cracks, (b) normal stress at main
crack, (c) normal stress at secondary crack, (d) average normal strain, (e) pressure at rebar.
 546 Molina, Alonso and A n d r a d e

i--7
..~,,, " . .o -'" . "" [ STRESS N/rr~2

4"888 I

......
J/;-}
3.888.
: .... .g.~. <,, i -x !

/__ ..... (.' 2.888-

. \, ',,

1.888

i! . ""

..~18 ..~28 8.~38 8.~48 ..~s8

(a) (b) PENETRATION X rnm

STRESS N/mm2 STRAIN rnm/m~

4.888~
,.,1, ............... i.............. 7 .............. T .............. r .............. :

8.889. ', : ', : :, ;

............. ================================
.................. . ~ ...................

3.888
I i

s.eee ............... -
,
.............
,,
~ ..............
,,
,~. . . . . . . . . . . . .
/
~
,,
...........
', ', ', /" 1 ', :

2.888 8. 885 ............... T ............. ~ .............. ~ ............ ,~ . . . . . . . . . . . . . . ',

', : / ' , ,
8.884 ............... ~ ............. ~ ....... "". . . . " .............. ~. . . . . . . . . . . . . "

, ,/ ,, ', ',
8.883 ............... + ............ ~ ........... r .............. r .............. ,
1.888 ,, .- , ', ,, ,,

8.882- ,'4:./ "1 r

/*

8.881 .............. ' ............. 1 .............. ; .............. L ...............

8."~8L8 8.~iB 8.~4e ' 8.8s8 8. og..i - - ~ / / ~ : " - - ........... :. . . . . . . . . . . . . . . :

(c) PENETRATION X .~ (d) PENET"ATION X

PRESSURE N/~2

lS.888-

//
s.eee :

8..~%,. 8.&18 ,328 8.&3, ,.&4, 8.&s,

PENETRATION X

(e)

Fig. E3 Results for specimen I (fracture energy augmented): (a) deformed mesh with cracks, (b) normal stress at main
crack, (c) normal stress at secondary crack, (d) average normal strain, (e) pressure at rebar.
Materials and Structures 547

STRESS N/met2

4"888T

'--..ii" 3..8, i," ",',

;/

1..,, /,, 8~1~ "i",,.

(a) (b) PENETRATION • mm

STRESS N/mm2 STRAIN g~/l~m

4.808- 8.fl18..............~............. ~..............9............... -. . . . . . . . . . . . .

6,8"9 . . . . . . . . . . . . . 4" . . . . . . . . . . . . . ~ . . . . . . . . . . . . . . " . . . . . . . . . . . . . . " . . . . . . . . . . . .

8.,8s .............. i............... - ............. ~ .............. '~..... ,,--<....

i i i i , " - ~ !
8.88, .............. L ............. ~ .............. ~ ............. ~:]_';_
~

8..~ .............. ~ ............. i .............. ! - - - • .......... -:

! i ~" i i
8.8.s ...............
',
,-: .............
;
~ ............
." t ...............
, ,
r ..............
,
,

8.8.4 ............... r ............. i-]-x ...... " .............. ,~.............. !

8 . 8 , 3 ................ i. . . . . . . . . . . . • ............. ! .............. ? .............. i

1 . . 8 8
............... :. . . . . . . <.__~ .............. ~ .............. ~ .............. :
8..~, : / : i i !
e. 8ez ................ i--~, .......... ~ .............. .~.............. .~ ............. .-:

8.,~8 i -ee- - - .... . ' . ' ~ J.e- ........ e-.~2-e ........ e-.~-3e .... e.~4e ' e.~se
(C) PENETRATION X rnm (d) PENETRATION X ram

PRESSURE N/mm2

l~'""R
I ,"~--~

.f;/ " ~

5,888

8"8~8.i.8 8.,118 ..... ~.~2, e.qi38 8.~..8 ,.~s8

PENETRATION X film

(e)

Fig. E4 Results for specimen I (Young's m o d u l u s of concrete diminished): (a) deformed mesh with cracks, (b) normal stress
at main crack, (c) normal stress at secondary crack, (d) average normal strain, (e) pressure at rebar.
548 Molina, Alonso and Andrade

v,/~ = 3.0 in Fig. El, seems to be an exact doubling of Education and Science of Spain for having supported
the speed in all the phenomena. In fact, that is precisely this research.
what Equation 12 would have predicted for the asymptotic
global crack-width rate, since the relative increment in the
REFERENCES
specific volume has been doubled with respect to the
original value. 1. Andrade, C., Alonso, C. and Molina, F. J. 'Cover cracking
If the tensile strength is diminished to fct = 2.5 M P a as a function of bar corrosion: Part I--Experimental
(Fig. E2), the peak values of the normal stresses and test', Mater. Struct. 26 (1993) 453-464.
2. Sygula, S. and Ryz, K. 'Longitudinal cracking and
pressures are likewise diminished, but the cracks keep
its relation to service life of reinforced concrete bridges',
non-zero normal stress at larger widths because the
in Proceedings of ACI-RILEM Symposium on Long-
fracture energy was not altered. However, the crack- Term Observation of Concrete Structures, Budapest,
width rate is only slightly affected. 1984, pp. 182-192.
Likewise, it can be seen in Fig. E3 how the effect of a 3. McLeish, A., 'Cracking due to corrosion'. Technical
greater fracture energy, i.e. Gr = 400 J m - 2, is very similar Note No. 1632 (Taywood Engineering Research Labor-
to the previous one, except for the fact that the peak atories, Southall, UK, 1986).
values of the stresses and pressures are unaltered with 4. Molina, F. J. and Alarc6n, E., 'Modelado num6rico
respect to the original. de la fisuraci6n for corrosi6n de armaduras', ICCET
In Fig. E4 the Young's modulus of the concrete was Hormigbn y Acero 175 (1990) 127-132.
diminished to Ec = 2 0 G P a which resulted in 'lazy' 5. Zienkiewicz, O. C,, 'The Finite Element Method' (McGraw-
behaviour because the cracks appeared later and were Hill, London, 1977).
6. Rots, J. G., Nauta, P., Kusters, G. M. A. and Blaawendraad,
also propagated later in a centrifugal sense. This effect
J., 'Smeared crack approach and fracture localization in
causes a delay in the maximum stresses and pressures
concrete', Heron 30 (1985) 1-48.
and in the first part of the crack growth graph, but once 7. Rots, J. G. and Blaawendraad, J., 'Crack models for
the cracks have been propagated, the response returns concrete: discrete or smeared. Fixed, multi-directional
to the original line and the measurable crack widths are or rotating?', ibid. 34 (1989) 1-59.
even a bit larger. 8. Reinhardt, H. W., Cornelissen, H. A. W. and Hordijk, D. A.,
Finally, the value of the bulk modulus of the rust was 'Tensile tests and failure analysis of concrete', ASCE
doubled to K r = 4 . 0 G P a which gave no significant J. Struct. Engn,q. 112 (1986) 2462-2477.
change in the results (not included). This surprising effect 9. Allan, M. L. and Cherry, B. W., 'Factors controlling
is probably due to the fact that the elastic deformation the amount of corrosion for cracking in reinforced
absorbed by the thin layer of rust is almost negligible in concrete', in 'Corrosion 91', Proceedings of Conference
and Show in Cincinnati, Ohio (NACE, Houston, Texas,
the whole model.
1991).
10. Hillerborg, A., Modeer, M. and Peterson, P. E., 'Analysis
ACKNOWLEDGEMENTS of crack formation and crack growth in concrete by
means of fracture mechanics and finite elements', Cem.
We are grateful to Dr J Rodriguez for his useful Concr. Res. 6 (1976) 773-782.
comments and to the Comisi6n Interministerial de l 1. Rots, J. G., 'Computational Modelling of Concrete Frac-
Ciencia y Tecnologia (CICYT) of the Ministry of ture', Dissertation, Delft University of Technology (1988).

RESUME combinaison de ddformations initiales et d'une modification

des propriOtks klastiques. Celles-ci sont respectivement
Fissuration du b6ton en fonction de la corrosion de
Oquivalentes `5 l'expansion et au ramollissement des
i'armature: Partie 2--Mod61e num6rique
~lOments d'acier `5 la surface de l'armature lorsqu'ils
On propose un modble numbrique bask sur des techniques s'oxydent.
normaliskes du finis pour la simulation de la Le module est appliquk `5 quatre exemples qui ont ktO,
fracture du de bOton armk lorsque leurs en mOme temps, essayds de faqon expbrimentale (voir la
armatures sont soumises ,5 ta corrosion. premiere partie de cet article), et on ktudie sdparOment
On utilise une technique de fissuration diffuse pour l'influence des divers paramOtres sur la rkponse. En
modbliser le comportement des klOments finis du bkton, particulier, on Ovalue quantitativement l'effet du volume
tandis que la corrosion, qui est considOrde comme la charge spOcifique de la rouille qui s'est formke sur la vitesse
appliquke sur la structure, est modOlisde au moyen d'une d'ouverture des fissures.