5th International Conference on Durability of Concrete Structures

Jun 30–Jul 1, 2016
Shenzhen University, Shenzhen, Guangdong Province, P.R.China

Probability Model of Corrosion-Induced Cracking Time in
Chloride-Contaminated Reinforced Concrete
Chunhua Lu, Jinmu Yang, and Ronggui Liu
Faculty of Civil Engineering and Mechanics, Jiangsu University, Zhenjiang 212013, P.R.China

ABSTRACT
Corrosion-induced concrete cover cracking caused by chloride ion is an important indication of durability limit
state for marine reinforced concrete (RC) structures and can ultimately determine the structural service life. In this
paper, considering the random nature of factors affecting the corrosion cracking process, a probabilistic model
which expands on the deterministic model of cover cracking time is developed by using Monte Carlo simulation
technique. The results showed that the time to corrosion cracking can be modelled by the Weibull distribution.
Finally, the probabilistic analysis for the cracking time is applied to an in-site RC bridge girder with four different
durability design specifications. It is found that the mean and 90% confidence interval of the cover cracking time
will increase with the improvement of durability design level, which means that the difficulty in precise prediction
with deterministic model will augment accordingly.
Keywords: reinforce concrete, corrosion, cover cracking, probability model, Monte Carlo simulation

1. INTRODUCTION 1998; Andrade, Alonso, & Molina, 1993; Liu & Weyers,
1998; Oh, Kim, & Jang, 2009; Rasheeduzzafar, Al-
For chloride-contaminated reinforced concrete
Saadoun, & Al-Gahtani, 1992; Vu & Stewart, 2005),
(RC) structures, reinforcement corrosion has been
and some theoretical models predicting the time of
recognized as one of the major causes of structural
cover cracking have been proposed (Bazant, 1979;
performance degradation (Liu & Weyers, 1998; Oh &
Bhargava, Ghosh, Mori, & Ramanujam, 2005; Li,
Jang, 2003; Vu & Stewart, 2000). With the progress
Melchers, & Zheng, 2006; Liu & Weyers, 1998;
of corrosion, the volume expansion of rust products,
Lu, Jin, & Liu, 2011; Maaddawy & Soudki, 2007;
which is about two to six times the volume of steel
Morinaga, 1988). Except for the above researches
consumed, produces increasing mechanical forces
about analytical models, some researchers tended
to the surrounding concrete, which can eventually
to model the cracking behavior using nonlinear
result in the damage to the structures in the form of
fracture mechanics or finite-element method (Ahmed,
(longitudinal) cracking, spalling and delamination of
Maalej, & Mihashi, 2007; Chen & Mahadevan, 2008;
concrete cover, and loss of bond between concrete
Hansen & Saouma, 1999; Pantazopoulou & Papoulia,
and reinforcement (Liu & Weyers, 1998; Steward &
2001). These efforts have made great contributions
Mullard, 2007; Vu & Stewart, 2005). The cracking of
in analyzing the corrosion of reinforcement and
concrete cover due to steel corrosion is an important
relevant cover cracking. However, there are still some
criterion for analyzing and evaluating the service life of
differences between the observed data and predicted
the corroded RC structures. Once these cracks in the
values made by the above-mentioned models.
cover concrete appear, a path for a quicker ingress of
Some researchers (Kirkpatrick, Weyers, Anderson-
aggressive elements to the steel bars may be provided,
Cooka, & Sprinkel, 2002; Suo & Stewart, 2009; Thoft-
and the corrosion-induced deterioration processes
Christensen, 2000; Val & Trapper, 2008) have used
will be accelerated. Therefore, the appearance of the
the probabilistic method to predict the service life
first corrosion crack is a key indication for the end-
of corroded RC structure, which can enhance the
of-service-life of structural concrete, and it is widely
accuracy and realism of time prediction. Therefore,
accepted that the state of cover cracking induced by
it is valuable to use probability method to predict the
corrosion is identified as the serviceability limit state
time from corrosion initiation to cover cracking.
of RC structures (Steward & Mullard, 2007; Vu &
Stewart, 2005). Based on the theoretical model of time to cover cracking
proposed by Lu et al. (2011), as given in Eq. 1, this
For the problem of the corrosion-induced cover
paper intends to propose a probability model for the
cracking in concrete, lots of laboratory tests and
cover cracking time by using Monte Carlo simulation
field investigations have been conducted by many
technology. Then, the given probabilistic procedure is
researchers (Alonso, Andrade, Rodriguez, & Diez,

309

the statistical parameters for ratio n have not been d + kc   c well understood yet. fct (MPa) 3. respectively Mullard (2007) (Lu et al. with the increase in cover thickness for the same strength concrete. n 2. corrosion current density. and 70 mm. Statistical parameters of random variable. which can provide results are given in Table 1. 48. the the ranges of cover cracking time of [0. and steel diameter (mm). & Diez (1998).15. which indicates statistical information given in literatures (Steward that the probabilistic model expressed by Weibull & Mullard. The observed times given Ft (τ ) = Pr {t ≤ τ } = ∫ t ≤τ f ( X ) dX (2) by Liu & Weyers (1998) also be included in Figure 3. Thus. theory to let such variables be random variables and Figure 2 shows the cumulative distribution to associate with them a probability density function functions (CDF) of the fitted Weibull distributions (PDF) based on data and other information.75/2. As seen in Figure 3.9. c and d are cover thickness (mm) Tikalsky et al. complexity of composition of corrosion products.41/ 0. To establish the probabilistic model of cracking Rodriguez.. 2. W (0. and it is assumed to follow tcr = 26.3 0. icorr is the corrosion current density (mA/cm2). 2011).49] years. and Ec are adopted from Liu and Weyers’s are in the range of 90% confidence interval and experiment (Liu & Weyers. 0).47. respectively.18.1  Probabilistic time modeling with c = 27.18 Normal Thoft-Christensen is the thickness of the porous zone which is typically cm2) 1. 48.16 Normal (2005) mean ratio of volume of corrosion products to that d0 (mm) 15 0. the time. n is the c (mm) 27/48/70 0. test times of tcr observed by Liu and Weyers (1998) icorr. 2  fct  (r0 + c )2 + r02  2δ 0    + vc  + 1+  − 1 (1) Ecef 2  (r0 + c ) − r0 2  d   Table 1.3 + 0. may be more logical to involve those uncertainties in W (1.80 ⋅   0. see Table 1. . 2000. From Figures 2 and 3. 2. as shown in Figure 3. Nevertheless. Random where tcr is the time from corrosion initiation to cover variable Mean COV Distribution Reference cracking (year). 4. mechanical cracking time tcr follows the Weibull distribution. nc is Poisson’s ratio of concrete with 0.03] cumulative distribution function (CDF) of the cracking years.13. etc. the service life predictions. The products. d0 icorr (mA/ 3. 1.79. & Marek. 1. of cover cracking time can be obtained. the distribution relationships cracking time. which is assumed to be 2.2. [0. and fct and Ecef are tensile strength and Ec (GPa) 27 0. which properties of concrete. Andrade.95. the histograms of CRACKING TIME frequency density with 0. To represent uncertainty 48. are obtained It is clearly known that all parameters affecting the and shown in Figure 1. it corresponding distribution functions. it can be found that cover thickness.78] years are time can be defined as obtained for c  = 27. 2.1a interval for specimens 2.68. 2005).5). composition of corrosion is also suggested by Thoft-Christensen (2000). it is customary when using probability Figure 1.29. respectively.7 0. and the with four durability design levels. respectively. such as cover level of significance of 5%.0) for c = 27. and [1. 1998). possess random nature. are also plotted in about a variable. which was recognized and also 2. Through the KS test with a corrosion process of reinforcement. 0.2. function is acceptable in predicting the cover Pustka.310  5TH INTERNATIONAL CONFERENCE ON DURABILITY OF CONCRETE STRUCTURES applied to a RC bridge girder in marine environment for these input variables are estimated. respectively.6  ⋅ (n − 1) ⋅ icorr   d normal distribution with mean = 2.20 Lognormal Thoft-Christensen of consumed iron.79 (2000) in the range of 10–20 mm (Maaddawy & Soudki. PROBABILITY MODEL OF COVER By Monte Carlo simulation. and W (2. For 90% confidence the Monte Carlo simulation technique.0 (2000) here. 2007.5–3. Based on the follow nearly with the mean curve. Thoft-Christensen. and 70 mm. it can be seen that the where X is the vector of the random variables to be mean and 90% confidence interval of tcr increase taken into account and f ( X ) is the joint PDF of vector X . Based on for three test specimens. Tikalsky. Mullard (2007) r0  = d/2+d0. fct.7 and coefficient of variation (COV) = 0. 1.20 Normal – 2. some basic analytical parameters about c.18–0.2  Establishment of probabilistic model confirmed by the data reported by Alonso.12 Normal Steward and effective elastic modulus of concrete.13 Normal Steward and 2007). Therefore. the joint PDF interval between 5% quantile and 95% quantile. given the some reference to engineering design. and 70 mm.

Then based on Liu (Liu & Weyers. may thus be expressed as a time- dependent variable (see Figure 4): icorr (t ) = icorr (1) ⋅ α ⋅ (t − t0 )β (t − t0 ≥ 1year) (3) where t - t0 is the time since corrosion initiation. icorr (1) = 37. it is assumed that during the period from corrosion current density is usually time variant due corrosion initiation to cover cracking.Probability Model of Corrosion-Induced Cracking Time in Chloride-Contaminated Reinforced Concrete 311 Figure 1. Frequency histogram and its fitting curves for specimens in the literature (Liu & Weyers. and test times of tcr. and the value of icorr is diffusion of the iron ions away from the steel surface supposed to be calculated by Eq. icorr(1) is the corrosion current density at the start of corrosion propagation (mA/cm2). 1998). the For simplicity. in which the W/C ratio for three . then a = 1 and b = 0. 4. which reduce the density would be time invariant. 2000). MODEL’S APPLICATION TO IN-SITE RC t0 = 20 years (Vu and Stewart. 90% confidence interval. 2005).85 and b = -0. Figure 4. Figure 3.8(1 − W / C )−1. 2000) suggest that corrosion current density. the corrosion current to the formation of rust products. 2000). then a = 0. Time-dependent icorr suggested by Vu and Steward with 3. Steward (Vu & Stewart. and for relative humidity of 75% and temperature of 20°C..64 / c (4) where W/C is the water–cement ratio of concrete. and a and b are two constants. CDFs of tcr for three test specimens. Vu and and Weyers’s experiment. icorr. it can be expressed as Figure 2.3 (Vu & Stewart. Mean. BRIDGE GIRDER For natural corrosion process in RC members. 1998. If corrosion rate reduces with time. If the corrosion rate is time invariant. Tikalsky et al.

40 (fc′ = 54 MPa. To illustrate the application of probabilistic model to an in-site structural scenario. 5.98 MPa. W/C  = 0.55.50 (fc′ = 40. With the improving of level of durability design specification. and (4) Extreme durability: c  = 80 mm. 10. W/C  = 0. W/C=0.6 GPa). Since specifications. (3) High durability: c  = 60 mm.5 Figure 6. 4. low durability design specification results in a very short time to corrosion-induced cover cracking. .43] years. as shown in Figure 5.28 MPa. 90% confidence interval of tcr for four durability design humidity of 75% and temperature of 20°C. high. The RC bridge girder is assumed to be exposed to a chloride contaminated environment with a relative Figure 7. the measured corrosion rates are compared with supposed to be achieved in order to use the deterministic the calculations predicted by Eq.43 and 0. completed by using Monte Carlo simulation. respectively. [0. 1.5 MPa.50 MPa. the time and confidence interval of tcr increase distinctly. concrete tensile strength. Figures 6 and 7 show the PDFs and 90% confidence interval of tcr for four durability design specifications. W/C ratio is estimated from Bolomey’s formula (Vu & Stewart. 0. Ec = 25.24. For 90% confidence interval of tcr.91 MPa. 4.64.35 (fc′ = 63. and [3. it can be concluded that the difficulty in precise prediction with deterministic model will augment for good durability design specification.88] years. 27 W /C = (5) fc′ + 13. Here.56] years. 2000). the W/C ratio. The statistical parameters for density expressed by Eq. As a result.6 MPa. It model of Eq. Ec = 35. corrosion initiation. fct=5. 4 is effective in predicting the the COV and distribution of input variables are same corrosion rate during the period of corrosion cracking and as those listed in Table 1.60 (fc′ = 31. W/C  = 0.9 GPa). uniform corrosion of reinforcement is 1998). Four durability design specifications are considered as followings: (1) Low durability: c  = 25 mm. 5. moderate. and elastic modulus of concrete are dependent variables on the concrete compressive strength. the ranges of [0. Ec = 28. where fc′ is the concrete compressive strength of a standard test cylinder in megapascal. PDFs of tcr for four durability design specifications.3 GPa). All the probability analysis is will be used in the following application. 1. [1. as expressed in Eq. a comparative analysis considering the variability of concrete cover and concrete compressive strength is developed for a RC bridge girder.5 MPa. (2) Moderate durability: c  = 40 mm. fct = 7. Figure 5.45 (Liu & Weyers.92] years are obtained for low. The corrosion current density can be can be concluded from Figure 5 that the corrosion current predicted based on Eq. For this study. and extreme durability design level.14.312  5TH INTERNATIONAL CONFERENCE ON DURABILITY OF CONCRETE STRUCTURES specimens varied between 0.1 GPa). fct = 5. respectively. Comparison between test values and calculations for icorr. Ec = 33. fct = 6. Clearly. which has large range of tcr.

cracking prediction updating of deteriorating Chen. Congress. 127(4). & Mihashi. (1988).. MI: Ahmed. R. induced cover cracking in reinforced concrete Engineering Structures. Y. Chloride-induced RC structures using inspection information. Materials and Modelling cover-cracking due to reinforcement Structures. (1998). Factors controlling cracking of concrete Pantazopoulou. G.. H. Al-Saadoun. the Structural Division. Y. D. 168–175. C. (2000). M. U. Melchers. K. (2009). China via Grant Nos. With the statistical results of the Liu. S. H. 105(6). 23) and 51378241 are greatly acknowledged. pp. corrosion amount to cause cracking of reinforced 104(2). A... relation to bar diameter. P. 103(4). B. (1998). Japan: Shimizu Corp. Thoft-Christensen. Journal Material and Civil Engineering. P. S. & Zheng. Al-Gahtani. Cement and Concrete Composites. & Liu.. 1943–1960. Alonso. and composition Analytical model for corrosion-induced crack width of corrosion products. 30(3). (2008)... 1457–1464. Suo. V. J. & Mahadevan. Cover Mechanics. 2000. Oh. Alonso.. 327–343. in concrete cover. 26. Philadelphia. 1155–1166. M. The application of probabilistic model to realistic Reinforcement corrosion-induced cover RC bridge girder shows that the mean and 90% cracking and its time prediction for reinforced confidence interval of corrosion cracking time will concrete structures. T. (1992). L. PA.. Critical due to corrosion of steel. 1340–1348. S. J. (1979). (2007). S. Physical model for steel corrosion quality.. corrosion rate. H. C. J. Numerical of the crack initiation time for reinforced concrete simulation of reinforced concrete deterioration: structures. Journal of Engineering Andrade. E. Corrosion Science. 94(8).. G. R. S.. Weyers. Prediction of service lives of The financial supports of the National Natural Science reinforced concrete buildings based on rate of Foundation of P. ACI Structural method to predict the corrosion-induced cracking time Journal. which mainly considers the cover thickness and Maaddawy. Mori. P. 95(6). Z.. B. properties of concrete. Modeling the time main influencing parameters. ACI Material Journal. In: Proceedings of ASCE Structures Part II-steel corrosion and concrete cracking. Rodriguez. J. & Soudki. W. level. & Jang. Chloride diffusion analysis of concrete structures considering the REFERENCES effects of reinforcements. corrosion in RC structures. 51278230 corrosion of reinforcing steel (Report No. It is found that the cracking time can be Journal. 53(4). ACI Materials simulation. Cement and Concrete Composites. Cement and Due to the random natures of the main influencing Concrete Research.. A. A.Probability Model of Corrosion-Induced Cracking Time in Chloride-Contaminated Reinforced Concrete 313 4. C. K. R. J. . S. dependent reliability analysis of corrosion damage S. ACKNOWLEDGMENTS Morinaga. 31. (1993). CONCLUSION Kirkpatrick. May 8–10. M. H.. (2009). cracking as a function of rebar corrosion: Part Rasheeduzzafar. T. simulation. 32(12). Materials and Structures. concrete structures. mechanical Li. for prediction of time from corrosion initiation to corrosion cracking. Steward.. 675–681.. M. & Ramanujam. ACI Materials Journal. Modeling of time to corrosion. parameters. & Weyers. structures. E. ACI. Q.. 82–83. 29(3)... a probabilistic model to to corrosion cracking in chloride contaminated predict the cover cracking time is built by Monte Carlo reinforced concrete structures. & Stewart. Andrade. F. K. Farmington Hills. 96(3). J. D. A model concrete strength. M. (2006).. in concrete sea structures-application. cover. K. 435–441. J. Corrosion cracking in 453–464. C. Corrosion 35(11). M. & Jang. B. B. 153–161. Anderson- Cooka. H.. pp 1–8.. Kim. Maalej. 331–338. & Mullard. & Saouma. 29(7). (2003). (2007). & I-experimental test. it is reasonable to use probability in reinforced concrete structures.. C. and the timing of first repair for RC structures. modeled by the Weibull distribution. (2007). J. (2002). 2203–2218. Ghosh. Stochastic modelling Hansen. & Diez. (2011).R. Journal of 4(4). R. affected by reinforcement corrosion. & Sprinkel. S. Cover cracking of reinforced concrete beams Oh. increase with the improvement of durability design 1337–1347. ACI Materials Journal. (2001). S. K. E. E. C. J.. Jin. Q. (1999). & Molina... 227–238. 333–339. reinforcement corrosion and concrete cracking Reliability Engineering and System Safety.. E. 106(4). F. (2005). Spatial time- Bhargava. S. 479–487. Cement and Concrete Research. Lu. E. G. such as cover thickness. and concrete Bazant. 342–351. C. M. The following conclusions can be drawn from the Probabilistic model for the chloride-induced present study: corrosion service life of bridge decks. & Papoulia.

364–372. K. 1681–1689. A. T.. ACI structural Journal. Safety. J. (2008).. Probabilistic Vu. Predicting evaluation of initiation time of chloride-induced the likelihood and extent of reinforced concrete corrosion. V. chloride-induced corrosion models. 313–333. D. Val. Pustka. K. T. . Structural Statistical variations in chloride diffusion in reliability of concrete bridges including improved concrete bridges. (2000). 102(3).. D. (2005). & Stewart. Reliability Engineering and System corrosion-induced cracking. A. 93(3). 22(4). Journal of Structural Safety. A. (2005).314  5TH INTERNATIONAL CONFERENCE ON DURABILITY OF CONCRETE STRUCTURES Tikalsky. G. & Marek.. P. Vu. & Trapper. P.. G. Engineering. 131(11). Structural 481–486. P. & Stewart. M. M.