Accepted Manuscript

Modeling basic creep of concrete since setting time

B. Delsaute, J.-M. Torrenti, S. Staquet

PII: S0958-9465(16)30260-8
DOI: 10.1016/j.cemconcomp.2017.07.023
Reference: CECO 2873

To appear in: Cement and Concrete Composites

Received Date: 14 June 2016

Revised Date: 19 July 2017
Accepted Date: 21 July 2017

Please cite this article as: B. Delsaute, J.-M. Torrenti, S. Staquet, Modeling basic creep of concrete
since setting time, Cement and Concrete Composites (2017), doi: 10.1016/j.cemconcomp.2017.07.023.

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.
 ACCEPTED MANUSCRIPT

1 Modeling basic creep of concrete since setting time

2 B. Delsaute1,2, J.-M. Torrenti2, S. Staquet1

3 1 Université Libre de Bruxelles (ULB), BATir, LGC, Avenue F.D. Roosevelt 50 CP194/04, 1050

4 Bruxelles, Belgium.

PT
5 2 Université Paris Est, IFSTTAR, Boulevard Newton, 77447 Marne la Vallée Cedex 2, France.

RI
6 Corresponding author: Brice Delsaute (bdelsaut@ulb.ac.be)

SC
7 Abstract

U
8 Modeling the early age evolution of concrete properties is necessary to predict the early age behaviour
AN
9 of structures. In case of restrained shrinkage or application of prestress load [1], creep plays an

10 important role in the determination of the effective stress. The difficulty lies in the fact that the
M

11 modeling of creep must be based on experimental data at early age and this data must be obtained

12 automatically because the hardening process of the concrete takes place rapidly during the first hours
D

13 and also the first days. This paper presents a new methodology to model basic creep in compression
TE

14 since setting. Two kinds of tests are used: classical loadings and repeated minute-scale-duration

15 loadings. The classical test is used to characterize the creep function for one age at loading and the
EP

16 repeated minute-scale-duration loadings test is used to define two ageing factors for the creep

17 function. A new model based on the physical mechanisms and the two ageing factors is presented. A
C

18 comparison with the Model Code 2010 is done and an advanced way to consider ageing with the
AC

19 Model Code 2010 is presented.

20 Keywords: Concrete, very early age, basic creep, modeling

21 Introduction

22 Nowadays, the construction phases of modern concrete structures (including high-rise buildings,

23 bridge piers, and storage tanks) are challenging due to their high performance requirements.
 ACCEPTED MANUSCRIPT
24 Consequently, for the design of concrete structures it is important to model accurately the early age

25 behavior of concrete, which influences the whole service life. Even though the mechanical behavior

26 of hardened concrete can usually be accurately estimated, it is not always the case for the early age

27 behavior of concrete, when the mechanical properties change rapidly in function of the advancement

28 of the hydration reaction. Among all the usual parameters (including measures of strengths and E-

PT
29 modulus) needed for the design of the concrete structures, creep and relaxation must also be taken into

30 consideration.

RI
31 For general concrete structures built in several phases, the evolution of the restrained strains is similar

SC
32 and is composed of two periods: a heating period followed by a cooling one (Figure 1a). The heating

33 period begins just after the initial setting [2] when the mechanical properties of concrete start to

U
34 develop. During and after the setting, a large amount of heat is produced by the hydration reaction
AN
35 which leads to an increase of the temperature inside the concrete element and thus to an increase of

36 the thermal strain. In the meantime, autogenous strain starts to develop. No systematic tendency can
M

37 be given for the autogenous strain, because during the heating period, autogenous deformation results

38 in swelling or shrinkage according to the mixtures proportions and content [3,4]. However thermal
D

39 strains are generally higher than autogenous strain (especially for massive structures) and thus a
TE

40 general swelling of the concrete occurs during the heating period. The cooling period starts when heat

41 of hydration decreases rapidly or, depending on the massivity of the structure, when the formwork is
EP

42 removed. During this period, both autogenous and thermal strains decrease. As a result of the

43 restriction of the strains of concrete, stresses are induced (Figure 1b). During the heating period,
C

44 concrete element is in compression and inversely during the cooling period the concrete is submitted
AC

45 to tension. However, temperatures are nonuniform in most cases. It is possible to have both tension

46 and compression during cooling. Contraction of cooling surfaces can induce compression in the

47 interior, for example.

48 Creep and relaxation often play a positive role for the design of concrete structures at early age.

49 However their impact at very early age as part of restrained deformation could be negative as

50 highlighted on Figure 1b. In a general view, no consideration of creep/relaxation leads to a global

 ACCEPTED MANUSCRIPT
51 overestimate of the stress. Thus creep/relaxation seems to play a general positive role for the design of

52 concrete structures at early age. However, at very early age, the creep/relaxation amplitude is very

53 significant and reduces strongly the compressive stresses caused by thermal expansion. Thereafter,

54 during the cooling period, stresses may switch rapidly to tension. During this period, an

55 underestimation of the creep/relaxation phenomenon leads to an underestimation of the tensile

PT
56 stresses which can cause cracking in the concrete structure. Hence it is important to model correctly

57 the basic creep since final setting time.

RI
58 State of the art

SC
59 During this last decade, different creep models have been developed for early age concrete. These

U
60 models are generally based on the decomposition of the total strain of a loaded concrete sample in
AN
61 a sum of the elastic strain , the creep strain , the autogenous strain and the thermal strain

62 as expressed in Equation 1. The effects of drying are significant (see e.g. [5]), but are not considered
M

63 in the following state of the art. For a constant uniaxial stress , the uniaxial mechanical strain can be

64 defined as in Equation 2 where t is the age of the concrete, t’ the age of the concrete at loading, J(t,t’)
D

65 is the compliance function, E(t’) is the elastic modulus and C(t,t’) is the specific creep. Generally
TE

66 authors use the creep deformation, the creep compliance, the specific creep or the creep coefficient

67 to model creep. The link between these three last parameters is given in Equation 3.
EP

1
C

2
AC

68 As concrete creep at early age is still not well understood, mathematical expressions are used to fit

69 experimental data. Several models developed for hardened concrete are given in [6]. Models are often

70 composed of two terms: an amplitude term and a kinetic term. For consideration of ageing both

71 parameters can be expressed in function of the age at loading [6–12], the advancement degree of
 ACCEPTED MANUSCRIPT
72 reaction at loading [13,14] or the equivalent age at loading [15–17]. The advancement degree of

73 reaction is defined as the relative amount of hydrated cement and can be computed as a function of

74 the heat release and the heat release at an infinite time (Equation 4). The equivalent time is

75 based on the Arrhenius equation in order to take into account the main temperature effect on the

76 hydration process. Equation 5 defines the equivalent time which is function of the age of the material

PT
77 t, the evolution of the temperature T (°C), a reference temperature (generally 20°C), the universal

78 gas constant R (=8.314 J/mol/K) and the apparent activation energy (material parameter in J/mol).

RI
4

U SC
AN 5

79 Ageing of the creep function can be linked to other parameters such as the mechanical strength
M

80 [18,19] and the effect of the stress level could be included [19].. For these models no real physical

81 mechanism is associated to the amplitude or kinetic terms. With only one amplitude term and one
D

82 kinetic term, it is assumed in these models that only one mechanism is responsible of creep or that all
TE

83 mechanisms evolve with same kinetics but different amplitudes.

84 Among many previous studies like those reported in [12–15,17,19–22], several theories were
EP

85 developed to clarify mechanisms related to creep behavior. However each theory alone does not allow
C

86 explaining all experimental observations. Globally each theory can be linked to two mechanisms:

87 direct mechanisms linked to the cement paste and responsible of the highest part of the creep
AC

88 amplitude and indirect mechanisms linked to the heterogeneity of the concrete. Direct mechanisms are

89 related to the water mobility and to the solidification of the material and can be separated in short and

90 long terms phenomena [23–26]. The short term phenomenon is reversible with a small characteristic

91 time of about 10 days, is linked to a stress-induced water movement towards the largest diameter

92 pores and to the solidification [26,27] of the material, and occurs under increasing volume for uniaxial

93 compression. The long term phenomenon is irreversible with a high characteristic time and related to
 ACCEPTED MANUSCRIPT
94 viscous flow in the hydrates and occurs under almost constant volume. The creep rate of this long

95 term phenomenon evolves as a power function [28–30] with an exponent between -1 and -0.9

96 according to [31], between -0.72 and -0.69 according to results of [32] on concrete and an exponent

97 between -0.86 and -0.6 on cement paste according to results of [22]. Nanoindentation tests were

98 carried out on C-S-H by Vandamme, et al [33]. It was shown that C-S-H exhibits a logarithmic creep

PT
99 which is in agreement with results obtained on concrete [18,34]. Vandamme [35] compared also this

100 logarithmic behavior with other heterogeneous and porous materials with porosity including several

RI
101 orders of magnitude (soils and wood). For these non-ageing materials, a logarithmic long-term creep

102 was also observed. It can then be assumed that this long term creep is not linked to a hydration

SC
103 process or any chemical specificity of the C-S-H. The indirect mechanisms are due to micro-cracks

104 which occur progressively in the cement paste and at the interface between cement paste and

105
U
inclusions. Their presence can cause a redistribution of the stresses in the material [36].
AN
106 On basis of the results coming from the literature presented above, it could be considered that creep is
M

107 separated in three functions which correspond to the short term creep, the long term creep and a term

108 related to the formation of micro-cracks. In the model B3 of Bazant [31,37], both short and long term
D

109 creep are considered separately as expressed in Equation 6 and 7. Complete details about the different
TE

110 parameters of the model can be found in the references [31,37]. The short term creep is linked to the

111 solidification theory [27,38] and corresponds to the terms with parameter and in the equations.
EP

112 The long term creep is linked to the microprestress theory [26,39] and corresponds to the term with

113 parameter in the equation.

C
AC

114 Another way to model the viscoelastic behaviour of concrete is the use of rheological model where

115 the creep behavior is associated to spring and dashpot in series or in parallel. There is an infinite

116 number of associations of springs and dashpots, several examples are given in [40]. Bazant advises to
 ACCEPTED MANUSCRIPT
117 use Kelvin Voigt chains in series for creep [41] and Maxwell chains in parallel for relaxation [42]. To

118 consider ageing creep, spring and dashpot parameters have to evolve according to the age of the

119 concrete. De Schutter [13] linked the creep coefficient to a Kelvin-Voigt chain by considering that the

120 spring and the dashpot evolve according to the amplitude term of a model depending on the

121 advancement degree of hydration. This model has the advantage of being very simple because it uses

PT
122 only two parameters. However this model is limited for early age study and for a limited duration of

123 loading. To extend the use of the model of De Schutter, Benboudjema and Torrenti [43] generalized

RI
124 this model by means of several Kelvin-Voigt chains in series with different springs and dashpots. The

125 ageing of each element of each Kelvin-Voigt chain depends on the evolution of one parameter which

SC
126 is function of the advancement degree of hydration. In case of complex loadings with partial or total

127 unloadings, Briffaut [21] added a dashpot in series to the three Kelvin-Voigt chains where the creep

128
U
strains predicted by the additional dashpot are totally irreversible. Then the Kelvin-Voigt chains are
AN
129 linked to the short term creep and the dashpot is linked to the long term creep. A more general

130 rheological scheme of the concrete behavior is given in Figure 2 [26,44] where elastic, creep (short
M

131 term and long term), shrinkage and thermal strain are represented on a same rheological model.
D

132 Another way to consider ageing by using rheological model is done by Hermerschmidt [45] who uses
TE

133 4 Maxwell units and 1 single spring in parallel to model tensile creep test with several histories of

134 loading. The influence of the hardening process on the viscoelastic behavior is carried out by
EP

135 increasing the stiffness of the single spring and the viscosities of the dashpots according to the age of

136 the concrete.

C

137 Therefore several approaches were proposed in the literature and in the design codes to model creep
AC

138 behavior for early age or for hardened concrete.

139 Research significance

140 This paper is a continuation of the previous work by Delsaute, et al. [46] and reports on a new

141 strategy for the modeling of the basic creep since setting time. This strategy uses as a basis the results

142 from two kinds of test: classical creep test with permanent load applied during one week or more and
 ACCEPTED MANUSCRIPT
143 repeated minute-scale-duration loadings. With the consideration of the existing models and the

144 physical mechanisms presented above, new models are developed in order to accurately reproduce the

145 basic creep since the final setting time. For that purpose, experimental results on an ordinary concrete

146 studied in isothermal conditions at 20°C and in sealed conditions reported in [46] are used. A

147 comparison between the experimental results and the Model Code 2010 is carried out in order to point

PT
148 out a certain deficiency in application to very early age and to suggest a modification which may

149 eliminate this deficiency. This research is original since there is no model for the basic creep which is

RI
150 simultaneously based on experimental results at very early age and considers the different physical

151 mechanisms occurring during the hydration process.

SC
152 Experimental details

U
AN
153 Materials and mixtures

154 The results used in the present paper were performed on an ordinary concrete for which mix
M

155 proportions are given in Table 1. All materials come from the same batch of production. An ordinary

156 Portland cement of type CEMI 52.5 N was used. Its chemical composition is given in [46]. Siliceous
D

157 sand and gravel coming from Sandrancourt (France) were used. Sand and gravel were dried. The
TE

158 effective water-to-cement ratio is 0.45.

EP

Components

CEMI 52.5 N – SR 3 CE PM-CP2 NF (kg /m3) 340

C
AC

Sand (Sandrancourt 0/4) (kg /m3) 739

Gravel (Sandrancourt 6.3/20) (kg /m3) 1072

Added water (kg /m3) 184

Wadded/C 0.54
 ACCEPTED MANUSCRIPT
Weff/C 0.45

159 Table 1 – Mixture proportions and materials properties of the concrete

160 Different mechanical properties have already been characterized at IFSTTAR and ULB. The

161 evolution of the tensile and compressive strengths, the heat release, the Young’s modulus and the time

PT
162 of setting are presented in [47,48] and creep properties are presented in [46].

163 Assessment of the viscoelastic properties

RI
164 The devices used for the characterisation of the creep properties are a TSTM, the BTJASPE and 16

SC
165 compressive creep rigs [46–49]. All tests were performed in a climatic chamber with a temperature of

166 20 ± 1 °C and a relative humidity of 50 ± 5 %. Special attention was taken in order to assure that no

167
U
loss of water occurred during the test. Specimens tested on compressive creep rigs were weighted
AN
168 before and after the test. With two layers of self-adhesive aluminium sheet, no mass change was

169 observed for typical duration of loading of two weeks. With the TSTM and the BTJASPE, it is
M

170 assumed that no drying occurred due to the continuous presence of the mould during the whole test
D

171 duration.
TE

172 Compressive creep rigs are used to apply permanent loading during at least 6 days with a

173 stress/strength ratio of 40% at the age at loading. For each age, tests were carried out on 3 samples to
EP

174 demonstrate the good repeatability of the results. For the monitoring of the first term of the short term

175 creep, a new test so-called repeated minute-scale-duration loadings test was developed. Just after the
C

176 setting, every 30 minutes a load corresponding to 20% of the compressive strength is applied for 5
AC

177 minutes. During the 5-minutes load duration, the specific creep is computed from raw force and

178 displacement measurements after having removed the free strain from the dummy specimen (thermal

179 and autogenous strain). Complete details about the protocol of loading and the devices can be found

180 on the reference [46].

 ACCEPTED MANUSCRIPT
181 Previous results and observations

182 In a previous research work [46], results on compressive creep are reported for ages at loading of 15 –

183 20 – 24 – 40 and 72h00 (Figure 3a). It was observed that:

184 - The amplitude of the basic creep coefficient is strongly influenced by ageing during the first

PT
185 hours after loading. Earlier is the application of the loading and higher is the basic creep

186 coefficient. On the contrary, no ageing effect is detected for the kinetic of the basic creep

RI
187 coefficient (Figure 3b).

SC
188 - After the first hours of loading, no significant effect of the ageing is observed on the kinetic

189 and the amplitude of the basic creep coefficient. However a general trend shows that, after

U
190 few hours of loading, ageing increases the amplitude of the basic creep coefficient as it is
AN
191 shown in Figure 3c. All basic creep coefficients are set to zero after two hours of loading to

192 highlight this observation.

M

193 Results from one repeated minute-scale-duration loadings test are presented in Figure 4. The specific

194 creep curves coming from each minute-scale-duration loading are superimposed according to the age
D

195 after loading in Figure 4a. Two parameters were studied: the evolution of the kinetic and the
TE

196 amplitude according to the age at loading. In Figure 4b, each creep curves are normalized by their

197 value obtained after 5 minutes of loading. Thus each normalized creep curve has a value of 1 after
EP

198 duration of loading of 5 minutes. Each normalized creep curves are very close and no ageing effect is
C

199 detected. In Figure 4c, the amplitude of the creep coefficient after 5 minutes of loading (coming from

200 the repeated minute-scale-duration loadings test) is compared to the amplitude of the creep coefficient
AC

201 after 2 hours of loading (coming from the permanent loading test). Both amplitudes follow the same

202 evolution according to the age at loading. Thus it was concluded that the first term of the short term

203 creep can be defined with repeated minute-scale-duration loadings test and is divided in two terms: a

204 dimensionless kinetic term which is constant and an amplitude term which is function of the age of

205 concrete at loading.

 ACCEPTED MANUSCRIPT
206 Through a comparison of the experimental results and outcomes from microstructural simulation

207 carried out with VCCTL (version 9.5) [46], it is noted that short term creep is divided in two parts:

208 - A first term is linked to the state of the cement paste when the load is applied and more

209 specifically to the volume fraction of capillary pores of the cement paste (diameter between 5

210 and 9 µm) and also to the volume fraction of the CSH when the load is applied. A second

PT
211 term is linked to the development of the CSH and the theory of solidification of Bazant

212 [26,27]. The consolidation creep depends on the evolution of the properties of cement paste.

RI
213 Modeling

SC
214 Part1: Modeling of the repeated minute-scale-duration loadings test

215
U
Two models were developed to best fit the creep coefficient for each repeated minute-scale-duration
AN
216 loadings. A first model was introduced in [50]. The model is given in Equation 8. The model is a
M

217 power law and is inspired from the work of Gutsch [17] and De Schutter [13]. The model has the

218 advantage to consider only two parameters: a parameter linked to the amplitude of the creep
D

219 coefficient after 5 minutes of loading and a parameter linked to the kinetic evolution of the creep
TE

220 coefficient.
EP

8
C

221 where . The second model developed is a logarithmic law and is inspired from
AC

222 the Model Code 2010 (Equation 9). The model is also based on an amplitude parameter and linked

223 to the amplitude of the creep coefficient after 5 minutes of loading and a parameter linked to the

224 kinetic evolution of the creep coefficient.

9
 ACCEPTED MANUSCRIPT
225 where . The calculation of both parameters was carried out in two steps.

226 Firstly, each creep curve was fitted with three Kelvin-Voigt chains in order to remove noise from the

227 measurement. The value of the fitted curve after 5 minutes of loading corresponds to the amplitude

228 parameter for each age of loading. Secondly, the value of the kinetic coefficient is computed with

229 method of least squares (by using the fminsearch function in Matlab©). To compare the power and

PT
230 logarithmic expression in term of performance, the error is computed by considering the experimental

231 and modeling value of the creep coefficient for each cycle at the same time according to Equation 10.

RI
232 and are respectively the number of cycles during the test and the number of measuring points

233 for each cycle. The error is 0.0020 for the power law and 0.0014 for the logarithmic law. A reduction

SC
234 of error by a factor 1.5 is observed. This is coherent with results of long duration which show a

U
235 logarithmic trend for each age of loading since setting time during the first hours of loading. Then,
AN
236 only results from the logarithmic law will be considered on the next parts of this paper.
M

10
D
TE

237 Coefficient of the logarithmic model is given in Figure 5. The amplitude parameter decreases strongly

238 with the age at loading and has an evolution inversely proportional to the age at loading as indicated
EP

239 in Equation 11. The kinetic parameter is relatively constant as expected with experimental results

240 obtained. Then, for this concrete, a constant value of 0.35 is considered for the kinetic parameter .
C

241 The evolution of both parameters is given in function of the age at loading and not in function of the
AC

242 advancement degree of reaction as it has been done in [13] due to technical reasons. Calorimetry

243 testing in isothermal, semi-adiabatic or adiabatic condition is generally used to define the evolution of

244 the advancement degree of reaction. These technics are very accurate for the early age but not for later

245 ages. Indeed, after several weeks the heat flow of the hydration of the cement paste is very low and is

246 very difficult to assess experimentally. Thus after few weeks, the advancement degree of hydration

247 does not evolve significantly whereas the mechanical properties such as the compressive strength and

248 the elastic modulus continue to evolve significantly [47]. That is the reason why it has been chosen
 ACCEPTED MANUSCRIPT
249 that the expression of all parameters is given according to the equivalent time. A value of 35 kJ/mol is

250 used for the activation energy [51]. In Figure 6, modeled and experimental data are compared for

251 several ages at loading. A very good agreement is obtained between both.

11

PT
252 where a = 11.26, b = -19.62 and = 1h.

RI
253 Part 2 : Modeling of the permanent loading

SC
254 Based on experimental observations explained in section “Previous results and observations”, new

255 approaches are used for the modeling of the basic creep function at early age.

U
AN
256 a. Rheological model with ageing factor

257 Several Kelvin-Voigt units in series were used. The stiffness E and the viscosity of the chain with
M

258 retardation time

D

259 lower than 2 hours depend on the evolution of the amplitude parameter which is determined by
TE

260 means of a minute-scale-duration loadings test. The Figure 7 illustrates the different models proposed

261 before by De Schutter [13], Benboudjema & Torrenti [43] and the one developed in this study.
EP

262 In order to optimize the number of Kelvin-Voigt units, the modeling of the creep curves were carried

263 out in several steps. First the retardation time of each chain must be defined. For that, a first
C

264 modeling was carried out with 10 Kelvin-Voigt units in series for which the retardation time is
AC

265 defined according to Equation 12 which is adapted from the work of Bazant [41] in which is the

266 retardation time of the first Kelvin-Voigt chain and has a value of 1 minute. The value of the

267 coefficient depends on the value of the characteristic time of the first and last chain. The maximum

268 retardation time is the value of the maximum duration of loading considered in the modelling. As only

269 data of the first 145 hours after loading are considered, the value of the last Kelvin-Voigt unit
 ACCEPTED MANUSCRIPT
270 corresponds to a value of 145 hours. A same weight is given for each creep curve. Then, the value of

271 is computed by considering and and is equal to 2.24.

12

272 The stiffness of the spring of each unit (expressed in MPa) is imposed to be positive by using

PT
273 Equation 13 in which the parameter d is computed by using method of least squares (with the

274 fminsearch function of Matlab©) by considering each creep curve at the same time. Then units having

RI
275 a spring with high stiffness are removed because limited creep is produced by those units. Only 4

276 Kelvin-Voigt chains are necessary to model the ageing of each creep curve for a period of 145 hours.

SC
277

U 13
AN
278
M

279 Finally a new computation of the four Kelvin-Voigt chains is done by considering the four

280 characteristic times defined just before and by using again the method of least squares as done before.
D

281 Values of the parameters of the 4 Kelvin-Voigt chains are given in Table 2 and the results of the
TE

282 modeling are compared to experimental results in Figure 8.

(h) (h) (h)

EP

0.34 24.58 0.94 27.33 19.32 1.66 145 2.80

283 Table 2 Parameter values of the basic creep coefficient for 4KV model.
C
AC

284 A good agreement is found between the experimental results and the modeling. This way to model

285 creep curve has the advantage to be easily implemented in finite element software. For the study of

286 loading of short duration, as the case of restrained shrinkage, this way to model creep is convenient.

287 However it has also several disadvantages:

288 - The modeling is not able to predict creep for long duration of loading.
 ACCEPTED MANUSCRIPT
289 - The modeling needs several creep tests of long duration to be calibrated. If only two tests are

290 used to fit the parameters, the model cannot accurately predict results for other ages.

291 - The model does not highlight correctly the mechanisms which occur during the loading and

292 does not distinguish between reversible and irreversible creep.

293 b. Model with separation between short term and long term creep

PT
294 To improve the model, physical mechanisms must be taken into account. Long term creep was

295 identified by several authors [25,31,34] as logarithmic. It is particularly highlighted by comparing the

RI
296 evolution of the time derivative of the creep function of several creep curves according to the age of

SC
297 the concrete (Figure 9). For a given composition, the time derivative of the creep function follows a

298 same power function trend after several days for early age loading or after several weeks for loading

U
299 at later ages. Similar observations are done with the experimental results of the concrete studied in
AN
300 this article for very early age loadings and loading applied at an age of 28 days [52] as shown in

301 Figure 10. With very early age loading, long term creep is identified easily, accurately and with a
M

302 short duration of loading. The long term creep corresponds to a power expression and is defined by

303 equations 14 and 15.

D
TE

14

15
EP

304 where = 9E-6 MPa-1 and =-1.048. The parameter corresponds to the long term creep rate for an
C

305 age after loading of 1 hour. The amplitude of the long term creep is mainly governed by the exponent
AC

306 parameter . The value of is in agreement with observation of Bazant [31] (between -1 and -0.9). In

307 Figure 11, the long term creep is removed from the experimental results by using Equation 15 in order

308 to isolate the short term creep. For very early age loading, the short term creep evolves strongly

309 during the first hours after loading with a logarithmic trend and after approximately ten hours no short

310 term creep occurs anymore. For later ages at loading, a logarithmic evolution is observed during the

311 first week of loading. Short term creep can then be separated into two components:
 ACCEPTED MANUSCRIPT
312 - Initial short term creep, , which is represented by a logarithmic function associated with

313 the actual state of the material (capillarity porosity, CSH) when the load is applied. The

314 function is linked to the amplitude term of the minute-scale-duration loadings test which is

315 strongly correlated to the evolution of the largest diameter pores. The function has two

316 materials parameters.

PT
317 - Solidification creep, , which is linked to the solidification of the material and associated to

318 the quantity of formed CSH during loading and ipso facto to the decrease of the capillarity

RI
319 pores.

SC
16

320 Firstly the initial short term creep is modelled by a logarithmic expression as given in Equation 16

321 where = 20 and = 0.08. The parameter

Uis an amplitude parameter and the parameter is a

AN
322 kinetic parameter. Both parameters are function of the concrete composition. The equation used to

323 model the initial short term creep is inspired from the second term of the model B3 [31] as for the
M

324 modeling of the creep coefficient with the minute-scale-duration loadings test. The modeling of the

325 initial short term creep is compared to experimental results for which the long term creep is removed
D

326 in Figure 12. For early age loading, a very good agreement is observed during the first hours of
TE

327 loading. For later ages (here loadings at 2 or 3 days), the agreement is good till one week of loading.

328 In Figure 12, it is also shown that the model developed for minute-scale-duration loadings test
EP

329 (Equation 11) is not able to predict alone the creep behaviour for long duration. This can easily be
C

330 explained by the fact that only the initial state of the material is considered and not the mechanisms

331 which evolve during the hardening process as the solidification of the material and the viscous flow in
AC

332 the hydrates. The results presented here are in good agreement with the observation carried out by

333 Irfan, et al. [53] who show that creep results from repeated minute-scale-duration loadings on cement

334 paste are enough to predict the evolution of the creep function for several days.

335 Secondly, the initial short term creep is removed also from the experimental results in order to

336 identify the solidification part of the creep for each age of loading. In Figure 12, results of this

337 subtraction are given. For early age loading, the solidification term creep is very important and
 ACCEPTED MANUSCRIPT
338 evolves strongly during the first days of loading. For later ages of loading, the amplitude of this

339 parameter seems very low. In order to compare only the kinetics aspect of the solidification term

340 creep, each solidification creep curve is normalized at a time corresponding to an age after loading of

341 145 hours. In Figure 13, it is observed that the kinetics evolution of the solidification term creep does

342 not depend on the age at loading of the concrete. Results from the loading applied at an age of 72

PT
343 hours are not included in Figure 13 because for later ages corresponding to an high advancement

344 degree of reaction (the advancement degree of reaction for an age of 72 hours has a value of 0.69), the

RI
345 solidification phenomenon is very low and therefore is very difficult to analyse. The kinetic of the

346 solidification term creep is modelled by the use of Kelvin-Voigt chains in series which best fits the

SC
347 experimental data (Figure 13). Three Kelvin-Voigt chains in series are used for the modeling of the

348 solidification term. Values of the parameters of the Kelvin-Voigt chains are given in Table 3.

U
AN
(h) (MPa) (h) (MPa) (h) (MPa)

1 8344450 11.94 2788404 142.5 1153883

M

349 Table 3 - Solidification creep parameters for 3 KV model

D

350 Then, the evolution of the amplitude of the solidification term creep is given in Figure 14. The
TE

351 amplitude is defined as the value of the solidification term creep at an age after loading of 145 hours.

352 The solidification mechanism is not directly dependent of the state of the material when the load is
EP

353 applied. This is dependent of how the material will harden during its hydration. That is why a

354 macroscopic parameter such as the strength and the E-modulus or a microscopic parameter such as the
C

355 porosity and the CSH cannot directly be used to characterize the amplitude of the solidification creep.
AC

356 As the amplitude of the solidification is linked to a mechanism which evolves during the hydration,

357 this parameter should be linked to the evolution rate of one parameter such as the CSH at

358 microstructural scale [46] or the elastic modulus at macroscopic scale. Parameters such as the

359 compressive strength could not be considered because there are not relative to the general behaviour

360 of the concrete but relative to the behaviour of the material at failure which is not what is studied here.

361 In Figure 14, the amplitude of the solidification term creep is plotted according to the time derivative

362 of the elastic modulus. A linear relation is observed between both parameters. Therefore it could be
 ACCEPTED MANUSCRIPT
363 concluded that both parameters evolve with a same kinetics and that the solidification term creep and

364 the time derivative of the elastic modulus are directly proportional.

365 Finally the solidification term creep can be defined by the multiplication of two functions:

366 - One function which is linked to the kinetics evolution of the solidification and is independent

PT
367 of the age at loading. Three Kelvin-Voigt chains are used to model it.

368 - One function which is linked to the amplitude of the solidification, dependent of the age at

RI
369 loading and directly proportional to the evolution of the time derivative of the elastic

370 modulus. This amplitude term is noticed and is given in Equation 17.

SC
17

U
371
AN
372 Where = 0.6615 and = -0.054. In Figure 15, the contribution of the very short term, the

373 solidification term and the long term are summed and compared to the experimental results. A very
M

374 good agreement for each age at loading is observed. It is then concluded that the creep function is
D

375 divided in three terms and can be defined since the setting with one early age loading test of long
TE

376 duration and one minute-scale-duration loadings test which begins just after setting. The methodology

377 used to identify each term has three steps:

EP

378 1. The long term creep is defined with one creep test for which the load is applied at early age

379 during one week or more. After one day or more, the creep compliance rate follows a power
C

380 law which corresponds to the long term creep.

AC

381 2. The initial short term creep is obtained after having removed the long term creep and

382 corresponds to a logarithmic law. The kinetic is constant and the amplitude depends on the

383 amplitude parameter defined with the minute-scale-duration loadings test. This term fits well

384 the results obtained during the first hours of loading or more. This term is function of the

385 actual state of the material (capillarity porosity, CSH) when the load is applied. The function
 ACCEPTED MANUSCRIPT
386 is linked to the amplitude term of the minute-scale-duration loadings test which is strongly

387 correlated to the evolution of the largest diameter pores.

388 3. The solidification creep is obtained after having removed the long term and the initial short

389 term creep. The kinetic is constant and the amplitude depends on the time derivative of the

390 elastic modulus at the age of loading.

PT
391

RI
392 c. Model Code 2010

SC
393 In the recent Model Code 2010 (MC2010) [18], basic creep is expressed as the multiplication of an

394 amplitude term which is linked to the mean compressive strength at an age of 28 days and a

U
395 kinetic term which is function of the age at loading, the age after loading and the type of
AN
396 cement (Equation 18 and 19). For one composition defined, all parameters are constant except the age

397 at loading which is the only parameter which considers the ageing. When the time approaches infinity,
M

398 the creep compliance rate approaches a value corresponding to the inverse of the time. Then MC2010

399 considers the long term creep,

D
TE

18
EP

19

400 where depends on the type of cement and is equal to 1 for CEM 52.5 N, is the mean
C

401 compressive strength at an age of 28 days (here = 48 MPa). In Figure 16, results from MC2010
AC

402 are compared to the experimental results. For age at loading of 40 and 72 hours, the predicted values

403 of MC2010 are quite close to the experimental results. However for earlier loadings, a significant

404 difference is observed. To understand this difference, results of the elastic modulus and its time

405 derivative are compared to the predictive value of MC2010 (Equation 20 and 21) in Figure 17.

406 Several significant differences are highlighted. The elastic modulus is highly overestimated till an age

407 of 20 hours. The time derivative of the elastic modulus is always underestimated and particularly
 ACCEPTED MANUSCRIPT
408 during the two first days. This underestimation is also noted on the amplitude of the creep compliance

409 (Figure 16). As mechanical properties are not well predicted for the very early age, it is normal that

410 MC2010 is not able to predict correctly a parameter such as creep compliance at very early age. In

411 order to improve the prediction of the creep compliance it is necessary to adapt the model by changing

412 the effect of the age at loading.

PT
20

RI
21

SC
413 Where is the elastic modulus in MPa at an age in days, =36300 MPa is the elastic

414 modulus at an age of 28 days and depends on the concrete grade, s=0.2 depends on the strength class

415 of cement.

U
AN
416 d. Adapted Model Code 2010
M

417 Le Roy et al. [34] put forward that MC2010 can be expressed as in Equation 22. The parameter is

418 constant for one concrete composition and is the only parameter which changes according to the age
D

419 at loading. By adjusting both parameters for different set of results coming from tests carried out on
TE

420 ordinary and high performance concrete, Le Roy et al. show that MC2010 could be able to predict the

421 basic creep coefficient. The same methodology is used on the concrete studied in this paper and
EP

422 results of the fitting of the creep curve are given in Figure 18. The model is able to fit very well the

423 experimental results. A value of =217.8 GPa is found. The value of the parameter for the different
C

424 age of loading is given in Table 4.

AC

22

425

15 20 24 40 72

2.49E-07 8.20E-07 3.90E-04 1.62E-02 4.76E-02

426 Table 4 - Value of parameter from Equation 22.

 ACCEPTED MANUSCRIPT
427 Results of the parameter are compared to the inverse of the time derivative of the elastic modulus in

428 Figure 19. A power law linking the parameter and the time derivative of the elastic modulus is

429 highlighted. Therefore the time derivative of the elastic modulus seems to be a very good indicator of

430 the ageing of the creep function. As for the model separating short and long term creep , this

431 observation can be explained by the link between the rate of the elastic modulus and the velocity of

PT
432 the hardening of the cement paste and thus to the solidification of the material. In comparison, the

433 predictive values of are plotted according to the inverse of the time derivative of the elastic modulus

RI
434 predicted by MC2010 in Figure 20. An excellent linear correlation is done between both parameters.

435 In Figure 21, data from experimental results and MC2010 are superimposed. Two kinds of trend are

SC
436 observed. For very early age (between setting up to 40 hours), the relation between the parameter

437 and the inverse of the time derivative of the elastic modulus follows a power trend. Values by

438
U
MC2010 are not able to predict this trend. For higher ages (after 40 hours), experimental results and
AN
439 predicted values of MC2010 are very close and follow a same linear trend. Through results presented

440 in the model separating the short and long term creep, this change of trend at an age at loading of 40
M

441 hours can be interpreted as an effect of the solidification of the material. Indeed, before an age of 40
D

442 hours, the amplitude of the solidification of the material is significant on the evolution of the creep
TE

443 coefficient. For later ages at loading, the solidification of the material has a low impact on the value of

444 the evolution of the creep coefficient. The mathematical expression linking the parameter and the
EP

445 time derivative of the E-modulus is given in Equation 23 and 24 where n=0.0884 GPa, q=22542 d,

446 p=-7.41 and =1 d.GPa-1.

C

23
AC

24

447
 ACCEPTED MANUSCRIPT
448 Conclusions

449 This paper presents a new methodology to model basic creep in compression since setting by means

450 of minute-scale-duration loadings test with low duration of loading (5 minutes) and classical creep

451 tests with duration of loading of 6 days or more. Results from the minute-scale-duration loadings test

PT
452 allow defining the elastic modulus and an ageing creep factor corresponding to the value of the

453 creep coefficient after 5 minutes of loadings. Based on experimental observations, a new approach is

RI
454 developed for the modeling of the creep function.

SC
455 Several approaches are used to model creep on long duration (6 days or more):

456 1. Four Kelvin-Voigt units in series are used. Only units with a characteristic time lower than 2

457
U
hours have an ageing factor which corresponds to the value of the creep coefficient after 5
AN
458 minutes of loading. This way to model creep curve has the advantage to be easily

459 implemented in finite element software. For the study of loading of short duration, as the case
M

460 of restrained shrinkage, this way to model creep is convenient. For long duration of loading,

461 however, this way of modelling creep is not accurate. Several creep tests of long duration are
D

462 needed for calibration and the model does not distinguish between reversible and irreversible
TE

463 creep.

464 2. In order to consider the physical mechanisms occurring on concrete under loading during the
EP

465 hydration process, creep is divided in three terms. Two terms are related to the short term
C

466 creep and one term is related to the long term creep. Each term is identified since the setting
AC

467 in three steps:

468 a. The long term creep is defined with one creep test for which the load is applied at

469 early age during one week or more. After one day or more, the creep compliance rate

470 follows a power law which corresponds to the long term creep.

471 b. The initial short term creep is obtained after having removed the long term creep and

472 corresponds to a logarithmic law. The kinetic is constant and the amplitude depends

473 on the amplitude parameter defined with the minute-scale-duration loadings test. This
 ACCEPTED MANUSCRIPT
474 term fits well the results obtained during the first hours of loading or more. This term

475 is function of the actual state of the material (capillarity porosity, CSH) when the load

476 is applied.

477 c. The solidification creep is obtained after having removed the long term and the initial

478 short term creep. The kinetic is constant and the amplitude depends on the time

PT
479 derivative of the elastic modulus at the age of loading.

480 The Model Code 2010 is used and compared to experimental results. For ages at loading of 40

RI
481 and 72 hours, the predicted values of MC2010 are quite close to the experimental results.

482 However for earlier loadings, a significant difference is observed. An adapted version of the

SC
483 Model Code 2010 is proposed. In this version, the effect of the age at loading is considered by

484 using the inverse of the time derivative of the elastic modulus.

U
AN
485 This work is however limited to medium stress levels (up to 40 % of compressive strength). A next

486 step of this research is the study of creep in tension, higher stress levels and the coupling between
M

487 damage and creep at early age. More sophisticated tests with more complex histories of loadings or

488 temperatures are still needed for computational purposes. The measurement of creep effects on
D

489 transverse strain is also of interest.

TE

490 Acknowledgements
EP

491 Special thanks are addressed to Claude Boulay and Florent Baby, from IFSTTAR, Bernard Espion
C

492 and Thierry Massart, from Université Libre de Bruxelles, BATir Department, and Farid Benboudjema
AC

493 from ENS Cachan for their fruitful discussions.

494 References

495 [1] S. Asamoto, K. Kato, T. Maki, Effect of creep induction at an early age on subsequent

496 prestress loss and structural response of prestressed concrete beam, Constr. Build. Mater. 70

497 (2014) 158–164. doi:10.1016/j.conbuildmat.2014.07.028.

 ACCEPTED MANUSCRIPT
498 [2] P. Kumar Mehta, P.J.M. Monteiro, Concrete: Microstructure, Properties, and Materials, Fourth

499 Edition, Third, The McGraw-Hill Companies, 2006.

500 [3] A. Bentur, Terminology and definitions, in: A. Bentur (Ed.), Early Age Crack. Cem. Syst. -

501 Rep. RILEM Tech. Comm. 181-EAS - Early Age Shrinkage Induc. Stress. Crack. Cem. Syst.,

502 RILEM Publications SARL, 2003: pp. 13–15. doi:10.1617/2912143632.002.

PT
503 [4] B. Delsaute, S. Staquet, Decoupling Thermal and Autogenous Strain of Concretes with

RI
504 different water/cement ratios during the hardening process, Adv. Civ. Eng. Mater. 6 (2017) 22.

505 doi:10.1520/ACEM20160063.

SC
506 [5] K. Kovler, Drying creep of concrete in terms of the age-adjusted effective modulus method,

507 Mag. Concr. Res. 49 (1997) 345–351. doi:10.1680/macr.1997.49.181.345.

U
AN
508 [6] Z.P. Bazant, E. Osman, Double power law for basic creep of concrete, Matériaux Constr. 9

509 (1976) 3–11. doi:10.1007/BF02478522.

M

510 [7] A.D. Ross, Concrete creep data, Struct. Eng. 15 (1937).

511 [8] W.R. Lorman, The theory of concrete creep, Proc. ASTM. 40 (1940) 1082–1102.
D
TE

512 [9] Thomas Thelford, Comité Eurointernational du Béton CEB/FIB Model Code 90, Design Code,

513 1993.
EP

514 [10] H. Straub, Plastic flow in concrete arches, Proc. Am. Soc. Civ. Eng. 56 (1930) 49–114.

515 [11] J.R. Shank, Plastic flow of Portland Cement Concrete, Ind. Eng. Chem. 27 (1935) 1011–1014.
C
AC

516 [12] D.S. Atrushi, Tensile and Compressive Creep of Early Age Concrete : Testing and Modelling,

517 PhD thesis, The Norwegian University of Sciences and Technology, Trondheim,Norway,

518 2003.

519 [13] G. De Schutter, Applicability of degree of hydration concept and maturity method for thermo-

520 visco-elastic behaviour of early age concrete, Cem. Concr. Compos. 26 (2004) 437–443.

521 doi:10.1016/S0958-9465(03)00067-2.
 ACCEPTED MANUSCRIPT
522 [14] W. Jiang, G. De Schutter, Y. Yuan, Degree of hydration based prediction of early age basic

523 creep and creep recovery of blended concrete, Cem. Concr. Compos. (2013).

524 doi:10.1016/j.cemconcomp.2013.10.012.

525 [15] P. Laplante, Propriétés mécaniques des bétons durcissants: analyse comparée des bétons

526 classiques et à très hautes performances, PhD thesis, Ecole Nationale des Ponts et Chaussées,

PT
527 1993.

RI
528 [16] I. Guenot, J.-M. Torrenti, P. Laplante, Stresses in Early-Age Concrete: Comparison of

529 Different Creep Models, ACI Mater. J. 93 (1996) 254–259.

SC
530 [17] A.-W. Gutsch, Properties of early age concrete-Experiments and modelling, Mater. Struct. 35

531 (2002) 76–79. http://link.springer.com/article/10.1007/BF02482104 (accessed December 17,

532 2013).
U
AN
533 [18] H. Muller, I. Anders, B. R., Concrete: treatment of types and properties in fib Model Code
M

534 2010, Struct. Concr. 14 (2013) 320–334. doi:10.1002/suco.201200048.

535 [19] J.P. Forth, Predicting the tensile creep of concrete, Cem. Concr. Compos. 55 (2014) 70–80.
D

536 doi:10.1016/j.cemconcomp.2014.07.010.
TE

537 [20] R. Le Roy, Déformations instantanées et différées des bétons à hautes performances, PhD

538 thesis, Ecole Nationale des Ponts et Chaussées, Paris, France, 1995.
EP

539 [21] M. Briffaut, F. Benboudjema, J.-M. Torrenti, G. Nahas, Concrete early age basic creep:
C

540 Experiments and test of rheological modelling approaches, Constr. Build. Mater. 36 (2012)
AC

541 373–380. doi:10.1016/j.conbuildmat.2012.04.101.

542 [22] B.T. Tamtsia, J.J. Beaudoin, J. Marchand, The early age short-term creep of hardening cement

543 paste: load-induced hydration effects, Cem. Concr. Compos. 26 (2004) 481–489.

544 doi:10.1016/S0958-9465(03)00079-9.

545 [23] F.-J. Ulm, F. Le Maou, C. Boulay, Creep and shrinkage coupling : New review of some

546 evidence, in: ACI-RILEM Work. Creep Shrinkage Concr. Struct., 1999: pp. 21–37.
 ACCEPTED MANUSCRIPT
547 [24] O. Bernard, F.-J. Ulm, E. Lemarchand, A multiscale micromechanics-hydration model for the

548 early-age elastic properties of cement-based materials, Cem. Concr. Res. 33 (2003) 1293–

549 1309. doi:10.1016/S0008-8846(03)00039-5.

550 [25] P. Acker, F.J. Ulm, Creep and shrinkage of concrete: Physical origins and practical

551 measurements, Nucl. Eng. Des. 203 (2001) 143–158. doi:10.1016/S0029-5493(00)00304-6.

PT
552 [26] Z.P. Bažant, A.B. Hauggaard, S. Baweja, F.-J. Ulm, Microprestress-Solidification Theory for

RI
553 Concrete Creep. I Aging and Drying Effects, J. Eng. Mech. 123 (1997) 1188–1194.

554 [27] Z.P. Bazant, S. Prasannan, Solidification theory for aging creep. I: Formulation, J. Eng. Mech.

SC
555 115 (1989) 1691–1703. doi:10.1016/0008-8846(88)90028-2.

U
556 [28] Z.P. Bažant, Double-power logarithmic law for concrete creep, Cem. Concr. Res. 14 (1984)
AN
557 793–806.

558 [29] F.H. Wittmann, Creep and Shrinkage Mechanisms, in: Z.P. Bažant, F.H. Wittmann (Eds.),
M

559 Creep Shrinkage Concr., 1982: pp. 129–161.

560 [30] O. Bernard, F.-J. Ulm, J.T. Germaine, Volume and deviator creep of calcium-leached cement-
D

561 based materials, Cem. Concr. Res. 33 (2003) 1127–1136.

TE

562 [31] Z.P. Bažant, Creep and shrinkage prediction model for analysis and design of concrete
EP

563 structures - Model B-3, Mater. Struct. 28 (1995) 357–365.

564 [32] F.H. Wittmann, Useful Fundamentals of Shrinkage and Creep of Concrete, in: Concreep 10,
C

565 2015: pp. 84–93.

AC

566 [33] M. Vandamme, F.-J. Ulm, Nanogranular origin of concrete creep., Proc. Natl. Acad. Sci. U. S.

567 A. 106 (2009) 10552–7. doi:10.1073/pnas.0901033106.

568 [34] R. Le Roy, F. Le Maou, J.M. Torrenti, Long term basic creep behavior of high performance

569 concrete: data and modelling, Mater. Struct. 50 (2017) 85. doi:10.1617/s11527-016-0948-8.

570 [35] M. Vandamme, A Few Analogies Between the Creep of Cement and of Other Materials, in:

571 Concreep 10, 2015: pp. 78–83.

 ACCEPTED MANUSCRIPT
572 [36] J.M. Torrenti, Basic creep of concrete-coupling between high stresses and elevated

573 temperatures, Eur. J. Environ. Civ. Eng. (2017) 1–10. doi:10.1080/19648189.2017.1280417.

574 [37] Z.P. Bazant, S. Baweja, Creep and Shrinkage Prediction Model for Analysis and Design of

575 Concrete Structures : Model B3, ACI Concr. Int. 83 (2001) 38–39.

576 [38] Z.P. Bažant, S. Prasannan, Solidification Theory for Concrete Creep. II: Verification and

PT
577 Application, J. Eng. Mech. 115 (1989) 1704–1725. doi:10.1061/(ASCE)0733-

RI
578 9399(1989)115:8(1691).

579 [39] Z. Bazant, A.B. Hauggaard, S. Baweja, Microprestress-Solidification Theory for Concrete

SC
580 Creep. II: Algorithm and Verification, J. Eng. Mech. 123 (1997) 1195–1201.

U
581 [40] Y.-S. Yoon, S.-K. Lee, M.-S. Lee, J.-K. Kim, K.-Y. Kwon, Rheological concrete creep
AN
582 prediction model for prestressed concrete bridges constructed by the free cantilever method,

583 Mag. Concr. Res. 63 (2011) 645–653. doi:10.1680/macr.2011.63.9.645.

M

584 [41] Z.P. Bažant, A. Asghari, Computation of Kelvin chain retardation spectra of aging concrete,

585 Cem. Concr. Res. 4 (1974) 797–806.

D

586 [42] Z.P. Bažant, A. Asghari, Computation of Age-Dependent Relaxation spectra, Cem. Concr.
TE

587 Res. 4 (1974) 567–579.

EP

588 [43] F. Benboudjema, J.M. Torrenti, Early-age behaviour of concrete nuclear containments, Nucl.

589 Eng. Des. 238 (2008) 2495–2506. doi:10.1016/j.nucengdes.2008.04.009.

C

590 [44] M. Jirásek, P. Havlásek, Microprestress-solidification theory of concrete creep: Reformulation

AC

591 and improvement, Cem. Concr. Res. 60 (2014) 51–62. doi:10.1016/j.cemconres.2014.03.008.

592 [45] W. Hermerschmidt, H. Budelmann, Creep of Early Age Concrete under Variable Stress, in:

593 CONCREEP 10, American Society of Civil Engineers, Reston, VA, 2015: pp. 929–937.

594 doi:10.1061/9780784479346.111.

595 [46] B. Delsaute, C. Boulay, S. Staquet, Creep testing of concrete since setting time by means of

596 permanent and repeated minute-long loadings, Cem. Concr. Compos. 73 (2016) 75–88.
 ACCEPTED MANUSCRIPT
597 doi:10.1016/j.cemconcomp.2016.07.005.

598 [47] C. Boulay, S. Staquet, B. Delsaute, J. Carette, M. Crespini, O. Yazoghli-Marzouk, et al., How

599 to monitor the modulus of elasticity of concrete, automatically since the earliest age?, Mater.

600 Struct. 47 (2014) 141–155. doi:10.1617/s11527-013-0051-3.

601 [48] B. Delsaute, C. Boulay, J. Granja, J. Carette, M. Azenha, C. Dumoulin, et al., Testing Concrete

PT
602 E-modulus at Very Early Ages Through Several Techniques: An Inter-laboratory Comparison,

RI
603 Strain. 52 (2016) 91–109. doi:10.1111/str.12172.

604 [49] S. Staquet, B. Delsaute, A. Darquennes, B. Espion, Design of a Revisited Tstm System for

SC
605 Testing Concrete Since Setting Time Under Free and Restraint Conditions ., in: Concrack 3 -

606 RILEM-JCI Int. Work. Crack Control Mass Concr. Relat. Issues Concern. Early-Age Concr.

607 Struct., 2012: p. 12.

U
AN
608 [50] B. Delsaute, J. Carette, S. Staquet, Monitoring of the creep and the relaxation at very early age:
M

609 complementary results on the CEOS concrete, in: VIII Int. Conf. Fract. Mech. Concr. Concr.

610 Struct., 2013: pp. 453–458.

D

611 [51] S. Staquet, M. Azenha, C. Boulay, B. Delsaute, J. Carette, J. Granja, et al., Maturity testing
TE

612 through continuous measurement of e-modulus: an inter-laboratory and inter-technique study,

613 in: Proc. ECO-CRETE Int. Symp. Sustain., 2014: p. 8.

EP

614 [52] P. Rossi, J.L. Tailhan, F. Le Maou, Creep strain versus residual strain of a concrete loaded
C

615 under various levels of compressive stress, Cem. Concr. Res. 51 (2013) 32–37.

616 doi:10.1016/j.cemconres.2013.04.005.
AC

617 [53] M. Irfan-ul-Hassan, B. Pichler, R. Reihsner, C. Hellmich, Elastic and creep properties of

618 young cement paste, as determined from hourly repeated minute-long quasi-static tests, Cem.

619 Concr. Res. 82 (2016) 36–49. doi:10.1016/j.cemconres.2015.11.007.

620

621
 ACCEPTED MANUSCRIPT
622

623 List of figures

PT
RI
SC
624

U
625 Figure 1 - (a) Free strain inside concrete element. (b) Stress induced by restriction of the concrete
AN
626 displacement.

627
M
D
TE
C EP

Figure 2 - Rheological scheme of the concrete behavior [44]

AC

628
 ACCEPTED MANUSCRIPT
629

PT
RI
(a) (b) (c)

SC
Figure 3 – (a) Creep coefficient for several ages at loading – (b) First two hours after loading of the

creep coefficient – (c) Creep coefficient set to zero at an age after loading of 2 hours [46].

630
U
AN
M
D
TE
EP

Figure 4 – Repeated minute-scale-duration loadings. a: for each minute-scale-duration loading, creep

curves are superimposed and set to zero when the load is kept constant. b: for each minute-scale-
C

duration loading, creep curves are normalized by their value obtained after 5 minutes of loading. c:
AC

Evolution of the amplitude of the creep coefficient after 5 minutes of loading (coming from the

repeated minute-scale-duration loadings tests (RMSL)) and 2 hours of loading (coming from the

permanent loading tests) according to the equivalent time [46].

631
 ACCEPTED MANUSCRIPT

PT
RI
Figure 5 – Evolution of the amplitude and kinetic Figure 6 – Experimental data of the creep

SC
parameter of Equation 9 according to the coefficient and modeled curves for several ages at

equivalent age at loading. loading from Equation 9.

632
U
AN
M
D
TE
C EP
AC

Figure 7 - Comparison between the approach of De Schutter [13], Benboudjema & Torrenti [43] and

the new strategy developed

 ACCEPTED MANUSCRIPT
633

PT
RI
SC
Figure 8 – Comparison between 4 KV chains model developed with minute-scale-duration loading

testing for duration of one week and experimental results of permanent loading tests.

U
AN
634
M
D
TE
C EP
AC

Figure 9 – Time derivative of the creep Figure 10 – Time derivative of the creep function

function according to the age of the concrete according to the age of the concrete for results

for a normal strength concrete (W/C=0.5) and obtained by Delsaute [46] and results obtained by

a high strength concrete (W/C=0.33) from Rossi for an age at loading of 28 days [52].

[33].

635
 ACCEPTED MANUSCRIPT

PT
RI
Figure 11 – Evolution of the creep coefficient Figure 12 – Evolution of the creep coefficient

SC
with subtraction of the long term part and with subtraction of the long term and very short

modeling of the very short term creep (dashed term part (continuous lines) and modeling of the

U
lines correspond to the modelling) consolidation creep (dashed lines).
AN
636
M
D
TE
C EP

Figure 13 – Normalization at 145 hours of the Figure 14 – Comparison between the amplitude
AC

creep coefficient after subtraction of the long of the solidification term creep after 145 hours

term creep and the very short term creep. of loading and the time derivative of the elastic

Modeling of the kinetics of the solidification modulus according to the age at loading.

term creep.

637
 ACCEPTED MANUSCRIPT

PT
RI
Figure 15 – Modeling of the creep coefficient by

SC
identification of the long term creep, the short

term creep and the solidification creep

U
(continuous lines correspond to experiments and
AN
dashed lines to the modeling).
M

638
D
TE
C EP
AC

Figure 16 – Comparison between predicted value Figure 17 – Comparison between predicted value

of the creep compliance obtained with MC2010 of the elastic modulus and the time derivative of

(dashed lines) and experimental results the elastic modulus obtained with MC2010 and

(continuous lines). experimental results.

639
 ACCEPTED MANUSCRIPT
640

PT
RI
SC
Figure 18 – Comparison between experimental Figure 19 – Comparison between the time

creep curves, creep curves coming from a model derivative of the elastic modulus and the

U
inspired by the model code 2010 (continuous parameter of the adapted model code 2010
AN
lines correspond to experiments and dashed lines

to the modeling).
M

641
D
TE
C EP
AC

Figure 20 – Evolution of the parameter τ from Figure 21 - Comparison of the relation between

Model code 2010 according to the time derivative the inverse of the time derivative of the elastic

of the elastic modulus. modulus and τ from experimental results and

Model Code 2010.

 ACCEPTED MANUSCRIPT
642

PT
RI
U SC
AN
M
D
TE
EP
C
AC