Fracture Mechanics of Concrete and Concrete Structures Recent Advances in Fracture Mechanics of Concrete - B. H. Oh, et al.

(eds) 2010 Korea Concrete Institute, Seoul, ISBN 978-89-5708-180-8

Energy based fatigue crack propagation model for plain concrete

Sonalisa Ray & J. M. Chandra Kishen
Indian Institute of Science, Bangalore, India

ABSTRACT: In this work, an analytical model is proposed for fatigue crack growth in concrete using the concepts of dimensional analysis and includes the following parameters size independent fracture energy, structural size, initial crack length, loading ratio and change in energy release rate. By knowing the governed and the governing parameters of the physical problem and by using the concept of self-similarity, the relationship among involved parameters is obtained. The coefficients of the law are calibrated using available experimental data in the literature. It is shown that the proposed fatigue law can capture the size effect and agrees well with the experimental results. 1 INTRODUCTION Repeated loading causes the crack to grow and finally leading to the failure of the structure. This phenomenon is known as fatigue fracture. Fatigue behavior is well understood for metallic structures, where it causes irreversible material damage (Paris & Erdogan 1963). But for concrete, the behavior is more complicated due to the presence of large size fracture process zone (FPZ) at the crack tip as shown in Figure 1. FPZ is a zone wherein the cement matrix is intensively cracked. Along the FPZ, there is a discontinuity in displacement but not in the stresses. Stresses are themselves a function of crack opening displacements. At the tip of the FPZ, the tensile stress is equal to tensile strength ( f ) of the material and gradually reduces to zero at the tip of the true crack. It is assumed that under low cycle fatigue loading the decrease in load carrying capacity and stiffness degradation occurs primarily in the FPZ and not in the undamaged material (Foreman et al. 1967). Development of mechanistic approaches using the concepts of fracture mechanics, for the study of crack propagation due to fatigue loading had started with the well known Paris law (Paris & Erdogan 1963) wherein crack growth increment per load cycle is a function of applied stress intensity factor amplitude. Attempts have been made by many researchers (Swartz 1978, Perdikaris 1989 and Baluch 1989) to model crack growth of concrete by applying Paris law (Paris & Erdogan 1963). However, one important aspect, namely - the size effect has been scarcely reported in the literature concerned with fatigue of concrete. The first attempt was made by Bazant and Xu (Bazant & Xu 1991). They combined their size effect law to the existing Paris law
'

and proposed a size adjusted Paris law applicable for plain concrete, given by:

da = C K dN K IC

(1)
2 If 1+
1

where K IC = K
If

d d 0

(2)

K is the fracture toughness of an infinitely large structure, is the characteristic dimension of the structure and 0 is an empirical constant. It gives the crack length increment per cycle as a power function of the amplitude of a size adjusted stress intensity factor. The numerical value of the transitional size is different for fatigue loading than the monotonic one. So, to take care of fatigue loading in the computation of fatigue law coefficients, transitional size has been adjusted and the adjustment factor used for 0 happens to be ten times the monotonic one which really has no significance. A similar law has been proposed describing the fatigue fracture of high strength concrete for predicting crack growth (Bazant & Shell 1993). However, concrete exhibits typically nonlinear fracture processes due to the large size process zone and makes this linear elastic fracture mechanics approach (LEFM) questionable. Slowik et al. (Slowik et al. 1996) proposed a linear elastic fracture mechanics based fatigue crack propagation law which included parameters such as the fracture toughness, loading history, and specimen size. This law applicable to variable amplitude fatigue loading is described by
d d d

= D ( h , T ) h

(1)

The proportionality coefficient D(h,T) is called moisture permeability and it is a nonlinear function of the relative humidity h and temperature T (Baant & Najjar 1972). The moisture mass balance requires that the variation in time of the water mass per unit volume of concrete (water content w) be equal to the divergence of the moisture flux J
= J
w t
Figure 1. Fracture Process zone.

explicitly accounts the evolution ofused. hydration onset of rapid crackfor growth have been Sain reaction and Kishen SF content. This sorption isotherm and Chandra (Sain & Chandra Kishen 2007) reads Slowik's law to include the effects of loadmodified ing frequency and overload function. The major limitation of the Slowik's model and the modifica tion made by Sain and (Sain Chandra Kishen ( we (h c c s ) + & dimens ) = G 2007) Chandra Kishen is that, (these are not g c c )h sionally homogeneous due nature the e empirical to of (4) proposed equations.
1 , , , 1 1 10 1

(2)

2 SCALING LAWS

K ( c s )e
, 1

10(

g c c )h
1

The water content w can be expressed as the sum of the evaporable water we (capillary water, water m n da = CandKadsorbed vapor, water) and the non-evaporable Imax K I (3) ( a ) p +F (chemically bound) water wn (Mills 1966, dN K K ( IC ) Pantazopoulo &I Mills 1995). It is reasonable to assume that the evaporable water is a function of where is the maximum stress intensityc factor , and relativeKIsup humidity, h, degree of hydration, ever reached by the structure in its past loading ,s) degree of silica fume reaction, s, i.e. we=we(h,chistory; KIC is the fracture toughness; KImax is the = age-dependent sorption/desorption isotherm maximum stress intensity factorthis in assumption a cycle; N and is (Norling Mjonell 1997). Under the number of cycles; F a is a function that ( ) by substituting Equation 1 into Equation 2 one takes into account the sudden overload on the crack obtains propagation and m, n, p are constants for all structural Similar model w based on w sizes. we the variation of h has been proposed e Paris by Kolluru et & & &al. (Kol + + ( D h ) = e law (3) c s + wn crack h t 2000).hIt is observed luru et al. that growth due c s to fatigue loading comprises deceleration stage followed they developed /hacceleration is the slope stage of theand sorption/desorption where by wean analytical expressions for the crack growth in both isotherm (also called moisture capacity). The the stages. But, the most important aspect size efgoverning equation (Equation 3) must be of completed fect has not been reportedand in this study. by appropriate boundary initial conditions. In the recent years, the concepts The relation between the amountof ofdimensional evaporable analysis and self-similarity are used many rewater and relative humidity is calledby adsorption searchers to study fatigue behavior of plain isotherm if measured with increasing concrete. relativity Carpinteri and desorption Spagnoli (Carpinteri & Spagnoli 2004) humidity and isotherm in the opposite have a size propacase. proposed Neglecting theirdependent differencefatigue (Xi et crack al. 1994), in gation law for concrete that expresses the the following, sorption isotherm will be usedcrack with growth rate the stress factor range. reference to against both sorption and intensity desorption conditions. The concepts of fractal geometry were used together By the way, if the hysteresis of the moisture with a new definition of into fracture energy and stress isotherm would be taken account, two different intensity factor (SIF) based on physical dimension relation, evaporable water vs relative humidity, must different from the classical ones. (Spagnoli be used according to the sign of Spagnoli the variation of the 2005) has derived a crack size dependent Paris law relativity humidity. The shape of the sorption using similarity methods and fractal concepts. The isotherm for HPC is influenced by many parameters, form of the fatigue law is proposed based upon especially those that influence extent and rate of the the assumption of an incomplete Thepore auchemical reactions and, in self-similarity. turn, determine thor has only shown the dependency of the fatigue structure and pore size distribution (water-to-cement parameters on growth ratecomposition, and has not obtained any ratio, cement chemical SF content, closed form expression for the fatigue crack propacuring time and method, temperature, mix additives, gation model. Carpinteri and Paggi (Carpinteri & etc.). In the literature various formulations can be Paggi 2007) have proposed an approximate relationfound to describe the sorption isotherm of normal ship between theal. Paris law coefficients C and m. concrete (Xi et 1994). However, in the present Two independent approaches, self-similarity conpaper the semi-empirical expression proposed by cepts and Mjornell the condition that the lawbecause instability Norling (1997) is Paris adopted it corresponds to the Griffith-Irwin instability at the
,
sup

where the term (gel which isotherm) represents the Scaling lawsfirst or power-laws describe the powerphysically bound between (adsorbed) water and the second law relationship different quantities give termevidence (capillary represents the capillary the of isotherm) a very important property of selfwater. This expression is valid only for low content similarity, wherein a phenomenon reproduces itself the amount of of different SF. The time coefficient G1 represents on and space scales. In construction water per unit volume in the gelto pores 100% an analytical model, it held is impossible take at into acrelative and it can be expressed (Norling count allhumidity, the factors which influence the phenomeMjornell 1997)model as is based on certain idealization non. So, every of the phenomenon. In constructing the idealizations, the phenomenon under study should (5) be c c+ ks s G ( c s ) = kat vg intermediate c vg s times and distances. considered Therefore, every mathematical model is based on intermediate asymptotic. In fact, self-similar ksvg are material parameters. solutions From the where kcvg and not only describes the behavior of the physical maximum amount of water per unit volume thatsyscan tems some capillary special conditions but also one defill allunder pores (both pores and gel pores), scribes the intermediate asymptotic behavior of the can calculate K1 as one obtains solution to broader classes of problems i.e. the behavior in the regions where these solutions have g h c c ceased to depend on the details of the initial condiw s + s G e c s tions and boundary conditions (Barenblatt 1996, 2004). (6) K (this )= In work, the phenomenon of fatigue crack c s similarity h g propagation on the basis of approach is e c c considered. The material parameters kcvg and ksvg and g1 can be DIMENSIONAL calibrated by fitting experimental data relevant to 3 ANALYSIS AND SELFfree (evaporable) water content in concrete at SIMILARITY various ages (Di Luzio & Cusatis 2009b). In any physical problem, we try to determine the relationship among evolution the physical quantities involved. 2.2 Temperature Let us consider, there exists a relationship between a Note that, at early age, since the chemical reactions quantity a which is to be determined from experiassociated with cement hydration SF reaction ments (governed parameter), andfield a and set of quantities are exothermic, the temperature is not uniform which are under experimental control (governing pafor non-adiabatic systems even if the environmental rameters), that can be written as temperature is constant. Heat conduction can be described in concrete, at least for temperature not a = f (a a a (Baant a) exceeding 100C & Kaplan 1996), (4) by Fouriers law, which reads where ( a1 a ) have independent physical dimensions, i.e. q = T none of these quantities have a dimension that can be represented in terms of a product (7) of powers of dimensions of the remaining quantities where q is the flux, T is as the a ) heat (a + and can be expressed the absolute product temperature, and is the heat conductivity; in this of powers of the dimensions of the parameters
, 1
10 0 0.188 0.22 1 1 1 , 1 10 1 1

,.......

k +1

,....

......

k 1

..........

Proceedings of FraMCoS-7, May 23-28, 2010

. This would mean that the dimension of the governed parameter a is determined by the dimensions of ( a a ) . Thereby, a + can be written as a + a a to make it dimensionless. Introducing the dimensionless parameters as:
( a1 ......ak )
p

......
r k

k 1

k 1

......

=
i

a+
k

a1

pk i

......

rk i k

(5) (6)

a
r a1 ........ak p

= ( 1

, ........

n k

(7)

where is a function of non-dimensional terms

a = f ( a1 ,...ak , ak +1 ,....an ) = a1p ......akr
ak +1 an ,,,, p +1 r +1 p r +1 ........a ........a a a k k 1 1
k k n k

(8)

Applying Buckingham theorem turns out to be a function of (n-k) variables only. The quanti are called similarity parameters, ties and the physical phenomenon is termed similar if the are identidimensionless parameters cal. We shall now discuss two important terms associated with dimensional analysis. (1) Self-similarity of first kind (2) Self-similarity of second kind. Let us consider the parameter a . This parameter is considered as non-essential if the corresponding dimensionless parameter is too large or too small (tend to zero or infinity), giving rise to a finite nonzero value of the function with the other similarity parameters remaining constant. The number of arguments can now be reduced by one and we can write
1

, ........

n k

, ........

n k

J = D ( h , T ) h 4 FATIGUE CRACK PROPAGATION FOR PLAIN CONCRETE AND SELF-SIMILARITY The proportionality coefficient D(h,T) The fatigue crack propagation phenomenon moisture permeability and itisisgena nonlinea erally analyzed in amplitudehrange and ofthe themedium relative humidity and temperature in this range the phenomenon is characterized an balanc & Najjar 1972). The moistureby mass intermediate asymptotic nature (Barenblatt 1996). that the variation in time of the water mas Assuming the crack growth rate (da(water /dN) ascontent the pa-w) be eq volume of concrete rameter to be determined in the phenomenon, which divergence of the moisture flux J is governed by the loading parameter characterized by change in energy release rate ( G ) and the = J characteristic dimension of the structure D . Also fatigue crack growth is governed by crack length (a), loading ratio (R) defined as thecontent ratio ofwminimum to The water can be expressed a maximum stress of amplitude and loading frequency the evaporable water we (capillary wa ( ) . Material properties considered are size indevapor, and adsorbed water) and the non-e pendent fracture energy ( G ) and tensile strength ( ) . (chemically bound) water wn (Mil Now, we can write the above dependence follows: Pantazopoulo & Mills as 1995). It is reas assume that the evaporable water is a fu da relative humidity, h, degree of hydration (11) Dsilica , t, R) = ( G, Gdegree , a , of , , fume reaction, s, i.e. we=w dN = age-dependent sorption/desorption (Norling Mjonell 1997). Under this assum The governing are Equation summarized with Equati byvariables substituting 1 into their physical dimensions expressed in the Lengthobtains Force-Time class (LFT) in Table 1. Considering a state of no explicit time dependence and w h we G , we e have independentphysical+ dimensions, & & ( D h) = dimensional c + s + w h h t analysis gives c s
w t

= 1 ( 2 , .............
1 1

n k

(9)
1

where is the limit of the function as 0 or . This is called complete self-similarity or self-similarity of first kind. On the other hand for 0 or , if tends to zero or infinity, then the quantity becomes essential, no matter how large or small it becomes. However in some cases, the limit of the function tends to zero or infinity, but the function has power type asymptotic representation which can be written as,
1 1 1

1 1 ( 2 , .............

n k

(10)

where the constant and the non-dimensional parameter cannot be obtained from the dimensional analysis alone. This is the case of incomplete self-similarity or self-similarity of second kind.
1

where w e/h is the slope of the sorption/ a D R (12) isotherm G (also called moisture capac governing equation (Equation 3) must be by appropriate boundary and initial conditi The non-dimensional are The quantities relation between the amount of e water and relative humidity is called G = R with increasing = = isotherm a 3 = if Dmeasured humidity and desorption isotherm in th G G G case. Neglecting their difference (Xi et al. following, sorption isotherm Now we need the to analyze whether the number ofwill be reference to both sorption and desorption arguments can be reduced further or not. Consider-of thec By the way, if the hysteresis smallwould in the be intermediate range ing , it is usually taken into account, two of fatigue crack isotherm growth and consideration of comrelation, evaporable water vs relative humi plete self-similarity will make the crack growth inbe used according to the sign of the varia dependent of G , so assuming incomplete self- of the humidity. The shape similarity in , relativity we can write isotherm for HPC is influenced by many p especially those that influence extent and da G G chemical reactions and, in turn, determ ( = (13) (water) structure and pore size distribution dN G ratio, cement chemical composition, SF curing time and method, temperature, mix Generally, theetc.). parameter is a large number In the literature various formulatio and experimental found results to (Spagnoli 2005) have shown describe the sorption isotherm the dependence of crack growth rate on a . Therefore, concrete (Xi et al. 1994). However, in th consideration of incomplete gives paper the self-similarity semi-empirical expression pro Norling Mjornell (1997) is adopted b
da dN
G G = , G G
f t f t

Proceedings of FraMCoS-7, May 23-28, 2010

J = 1. D (Governing h, T )h variables of the fatigue crack growth pheTable (1) nomenon in plain concrete. Variables Definitions Dimensions The proportionality G Stress intensity coefficient factor range D(h,T) FLis called moisture permeability and it is a nonlinear G Fracture toughness FL function of the relative humidity h and temperature a Initial crack length L T (Baant Structural size L requires &DNajjar 1972). The moisture mass balance Tensile strength FL per unit the variation that in time of the water mass Loading frequency T volume of concrete (water content w) be equal to the t Time T divergence of the moisture flux J R Loading ratio -3/2 -3/2
f

-2

-1

explicitly accounts for the evolution hydration properties) and constant loading ratio. of For finding reaction and SFthe content. This sorption the coefficients, input parameters are isotherm (da/dN), reads G and a. The coefficients m, p, s and are determined through an optimization process using the principle of least squares which means, the sum of squared residuals has its least value, a residual being )= G ( we (h c s c s ) + value difference between the observed value and the (g c )hvalues obtained from the model. m e bestc suited The (4) and p and s found to be equal to 4.4761 and and 5.4113
1 , , , 1 1 10 1

the first term (gel isotherm) represents the The water content w can be expressed as the sum where physically bound (adsorbed) water and the second (14) of the evaporable water we (capillary water, water term (capillary isotherm) represents the capillary and adsorbed vapor, and the non-evaporable water. This expression G a water) =G ( ) valid only for low content (chemically bound) water wn (Mills 1966, of SF. The coefficient is G1 represents the amount of Pantazopoulo &above MillsEquation, 1995). It is reasonable to water per unit volume held in the gel pores at 100% Rewriting the assume that the evaporable water is a function of relative humidity, and it can be expressed (Norling relative humidity, h, degree of hydration, c, and Mjornell 1997) as da p s m silica s reaction, , i.e. w =w (h, , ) degree s e e c s = G fof t ( G ) afume (15) dN age-dependent sorption/desorption isotherm = c c+ ks s (Norling Mjonell 1997). Under this assumption and G ( c s ) = k vg (5) c vg s where m = 1, p = , s = and = . by substituting Equation 1 into Equation 2 one The exponents , and the non dimensional paobtains Relationship between and the non-dimensional and ksvg are material parameters. From the where2. kcvg rameter , cannot be determined from the consid- Figure parameter. eration analysiswalone. These pa- maximum amount of water per unit volume that can w hof dimensionalw e e fill all pores (both capillary pores and pores), one & & & a best rameters can only either from ) = obtained + ( e Dh hbe (3) 0.0648 c + s + wn respectively andobtains indeed can begel considered as h t as one can calculate K 1 fitting procedure on experimental results, or accord- material constants. The parameter should acc s ing to numerical simulations. In the following sec- count for specimen size and geometry. Figure 2 tion, the are determined through a cali- shows the relationship between of the sorption/desorption where wparameters non g c e/h is the slope h diand the c G e s+ D G s bration process using the experimental results available isotherm (also called moisture capacity). The mensional wparameter c ( s ) derived for the small, in the literature. governing equation (Equation 3) must be completed medium (6) K ( c s ) =and large specimens. The resulting quad by appropriate boundary and initial conditions. ratic best fit is given by, g h The relation between the amount of evaporable e c c 5 CALIBRATION THE PROPOSED MODEL water and relative OF humidity is called adsorption s c Log D D vg = material (16) Log +and g1 isotherm if measured with increasing relativity Log The can k parameters G k vg and G In this section, the constants introduced during the be calibrated humidity and desorption isotherm in the opposite by fitting experimental data relevant to formulation of the model are determined case. Neglecting their difference (Xi et al. using 1994),the in free (evaporable) water content in concrete at D is the depth ofCusatis the specimen. Using the experimental results of Bazant & Xu (Bazant &with Xu where the following, sorption isotherm will be used various ages (Di Luzio & 2009b). equation one can obtain the value of the pa1991). They have tested and a series of geometrically reference to both sorption desorption conditions. above rameter for any size of specimen and grade of similar under fatigue loading. The By thethree-point way, if beams the hysteresis of the moisture concrete. 2.2 Temperature evolution dimension details of isotherm would be and takenthe intophysical account, properties two different these specimens arewater shown Tablehumidity, 2. Size inderelation, evaporable vs in relative must Note that, at early age, since the chemical reactions found to be of 0.038 pendent fracture energy be used according to the G signwas of the variation the associated with cement hydration and SF reaction OF THE PROPOSED FATIGUE N/mm (Bazant & Pfeiffer seen 6 relativity humidity. The 1988). shape It of could the be sorption areVALIDATION exothermic, the temperature field is not uniform MODEL AND DISCUSSION that the nondimensional parameter is a function isotherm for HPC is influenced by many parameters, for non-adiabatic systems even if the environmental of and R and can be assumed to rate be aof con( D G ) those especially that influence extent and the temperature is constant. Heat conduction can be order to in validate the proposed experimental stant for areactions particularand, material mix determine (same material chemical in turn, pore In described concrete, at least model, for temperature not structure and pore size distribution (water-to-cement exceeding 100C (Baant & Kaplan 1996), by ratio,2. cement chemical composition, SF content, Fouriers law, which reads Table Geometry and material properties of specimens. curing time and method, temperature, mix additives, Specimens Depth Span Thickness Notch size Tensile strength Youngs modulus Fracture energy q (t) = T etc.). In the literature can (D)various (S)formulations (B) (a) be (E) (Gf) (7) mm mmisotherm mm of normal mm MPa MPa N/mm found to describe the sorption Beam (Bazant 38.1 However, 95 38.1 6.35 2.86 27120 0.038 concrete (Xi 1991) et al. 1994). in the present where q is the 27120 heat flux, T is0.038 the absolute Beam (Bazant 1991) 76.2 191 38.1 12.7 2.86 paper the semi-empirical proposed by temperature, and is the heat conductivity; in this Beam (Bazant 1991) 152.4 expression 381 38.1 25.4 2.86 27120 0.038 Norling Mjornell is420 adopted Beam (Toumi 1998) (1997) 80 50.0 because 40 it 4.2 0.25 31600 2200 0.05
t f f

dN

J G G da = =
w t
f

G
1

a 2 ( 3 , 4 )

(2)

K ( c s )e
, 1

10(

g c c )h
1

1 1 2

10

0.188

0.22
t

10

1.3963

15.399

34.663

Proceedings of FraMCoS-7, May 23-28, 2010

results of Bazant & Xu (1991) and Toumi et al. (1998) have been used. Bazant & Xu have tested small, medium and large sized beam specimens in fatigue under three point bending. The specimen details are tabulated in Table 2. Figure 3 and 4 show the logarithmic plot of crack growth rate verses stress intensity factor amplitude normalized with size independent stress intensity factor (SIF) as well as size dependent SIF K . The experimental results reported by Bazant & Xu (1991) are plotted together with the proposed fatigue law, considering the case of constant amplitude loading.
(

IC )

J= D(h , T )h medium and large perimental study, the small, specimens failed at N = 974, 850 and 882 cycles respectively. The proportionality coefficient D(h,T) moisture permeability and it is a nonlinea of the relative humidity h and temperature & Najjar 1972). The moisture mass balanc that the variation in time of the water mas volume of concrete (water content w) be eq divergence of the moisture flux J

= J
w t

The water content w can be expressed a of the evaporable water we (capillary wa vapor, and adsorbed water) and the non-e (chemically bound) water wn (Mil Pantazopoulo & Mills 1995). It is reas assume that the evaporable water is a fu Figure 5. Calculated growth of relative crack length with the number of cycles. relative humidity, h, degree of hydration degree of silica fume reaction, s, i.e. we=w = age-dependent sorption/desorption Table 3. Fatigue law coefficients. (Norling 1997). Under this assum Specimens Proposed modelMjonellExperiment by substituting Equation 1 (Beam) Log C m Log C m into Equati Small -18.289 -16.7 11.78 obtains11.09
Medium Large -18.993 10.50 -19.594w 10.90
Log

Figure 3. Logarithmic plot of crack length increment per cycle verses the stress intensity factor amplitude normalized with K .
If

9.97 9.27 w w h e e & & + ( D h ) = e c + s + w h Table 4. Validation with data. h experimental t c s Stress ratios K da da
IC

-18.2 -18.2

Figure 4. Logarithmic plot of crack length increment per cycle verses the stress intensity factor amplitude normalized with K .
IC

From this plot, the Paris constants C (vertical axis intercept) and m (slope of the straight line) may be determined. Table 3 shows the values of Log C and m obtained from the proposed model and experimentally by Bazant & Xu. The agreement between the experimental data and the model predictions is noticeably good. In Figure 5 the relative effective crack length a, has been plotted as a function of number of load cycles for the proposed fatigue law. In the ex-

K dN dN the slope of the sorption/ where we/h is (Beam) Experiment Model isotherm (also called moisture capac F F = -0.136 -3.72 -3.97 3) must be governing equation (Equation -0.104 -3.37 -3.74 by appropriate and initial conditi -0.072 -3.03boundary -3.57 the amount of e F F = -0.16 The relation -3.98 between-3.94 -0.136 -3.71 is called water and -3.72 relative humidity -0.104 isotherm -3.37 if measured -3.65 with increasing -4.14 -4.58 F F = -0.216 humidity and desorption isotherm in th -4.08 -4.49 -0.208 case. Neglecting (Xi et al. -3.97 -4.41 their difference -0.2 the following, sorption-3.82 isotherm will be -3.57 -0.123 -3.7 -4.20 F F = -0.242 reference to both sorption and desorption c -3.4 -3.75 if the hysteresis -0.196 By the way, of the -3.14 -3.75 -0.168 isotherm would be taken-3.02 into account, two -2.83 -0.104 relation, evaporable water vs relative humi -2.91 -2.46 -0.069 be used according to the sign of the varia relativity humidity. The In the proposed model, the specimens failshape at N = of the for HPC is influenced by many p 880, 810 and 800isotherm cycles respectively. A good agreeespecially those that influence extent ment is seen between the experimental results and and chemical reactions in turn, the proposed fatigue model, thereby and, validating the determ structure and pore size distribution (watersame. ratio, cement composition, SF To verify the effectiveness of chemical the proposed model curing time and method, temperature, mix further, the experimental data reported by Toumi et Inisthe literature various formulatio al. (Toumi et al.etc.). 1998) used and a comparative found to describe the sorption study is carried out. The specimen details such isotherm as concrete properties (Xi et al. are 1994). However, dimensions and material tabulated in in th paper the semi-empirical expression Table 2. In the above experimental study, fatigue pro Norling Mjornell values (1997)ofisupper adopted b tests were performed at different
max
u
0.87

Log

Log

max

0.81

max

0.76

max

0.70

Proceedings of FraMCoS-7, May 23-28, 2010

J = D ( h,cyclic T )h load F (0.87 F , 0.81 F , 0.76(1) limit of F , 0.70 F ), keeping the lower limit load F constant for The all tests i.e. 0.23 F , coefficient where F is the peak load proportionality D(h,T) is called during static tests. Four various loading ratios ( ) moisture permeability and it is a nonlinear function are incorporated into Equation 16, leading to of the relative humidity h and temperature T (Baant & Najjar 1972). The moisture mass balance requires in water time of the that unit Log the = variation D D + mass + per R (17) Log Log G G ) be equal to the volume of concrete (water content w divergence of the moisture flux J The crack growth rate is plotted against the variation in the stress intensity factor for different stress (2) = and J compared with the experimental data ratios points as shown in Figure 6. Table 4 shows the results of water the crack growth rates varying The content w can be with expressed as normalthe sum ized stress intensity factor. is seen that there is (capillary water, water of the evaporable water we It good agreement between the experimental result and vapor, and adsorbed water) and the non-evaporable the proposed model. (chemically bound) water wn (Mills 1966, Pantazopoulo & Mills 1995). It is reasonable to assume that the evaporable water is a function of relative humidity, h, degree of hydration, c, and degree of silica fume reaction, s, i.e. we=we(h,c,s) = age-dependent sorption/desorption isotherm (Norling Mjonell 1997). Under this assumption and by substituting Equation 1 into Equation 2 one obtains
max
u u u u

min

explicitly accounts for the evolution of hydration through the present model matches closely with exreaction and SF content. This sorption isotherm perimental results for plain concrete specimens. reads REFERENCES
1 we (h, M. , ) 1 = G1 ( c + crack Baluch H., A. B. 1989. Fatigue Azad A. K. c , s )Quershi s& 10(g conference propagation in plain concrete. on c )h 1 c eInternational

1.3963

15.399

34.663

2.5

w t

w h e + ( D h) = we h h t

w e & & & c + s + wn c s

(3)

where we/h is the slope of the sorption/desorption isotherm (also called moisture capacity). The governing equation (Equation 3) must be completed Figure 6. Logarithmic plots of crack growth rate verses stress by appropriate boundary and initial conditions. intensity factor amplitude. The relation between the amount of evaporable water and relative humidity is called adsorption isotherm if measured with increasing relativity 7 CONCLUSIONS humidity and desorption isotherm in the opposite case. Neglecting their difference (Xi et al.model 1994),for in In this study, asorption fatigue crack propagation the following, isotherm will be used with plain concrete is developed using the concept of direference to both sorption and desorption conditions. mensional analysis. An important aspect of the proBy the way, if the hysteresis of the moisture posed model is that it is dimensionally homogeneous isotherm would be taken into account, two The different as compared to existing empirical models. prorelation, evaporable water vs relative humidity, must posed model takes into account a number of paramebe used according to thestrength, sign of the variation of the ters, such as, the tensile fracture toughness, relativity humidity. The shape of the sorption loading ratio and is most importantly theparameters, structural isotherm for HPC influenced by many size. In most cases, the crack growth rate is notof only especially those that influence extent and rate the a function of stress intensity range but also the mean chemical reactions and, in turn, determine pore stress level which is considered in the fatigue analystructure and pore size distribution (water-to-cement sis through stress ratio. composition, The analyticalSF prediction ratio, cement chemical content, curing time and method, temperature, mix additives, etc.). In the literature various formulations can be found to describe the sorption isotherm of normal concrete (Xi et al. 1994). However, in the present paper the semi-empirical expression proposed by Norling Mjornell (1997) is adopted because it
Proceedings of FraMCoS-7, May 23-28, 2010

Bazant Z. P. & Pfeiffer P. A. 1988. Determination of fracture energy from size term effect and number. ACI materiwhere the first (gelbrittleness isotherm) represents the als journals , 84 (6), 463480. physically bound (adsorbed) water and the second Bazant Z. P. & Xu K. 1991. Size effect in fatigue fracture term (capillary isotherm) represents the capillary of concrete. ACI materials journal , 88 (4), 390399. water. This expression is valid only for low content Bazant Z. P. & Shell W. F. 1993. Fatigue fracture of highstrength concrete and size effect. ACI materials journal, 90 of of SF. The coefficient G1 represents the amount (5), 472478. water per unit volume held in the gel pores at 100% Carpinteri A. & Spagnoli A. it 2004. fractal analysis of size efrelative humidity, and canAbe expressed (Norling fect on fatigue crack growth. International journal of faMjornell 1997) as tigue, 26 (4), 125-133. Carpinteri A. & Paggi M. 2007. Self-similarity and crack instability between the Paris conc in+the s correlation ( G growth Engineering k vg mechanics stants. , 76 , 10411053.(5) c s ) = k vg c c fracture ss Foreman R., Kearney V. & Engle R. 1967. Numerical analysis of crack in cyclic loaded structures. Journal of c propagation s are material parameters. From the where vg and k vg ASME basickengineering, , 89, 459464. maximum amount ofF., water perJ.unit that can Kolluru S. V., ONeil E. Popovics S. &volume Shah S. P. (2000). propagation in flexuralpores fatigue of concrete. Journal fillCrack all pores (both capillary and gel pores), one ofcalculate engineering mechnics, ASCE, 126 (9), 891-898. asF., one obtains can K1 Paris P. and Erdogan A Critical Analysis of Crack Propaga tion Laws, Journal of Basic Engineering, ASME, 85(3) (1963), 528-534 g h c c & Calomino e w C. s + A. s M. G 1989. Perdikaris P. Kinetics of crack c s frac growth in plain concrete. International conference on (6) )= K (ture of concrete and rock , Springer-Verlag, New York. 64 c s 69. g h c 2007. c e J. M. Sain T. & Chandra Kishen Prediction of fatigue strength in plain and reinforced concrete beams. ACI materials , 104 (5), 621628. The journals material parameters kcvg and ksvg and g1 can Slowik V., Plizzari G. A. & Saouma V. E.data 1996.relevant Fracture of be concrete calibrated byvariable fittingamplitude experimental to under fatigue loading. ACI mafree (evaporable) water content in concrete at terials journals, 93 (3), 272283. various ages (Di Luzio & Cusatis Spagnoli A. 2005. Self-Similarity and 2009b). fractals in the Paris range of fatigue crack growth. Mechanics of materials, 37, 519529. 2.2 Temperature evolution Swartz S. E., Hu K. K. & Jones G. L. 1978. Compliance monitoring of crack growth insince concrete. of engineering Note that, atASCE early age, theJournal chemical reactions mechanics, , 104 (3), 789800. associated with cement hydration and SF Toumi A., Bascoul A. & Turatsinze A. 1998. Crack reaction propagaaretion exothermic, temperature field is loading. not uniform in concrete the subjected to flexural cyclic Mateand structures , 31, 451458. forrials non-adiabatic systems even if the environmental
, 1
10 0 0.188 0.22 1 1 1 , 1 10 1 1

(4) Fracture of concrete and rock, Springer-Verlag, New York. 8085. 10(g )h Barenblatt G.I. 1996. Scaling, self-similarity and intermediate c 1 c K1 ( c , s ) e 1 asymptotic. Cambridge: Cambridge university press. Barenblatt G.I. 2004. Scaling. Cambridge university press.

temperature is constant. Heat conduction can be described in concrete, at least for temperature not exceeding 100C (Baant & Kaplan 1996), by Fouriers law, which reads
q

= T

(7)

where q is the heat flux, T is the absolute temperature, and is the heat conductivity; in this