Hydro-mechanical behaviour of soilcretes through a parametric laboratory study

Hydro-mechanical behaviour of soilcretes through a parametric laboratory study

Construction and Building Materials 166 (2018) 657–667 Contents lists available at ScienceDirect Construction and Building Materials journal homepag...

3MB Sizes 1 Downloads 33 Views

Construction and Building Materials 166 (2018) 657–667

Contents lists available at ScienceDirect

Construction and Building Materials journal homepage: www.elsevier.com/locate/conbuildmat

Hydro-mechanical behaviour of soilcretes through a parametric laboratory study Olivier Helson a, Javad Eslami a, Anne-Lise Beaucour a,⇑, Albert Noumowe a, Philippe Gotteland b a b

University of Cergy-Pontoise, Laboratoire de Mécanique et Matériaux du Génie Civil, EA4114, F-95000 Cergy-Pontoise, France Fédération Nationale des Travaux Public, 3 rue de Berri, 75008 Paris, France

h i g h l i g h t s  Soil-cement mixtures are much more sensitive to drying-shrinkage than ordinary concretes.  Permeability can be determined from the porosity and the characteristic pore diameter.  Mechanical damage under loading mainly depends on cement content.  Poisson’s ratio (0.23–0.37) increases with the soil clay content, mostly for low cement dosages.  Results underline the importance of density for estimating the Young’s modulus from UCS.

a r t i c l e

i n f o

Article history: Received 12 September 2017 Received in revised form 23 January 2018 Accepted 29 January 2018

Keywords: Mix-design Shrinkage Static modulus Pore size P-wave velocity

a b s t r a c t The study focuses on hydro-mechanical properties of soil-cement mixtures. This material produced by Deep Soil Mixing is used as foundations or cutoff walls and therefore must have appropriate mechanical properties and hydraulic conductivity. Laboratory soilcretes were manufactured with various amounts of cement and types of soil. Compared with ordinary concrete, drying shrinkage is particularly high. The results highlight the link between the hydraulic conductivity and the mix design parameters through the pore size distribution. Tests investigating Poisson’s ratio provide information on this understudied parameter. Several empirical relationships are proposed for estimating static modulus, strength and permeability from physical parameters. Ó 2018 Elsevier Ltd. All rights reserved.

1. Introduction The soil-mixing and the jet-grouting are soil improvement methods using a hydraulic binder. Many sophisticated mixing processes have emerged since their introduction in the 1950s. In Japan, these methods are mainly used for the embankment stability and reduction of settlement. In France, the method was first tested in 1986 [1], for railway line reinforcement. When the structures do not require high mechanical performance, the Deep Soil Mixing is now often used in Europe as an alternative to traditional methods of underpinning, cutoff wall, and foundations. It is thus essential to have a good knowledge of hydraulic conductivity, compressive strength, and Young’s modulus which are the main properties necessary for the soilcrete structure design [2].

⇑ Corresponding author. E-mail address: [email protected] (A.-L. Beaucour). https://doi.org/10.1016/j.conbuildmat.2018.01.177 0950-0618/Ó 2018 Elsevier Ltd. All rights reserved.

In the field of special foundations, the Water/Cement ratio (W/ C) depends on the soil clay content. This ratio is generally quite high, because the mixing process requires a self-compacting mixture. Overall, the high water content and the small particle size of the soils [3] considerably limit the mechanical properties of soilcretes. The high porosity of these materials makes them more vulnerable to different chemical aggressions. Thus, it is important to analyse the porous network characteristics that affect the transfer properties, the porosity and the hydraulic conductivity as the two most important durability indicators. Clayey soils are problematic soils due to their high deformability even after the soil treatment. The soilcretes are particularly sensitive to drying-shrinkage, so that protection is essential for structures exposed to air [4]. Soil-cement Young’s modulus may also significantly affect the distribution of load between soilcrete and surrounding ground [5]. Soilcrete is a much ‘‘softer” material than concrete and its Young’s modulus is 5 to 8 times lower than that of an ordinary concrete [6]. Actually, the Eurocode 2 (EN

658

O. Helson et al. / Construction and Building Materials 166 (2018) 657–667

all the mixtures (32 cm diameter mini-slump flow). Without clay in the soil, the desired workability cannot be obtained due to the segregation of Fontainebleau sand (uniform particle size). Thus, K0 soilcretes were designed with the same effective cement to water ratio (C/W) as the other mixtures. The C/W effective weight ratio corresponding to the amount of cement divided by the amount of water mixing reduced by the amount of water retained by clay (96% of its weight in water). Sand, clay and cement have previously been mixed in a dry state for 5 min and then with water for 10 min in a mortar CONTROLAB mixer. Cylindrical molds are filled in three layers by tapping method. For each layer, the mold was tapped against bench 15 times, except for K0 mixtures because the lower flowability requires higher energy to remove the entrapped air. Each layer was therefore vibrated for 20 s on a vibrating table. The samples are removed from the molds after 7 days, wrapped in wet textile and placed in sealed plastic bags. This storage method preserves moisture and prevents soilcretes from early drying. Thereafter, the designed mixtures will be identified using abbreviations related to their soil’s clay content and cement dosage. For example a soil containing 75% of SpeswhiteTM kaolinite and 25% Fontainebleau sand, and treated with 200 kg/m3 of cement will be named K25C200.

1992) proposes relationships that are currently used to estimate Young’s modulus but the predictive models proposed must be adapted to soilcrete’s specific behaviour. Whatever the method used, the design of mechanical properties is complicated [7,8]. According to Bellato (2013), the hydromechanical properties of soilcretes are mainly influenced by the quantity of cement injected and its hydration conditions [9]. This paper focuses on shrinkage, hydraulic conductivity and elastic properties, which are important characteristics but understudied according to the mix-design parameters. Although many previous studies examined the hydraulic conductivity and proposed large quantities of data [10,11], the transfer properties are rarely related to the soil type or to the cement dosage, let alone to the microstructure as we are going to suggest. Given the composition of soilcrete, the Poisson’s ratio is more likely to vary considerably. This parameter was therefore determined to provide a basis of modelling activities and determine as precisely as possible the dynamic modulus. A part of this work focuses also on mechanical damage assessment under mechanical loading and therefore allows a better understanding of the material’s behaviour in the operation phase. In this study, the large numbers of experimental results are linked to proposed correlations between the various physical and/or mechanical properties for different compositions of soilcretes. The aim is to propose mathematical relationships to determine the properties requiring expensive or time-consuming procedures from properties that can be obtained more quickly and simply.

2.2. Hydric behaviour 2.2.1. Water porosity and mercury intrusion porosimetry The water porosity under vacuum (g), the dry (qd) and wet bulk (qh) density are determined on cylindrical specimens (40 mm diameter  100 mm height) according to NF P18-459 standard [15]. The mercury intrusion porosimetry was used to analyze the microstructure of soilcretes. The tests were carried on a Micrometric AutoPore IV porosimeter. Cubic samples of 18 mm side were sawn and dried at 60 °C before the test. This testing includes two successive intrusion/extrusion phases. The maximum pressure applied was 200 MPa. Therefore the pores diameter accessible by mercury vary between 0.006 lm and 404 lm.

2. Experimental methods 2.1. Preparation of test specimens The soilcrete mixtures were designed as those presented by Helson et al. (2016). The different proportions and some properties of the different mixtures are given in Table 1. Six artificial soils are prepared by substituting sand with different volume proportions of clay. The soils produced in the laboratory are composed of Fontainebleau sand NE 0/1 (XP P 18-545 standard) and kaolinite SpeswhiteTM, which are widely used in the physical modeling of soils [12]. Two cement dosages were tested independently on the soil type. The cement used was a CEM III/C 32.5 N CE PM-ES NF ‘‘HRC” containing more than 81% of blast furnace slag [13], responsible for the slow development of strength, but whose latent character is favorable in terms of workability [14]. The amount of water is fixed according to the deep mixing methods requiring a self-compacting consistency. The mixing water amount is adjusted to keep a constant workability between

2.2.2. Shrinkage Shrinkage tests were carried out according to NF P 15-433 for hydraulic mortars [16]. For each mixture, shrinkage was measured on three soilcretes prisms 40  40160 mm3, removed from the molds after 3 days of moist curing and then placed in a climatic chamber at 20 °C and 50% relative humidity. Reference stainless steel studs are casted into the mid-points of the top and bottom faces of the prisms. The measurement is carried out using a metal frame equipped with a comparator. Calibration of the length is achieved by an Invar bar of 160 mm long.

Table 1 Parameters of the analyzed mixtures per cubic meter of soilcrete and workability. Kaolinite [%vol]

Water [%mass]

Cement [kg/m3]

Clay (kaolinite) [kg/m3]

Sand [kg/m3]

Water [kg/m3]

C/W mixing

C/W effective

Workability

0

20 27

10

31

15

36

25

47

50

74

0 0 67 63 125 115 173 158 243 220 347 299

1534 1441 1295 1215 1144 1059 998 912 743 672 353 305

352 353 417 414 451 452 488 492 557 559 664 667

0,57 0,85 0,48 0,73 0,44 0,66 0,41 0,61 0,36 0,54 0,30 0,45

0,57 0,85 0,57 0,85 0,60 0,87 0,61 0,87 0,61 0,85 0,59 0,78

Slump

5

200 300 200 300 200 300 200 300 200 300 200 300

[cm]

Diameter mini-slump flow

0,9 ± 0,6 3,3 ± 0,4 32,0 ± 1,4 33,0 ± 1,2 31,4 ± 0,8 33,4 ± 0,7 32,8 ± 1,0 33,0 ± 1,2 32,0 ± 2,0 31,6 ± 1,8 32,8 ± 1,5 33,6 ± 1,7

O. Helson et al. / Construction and Building Materials 166 (2018) 657–667

659

Fig. 1. Experimental device for hydraulic conductivity measurement: a) GDS controllers and tri-axial cell, b) Schematic representation of the cell [20].

Fig. 2. Determination of elastic properties: a) instrumentation with strain gauges and b) an example of stress strain curve during cyclic loading.

2.2.3. Hydraulic conductivity The protocol set up in the laboratory is based on the research work of Åhnberg (2006). The steady-state intrinsic permeability k of soilcrete was determined in a triaxial cell using the Darcy law:

kðm=sÞ ¼ RT 

Q lL  107 DPAq

ð1Þ

k (m/s) is the hydraulic conductivity, RT is a correction factor to take into account the temperature variation, Q (m3/s) is the flow rate, m (Pa s) is the water dynamic viscosity, L (m) is the sample length, DP (Pa) is the pressure difference between the sample inflow and outflow, A (m2) is the sample cross-sectional area, q (kg/m3) is the water density. The experimental device is illustrated in Fig. 1. According to the literature, the length of the tested samples varies from 20 to 100 mm [17–19]. In this study, the choice of the specimen length is based on less permeable soilcretes mixtures in order to limit the test duration to 24 h. The cylindrical specimens for permeability tests were Ø = 50 mm of diameter and L = 25 mm of height. The hydraulic conductivity measurement is carried out using a hydraulic gradient of one hundred. 2.3.Mechanical properties 2.3.1. Elastic static modulus determination The uniaxial compression tests are performed on cylindrical specimens (50 mm diameter  100 mm height), sanded at the ends to obtain two parallel bearing surfaces. The loading is then performed in controlled stress rate of 0.04 MPa/s, with an electromechanical InstronÒ press. Some soilcretes specimens are instrumented with strain gauges in order to determine the elastic

properties of the material (Fig. 2a), that is to say the Poisson’s ratio (m) and the static Young’s modulus (E). At mid-height of the specimen and diametrically opposed, two strain gauges are attached in axial position and the two others in a transverse position. Cyclic loading then enables access to mechanical damage in function of the applied load. A stress-strain curve illustrates the result of a cyclic compression test to determine the Young’s modulus at different stages of mechanical loading (Fig. 2b), ie for each loading/unloading cycle (up to 12 cycles have been carried out). The static Young’s modulus is the slope of the stress versus strain curve. This parameter is calculated from the linear part of the curve during the unloading phase and the first cycle is carried out at 1.5 MPa. The static moduli presented hereafter are determined for a loading level of 0.3UCS. 2.3.2. Elastic dynamic modulus determination The measurement of P-wave velocities (Vp) is performed using a Pundit7 (Fig. 3). Vp measurement on soilcrete specimens consists in emitting an ultrasound signal in the form of a pulse and analyzing its propagation in the sample. The transit time of the ultrasonic wave is measured between two piezoelectric transducers, a transmitter and a receiver, placed in contact with the sample facing each other. After that, dividing the distance travelled by the wave travel time help to determine the speed of propagation. The tests are performed at the wet state (after getting out from the sealed bag), as a function of curing time and before each compression test. This test gives information on porosity, cracking state, and the elastic properties. The dynamic elastic modulus (Eq. (2)) is especially calculated from the P-wave velocity (Vp), the density of hardened concrete (q) and Poisson’s ratio (m) the chosen value is that

660

O. Helson et al. / Construction and Building Materials 166 (2018) 657–667

Fig. 3. Determination of dynamic elastic modulus by ultrasonic measurement. Fig. 4. Relation between water-accessible porosity and density.

which was measured with strain gauges during the compression test.

Edyn ¼ qVp2

ð1 þ mÞð1  2mÞ ð1  mÞ

ð2Þ

3. Results and discussion

sentativeness of the laboratory soilcretes mixtures elaborated in our work.

g ¼ 0; 0359qd þ 95; 68

ð3Þ

g ¼ 0; 053qh þ 142; 1

ð4Þ

Two relationships are proposed based on the experimental results obtained after 28 days of curing time (Eqs. (3) and (4)).

3.1. Hydric behaviour 3.1.1. Porosity versus density The porosity accessible to water (g) is expressed as a function of the dry density (qd) and the wet density (qh) in Fig. 4. Porosity varies between 28% and 60% depending on the clay and cement content, which is 2–4 times higher than for ordinary structural concrete due to the high amounts of mixing water. The variation of the porosity with density (qd or qh), for the studied formulations in this work, is clearly linear and agrees with investigations carried out in Belgium on samples from 38 different construction sites of Deep Soil Mixing [11]. The good fit between our results and literature results based on construction sites confirms the repre-

3.1.2. Mercury intrusion porosity Fig. 5 shows the mercury porosity differential curves for different mixtures after 180 days of moist curing. The results reveal a bimodal pore sizes distribution for soilcretes without clay and unimodal distribution for specimens containing clay. For all the mixtures the characteristic diameter of pores, determined from the highest peak of pore size distribution curve, is close to the limit between the mesopores and macropores that is to say between 50 and 100 nm [21]. However, mixtures without clay (‘‘K0”), showing a bimodal distribution, have also a pore family with a greater pore size of 3 or 0.7 lm according to the cement dosage (Fig. 5). This larger pore size may be associated with the

Fig. 5. Differential intrusion volume versus pore diameter for soilcretes mixtures, after 180 days of endogenous curing (2 tests by mixture).

661

O. Helson et al. / Construction and Building Materials 166 (2018) 657–667 Table 2 Characteristics of the porous medium: characteristic values and proportions of the different pore families. Formulation

Diameter [nm] (peak) characteristic Ø

K0C200 K10C200 K25C200 K0C300 K10C300 K25C300

60 95 95 60 56 70

3000

700

Mercury porosity [%]

Capillary pore [%] (large) 10 lm > Ø > 50 nm

(medium) 50 nm > Ø > 10 nm

(C-S-H gel) Ø < 10 nm

27.1 ± 0.1 35.0 ± 2.3 48.8 ± 0.2 27.7 ± 1.3 36.5 ± 0.4 46.7 ± 0.2

65.0 ± 4.9 68.4 ± 1.7 63.0 ± 0.9 57.1 ± 0.9 48.9 ± 0.3 46.9 ± 0.5

21.0 ± 1.5 24.0 ± 2.1 29.1 ± 0.3 31.8 ± 1.5 41.3 ± 0.2 44.8 ± 0.7

4.0 ± 0,5 1.7 ± 0.7 2.6 ± 0.0 5.7 ± 0,1 5.5 ± 0.1 3.9 ± 0.2

Fig. 6. Water-accessible porosity, initial intrusion and trapped mercury porosity (intrusion/extrusion).

lower workability (stiff mixture) and the lack of fine particles in the sand. Moreover, for given clay content in the soil, there is an overall decrease of the pore size diameter with cement content increasing from 200 to 300 kg/m3 (Fig. 5). Table 2 shows the characteristic pore diameter and the different porosity ranges (capillary, hydrate) calculated from the mercury porosity cumulative curves. The results distinguish between large and medium capillary pores according to a concrete reference [22]. The results show that capillary porosity increases, especially with the soil clay content. This is due to the increase of the water to cement ratio (W/C) of mixes as a function of the clay content [23]. Indeed, in order to have the same workability, the higher the clay content the higher water content. Moreover, the size of the capillary pore decreases when the cement dosage increases (Table 2). In fact, the proportion of ‘‘medium” capillary pores to ‘‘large” capillary pores increase from one cement dosage to another (lowest to highest). Mercury intrusion porosimetry provides an indication of the amount of Calcium-silicate-hydrate (C-S-H), which is characterized by the presence of pores between 0.5 and 10 nm [24]. Table 2 shows an increase for pores with a diameter lower than 10 nm with the cement content. The addition of cement finally results in a refinement of the porous network in relation to a size of the capillary pores reduction and a greater quantity of C-S-H. Mercury intrusion porosimetry makes it possible to identify different types of porosity (accessible, free, and trapped) which have a strong effect on the transfer properties. Fig. 6 compares the wateraccessible porosity under vacuum with mercury initial intrusion porosity. The mercury porosity values are slightly lower than those of the porosity accessible to water (about 1.7% in absolute value). This is because all the connected pores are accessible to water, even the smallest pores, thanks to the high wetting properties of water. The volume that remain in the sample at the end of mercury extraction is named the trapped porosity and is linked to the

Hydrate pore [%]

Fig. 7. Hydraulic conductivity according to mix-design parameters and curing time.

heterogeneity of the porous network geometry. During the second injection, because the trapped porosity is still filled by mercury, the injection occurs only in the free porosity. The trapped mercury porosity is defined such as the difference of the total porosity (first injection) and free porosity (second injection). The results show that the trapped porosity is between 34 and 45% of the total porosity (Fig. 6). The highest proportion of trapped porosity was observed for low cement content and soil clay content (K0C200 and K10C200). 3.1.3. Hydraulic conductivity Fig. 7 shows the evolution of the hydraulic conductivity (k) of six different mixtures function of curing time (7, 28, 180, 365 days). The hydraulic conductivity values vary between 2.107 and 3.1011 m/s. This range of value is consistent with the values obtained in a research work carried out on Deep Soil Mixing (DSM) material, where the values of k were between 108 and 1012 m/s [11]. Regarding the effect of curing time on the permeability, the results didn’t show a clear trend for mixtures with clay. However, a significant decrease of permeability as a function of curing time was observed for the mixtures without clay (Fig. 7), that is consistent with their significant strength increase. The strength development of the mixtures without clay is slower than that of the mixtures with clay [23]. This could be related to the filler effect of clay that limits the pores size regardless of hydration kinetic. A physico-chemical nucleation-type effect of the clay can also be considered as it could accelerate hydration phenomena, especially at early stages. The results show also that the hydraulic conductivity increases as a function of the soil clay content and decreases when the cement dosage increases. Fig. 8 represents the evolution of the permeability and porosity for all specimens after 180 days of curing. A quasi-linear increase of the porosity as a function of the clay con-

662

O. Helson et al. / Construction and Building Materials 166 (2018) 657–667

Fig. 8. Comparison between the estimated hydraulic conductivity from the pore diameter and the measured conductivity.

tent may explain the observed increase of hydraulic conductivity. For the specimens with a same clay content, the hydraulic conductivity decreases with the cement dosage despite of no variation of pore volume. As the results of mercury porosity showed above, the increase of cement dosage can cause a decrease of the capillary pore size that can explain also a decrease of hydraulic conductivity. In the literature, some studies show that the influence of cement content on the soilcrete permeability is particularly important when the cement content varies between 5 and 10%. On the other hand, beyond 10% the cement has much less influence on the material transfer properties [19], which is the case for all of our mixtures. 3.1.4. Permeability-porosity relationship There are many models in literature, based on different hypotheses, to estimate the permeability of porous materials from the geometry characteristic of pore network. A classical modelling consists in considering the porous medium as an assembly of channels parallel to each other [25,26]. This model describes the hydraulic conductivity (k) as a linear function of the porosity (g) and a quadratic function of the size of the channels, and therefore of the characteristic pore size (Ø) in meters.

kðm=sÞ ¼ g£2 =32  107

ð5Þ

For mixtures containing clay whose pore distribution is unimodal, Ø corresponds to the pore diameter of the single observed peak. However, for mixtures without clay whose pore sizes distribution is bimodal, Ø corresponds to the pores diameter of peak with the lowest associate diameter. The comparison between experimental values of permeability ‘‘k exp” and estimated values ‘‘k est” according to Eq. (5) for all formulations after 180 days of curing are presented in Fig. 8. The good agreement between the estimated values of k and measured values can be explained by a very good interconnection of pores for a W/C ratio beyond 0.7 [27]. It is also interesting to note that the values of hydraulic conductivity are slightly overestimated for K0 mixtures. This could be caused for the K0 mixtures, peaks on the pore sizes distribution curve (Fig. 5) are more rounded and therefore the variation of pore diameter is more important in comparison with mixtures showing a sharply defined intrusion peak. 3.1.5. Shrinkage Fig. 9a shows the shrinkage values obtained as a function of time for different mixtures. The results show that the shrinkage depends mainly on the soil clay content and rather little on the amount of cement. In fact, with 25, 10 and 0% of clay content, the shrinkage values after 25 days of drying are on average 20,700, 9400, and 3000 lm/m, respectively (Fig. 9a). The values obtained are about 3 to 40 times higher than those of ordinary

Fig. 9. Shrinkage a) and mass loss b) of soilcrete prisms during drying at 50% relative humidity according to NF P15-433 standard.

concretes and self-compacting concretes [28,29]. The results obtained for K0, however, are rather consistent with those of Guimond-Barrett (2013). The shrinkage values measured by this author after 23 days of drying at 65% of relative humidity and 20 °C, are about 2800 lm/m for sand-cement mixtures and 27,900 to 21,000 lm/m for silt-cement mixtures [5]. The mixtures with clay (K10 and K25) have a shrinkage value 3 to 7 times higher than those without clay, depending of clay content. The kinetic of shrinkage is especially important during the first 10 days of exposure to drying and then slows down and finally it looks stabilized after about 25 days of drying. This is due to the high porosity of soilcrete, which leads to a significantly faster drying kinetics than ordinary concretes (Fig. 9b). The soilcrete specimens lose about 80% of their free water after only 5 days of drying in the climatic room. Increasing cement dosage also slightly slows down shrinkage kinetics (between 8 and 20 days). This delay is due to the slight reduction in porosity and hydraulic conductivity caused by the high dosage of cement. The total shrinkage and the mass loss seem to be linked by a power function, showing that departure of free water is one of the main driving forces of shrinkage (Fig. 10). 4. Mechanical behaviour 4.1. Unconfined compressive strength Compressive strength of soilcretes have already been studied in a previous research work [23,30]. Results showed that the unconfined compressive strength ranges from 0.7 to 21.3 MPa (Fig. 16)

663

O. Helson et al. / Construction and Building Materials 166 (2018) 657–667

Fig. 10. Evolution of shrinkage depending on mass loss of samples for several drying times. Fig. 12. Influence of cement content on static modulus.

according to the parameters of the designed mixtures and the curing time. An optimum strength was found with 15% of clay. A low substitution rate of sand by clay, improve the granular stacking and, therefore less water and cement are needed to fill voids. This improves the sand grains coating by the cement leading to a better interface. Given the high water retention capacity of clay, the quantity of water increases for higher clay contents which generates a capillary porosity causing the reduction of the unconfined compressive strength. Moreover, many researchers have shown that UCS is proportional to the logarithm of the curing time. Thus, it is possible to predict the evolution of compressive strength versus time knowing the strength at 28 days of curing [31]. 4.2. Static elastic moduli Fig. 11 shows the evolution of Poisson’s ratio (m) and static Young’s modulus of elasticity (E) as a function of the soil clay content for two cement dosages after 180 days of curing. Each of the points corresponds to an average value of two tested specimens. Static Young’s modulus decreases as a function of clay content. This loss is important for low clay content and the modulus drop can reach about 50% of the initial value for the soil containing 25% of clay. The evolution of static Young’s modulus (E) depending on the soil clay ratio (K) for two dosage of cement was fitted to a square root function:

Fig. 13. Axial elastic modulus normalized by E30 evaluated at 0,3UCS versus the stress level.

pffiffiffiffi E200 ðGPaÞ ¼ 16:50  15:98 K

r2 ¼ 0; 98

pffiffiffiffi E300 ðGPaÞ ¼ 20:74  20:26 K r2 ¼ 0; 99

Fig. 11. Evolution of static modulus and Poisson’s ratio depending on clay content (at 180 days).

ð6Þ

ð7Þ

The Poisson’s ratio increases with the soil clay content, mostly for low cement dosages (200 kg/m3), ranging between 0.23 and 0.37. For low clay contents (0–5% of clay), cement compensate for the lack of fine elements due to the uniform particle size distribution of Fontainebleau sand. Thus, the Poisson’s ratio coefficient of mixtures C300 is higher for K0 mixtures. Above 5% of clay in soil, the Poisson’s ratio value is lower for C300 mixtures because the cement improves the cohesion of soilcretes. Based on the values of the Poison’s ratio found by Terashi and Kitazume (2015), ranging between 0.2 and 0.45, the soilcrete mixtures in this work take into account most of the cases encountered in situ [32]. For the specimens without clay in the soil the static modulus varies between 16.6 and 20.7 GPa, for a cement content of 200 kg/m3 and 300 kg/m3, respectively. A cement content increasing from 200 to 300 kg/m3 allows keeping the Young’s modulus above 13.5 GPa for soil with a clay content of less than 15%. The low elastic moduli of Deep Soil Mixed materials (10–15 GPa) can be

664

O. Helson et al. / Construction and Building Materials 166 (2018) 657–667

Fig. 14. Poisson’s ratio versus the stress level: a) 200 kg/m3 and b) 300 kg/m3 of cement.

Fig. 15. Comparison between static and dynamic modulus: using real value (gauges) or various fixed Poisson’s ratio (0.23, 0.3, and 0.37).

interesting for rigid inclusions since a low stiffness limits the generation of internal stresses under horizontal loading [33]. Each point in Fig. 12 represents the static modulus of elasticity value obtained with 300 kg/m3 of cement as a function of that obtained with 200 kg/m3 of cement for a given mix design after 180 days of moist curing. A linear relationship with a very good coefficient of determination was determined between E200 and E300 (Eq. (8)). Therefore an increase of the cement content from 200 to 300 kg/m3, increases on average by 25% the static modulus of elasticity.

ð8Þ

Fig. 16. Correlation between UCS and P-wave velocities measurement: without taking account the density a) and considering the density b).

Uniaxial compressive-cycling tests of increasing amplitude have been performed to investigate and quantify the contribution of microcracking to the moduli of soilcretes. In this research, mechanical damage is assessed by measuring the tangent Young modulus evolution during the unloading phase of stress-strain curve after different mechanical loading levels (Fig. 13). The evolution of Young modulus during axial loading are normalized by the initial Young’s modulus determined at 0,3UCS (E30).

For all mixtures, a significant decrease of modulus is observed with increasing stress level, which indicates a growth and propagation of cracks. The decrease seems to be progressive from the beginning of the loading until the failure of the specimen, and more important for lower cement dosages (200 kg/m3). The E/E30 evolution highlights two spindle-shapes which depend mainly on the cement content. For a stress level of about 80% of the failure compressive strength, the Young’s modulus drops from 13to 25% and from 25to 37% for a cement content of 300 and 200 kg/m3,

E300 ¼ 1:25  E200

r2 ¼ 0; 97

4.3. Evolution of elastic moduli with increasing load

665

O. Helson et al. / Construction and Building Materials 166 (2018) 657–667

estimations of the static Young’s modulus (r2 = 0.96) from the Pwaves velocity and the density with the measured Poisson’s ratio for each mixture. Fig. 16a shows a correlation between the unconfined compressive strength (UCS) and the P-waves velocity (Vp) for all the mixtures. These parameters were measured between 7 days and 2 years of endogenous curing. The results obtained are compared with Goto (2000), Yessiler (2000), Porbaha (2005) studies [35]. The results are dispersed and the margin of error seems to increase as a function of P-wave velocity. However, the unconfined compressive strength seems to follow a power law according to Vp as:

UCSðMPaÞ ¼ 4; 14  1012  Vp ðm=sÞ3;539

Fig. 17. Comparison between prediction models and experimental measurements.

respectively. The drop of static Young’s modulus of elasticity is greater than that of ordinary concretes (C25/30), for which the expected damage is about 10% at the same stress level [34]. Fig. 14 shows the evolution of the Poisson’s ratio (m) as a function of the compressive stress with 200 kg/m3 of cement (Fig. 14a) and 300 kg/m3 of cement (Fig. 14b). The results show that m is relatively constant for all the mixtures throughout the loading and increases just before the failure. This evolution indicates an isotropic damage when the stress is low and when the ultimate rupture is impending, however, the coalescence of cracks occurs in the direction of the applied load. Regarding the influence of mixtures design, a high clay content or insufficient amount of fine particules with a low cement content cause an increase in Poisson’s ratio according to the applied stress level. Unlike other soilcretes, this indicates that cracking is mostly generated along the axis of the applied stress. 5. Correlations between various physical and mechanical parameters The results of the evolution of the uniaxial compressive strength (UCS), P-wave velocity (Vp) and dynamic elastic modulus until 180 days of curing for studied soilcretes in this work are presented in detail in Helson et al. (2016). Here, the results completed until 2 years of curing are used to establish some correlations between UCS, Vp and the static elastic moduli (E30) presented above. 5.1. Prediction of mechanical properties from non-destructive methods A comparison between the static and dynamic elastic moduli, determined after 180 days of moist curing, is shown in Fig. 15. The dynamic modulus was calculated for each mixture by substituting Vp, q, and m determined from the strain gauges into Eq. (2). Whatever the parameters of the designed mixtures, the dynamic modulus is about 16% higher than the static modulus. This slightly higher value of the dynamic modulus can be explained by the essentially elastic response of the sample over a very short period of time at a very low stress level. Based on in situ results and with a Poisson’s ratio value of 0.35 it appears that Estatic and Edynamic are not well correlated [11]. Based on laboratory tests, however, it is possible to make accurate

with r2 ¼ 0; 764

ð9Þ

The strong dispersion of the results observed in Fig. 16a is mainly due to the soil clay content. In fact, at a given value of unconfined compressive strength the P-waves velocity is higher for the lowest clay content. This is explained by a decrease of porosity and an increase of sand amount (Vp quartz is 6050 m/s versus Vp clay is between 1800and 2200 m/s). Calculations taking into account the density provide a significant improvement of the UCS prediction from the P-wave velocity measurements as shown in Eq. (10).

UCSðMPaÞ ¼ 7; 089  105 ðVp5;321 =qh

4;093

Þ with r2 ¼ 0; 918 ð10Þ

The different mixtures are distinguished from each other to show that the proposed relationship doesn’t depend on the cement dosage and the soil’s clay content (Fig. 16b). 5.2. Prediction of elastic modulus from UCS values Many empirical formulas for estimating the concrete modulus of elasticity from the compressive strength can be found in the literature [36]. These models must be adapted to the specific characteristics of soilcretes (low density and high porosity). Thus, Fig. 17 compares the experimental values of static Young’s modulus determined by strain gauges measurements after 180 days of curing, with those calculated from formulas coming from design codes for lightweight or normal weight concretes. The static Young’s moduli calculated according to Eurocode 2 (EN 1992) and American Concrete Institute – 318 (ACI-318) codes are compared to the experimental results (Fig. 17). According to their low density, soilcretes can be classified as lightweight concretes (qd < 2000 kg/m3). The results show that formulas established for lightweight concretes are better suited to soilcretes than models established for normal-weight concretes. The relation for normal-weight concrete provided by EN 1992 (EC2_OC) is clearly not adapted to soilcretes as in a previous study [37]. Indeed, there is little difference in density among normalweight concretes contrary to soilcretes where the variations of the amount of clay in the soil results in an increase of the amount of mixing water and therefore of the porosity of the hardened material that greatly influence the elastic modulus. The compressive strength is not sufficient for accurately estimating the Young’s modulus.

E30 ðGPaÞ ¼ 3; 024  105 qd

1;642

E30 ðGPaÞ ¼ 7  1010 qh

3;011

UCS0;377 r2 ¼ 0; 97

UCS0;361 r2 ¼ 0; 97

ð11Þ ð12Þ

Two formulas based on qd or qh are proposed based on the same principle as EC2 and ACI’s lightweight concrete equations (EC2_LC and ACI_LC respectively). The model parameters were fitted to correlate with experimental results using Excel solver based on the

666

O. Helson et al. / Construction and Building Materials 166 (2018) 657–667

least squares method (Eqs. (11) and (12)). The empirical relationship using qd and UCS is plotted on Fig. 17. The coefficient of determination close to one shows the accuracy of the prediction. 5.3. Relationships between hydric properties and non-destructive measurements The UCS can depend on complex phenomena such as the characteristics of the sand-paste interface, contrary to the elastic modulus which depends mainly on the porosity. Fig. 18 shows the porosity accessible to water evolution as a function of the P-wave velocity after 180 days of curing. The variation between these two parameters is clearly linear. Results of this study were compared with Guimond-Barrett work (2013) [5]. The samples taken from Deep Soil Mixing construction sites by coring or fresh sampling showed the same tendency (Fig. 18). Our study highlights in addition the influence of cement content and curing time. A 30% porosity difference between K50 and K0 soilcretes results in a P-wave increase from 700 to 1070 m/s. For a given value of porosity and a given curing time, increasing by from 200 to 300 kg/m3 the cement dosage leads to a higher P-wave velocity because the cement improves both the cementation and the adhesion between the sand grains. As previously shown, the transfer properties of the material depend both on porosity and on porous network geometry. Fig. 19 shows the hydraulic conductivity versus the P-waves velocity. The conductivity decreases when the velocity increases. Correlation between velocity and conductivity is significant because of the match between porosity and conductivity for soilcretes (Fig. 8).The value of the hydraulic conductivity can be estimated from non-destructive measurements with the following relation (Eq. (13)).

kðm=sÞ ¼ 3:75  103 expð0:0049  Vp Þr2 ¼ 0; 70

ð13Þ

6. Conclusion The effect of soil nature and particularly soil clay content on hydro-mechanical properties of soilcretes was studied. The comparison of the physical and mechanical properties obtained for the specimens manufactured in laboratory with those found in the literature for a worksite case study, confirms the good representativeness of the laboratory manufactured soilcretes. After 28 days of curing, the soilcretes have a sufficiently low hydraulic conductivity to be used as watertight barriers. For this

Fig. 18. Relationship between porosity and P-wave velocity.

Fig. 19. Relationship between hydraulic conductivity and P-wave velocity.

type of application, a hydraulic conductivity of less than 109 m/ s [18] or 108 m/s [38] is generally required. Despite of no effect of curing time on the hydraulic conductivity evolution for the specimens with clay after 7 days, a significant decrease of hydraulic conductivity as a function of time was observed for the specimens without clay. After 180 days of curing, k varies between 1.6  1010 m/s and 1.7  109 m/s with 200 kg/m3 of cement and between 1010 m/s and 7.3  1010 m/s with 300 kg/m3 of cement. The hydraulic conductivity increases as a function of the soil clay content due to porosity increase and decreases when the cement dosage increases due to the decrease of pore size. The results also show that the porosity varies linearly as a function of the P-wave velocity. Thus, it makes sense to estimate the hydraulic conductivity by non-destructive measurements. The results show that the shrinkage depends mainly on the soil clay content and the values obtained for soilcretes are about 3 to 40 times higher than those of ordinary concretes and selfcompacting concretes depending of soil’s clay content. The kinetic of shrinkage is also more important in comparison to ordinary concrete due to the high porosity and permeability. Because of the high sensitivity of soilcrete to the shrinkage phenomenon, protection seems indispensable for permanent structures exposed to air. Static Young’s modulus decreases as a function of clay content; the loss is about 50% for the soil containing 25% of clay. Moreover, an increase of the cement content from 200 to 300 kg/m3, increases on average by 25% the static modulus of elasticity. For all mixes, a significant decrease of modulus is observed with increasing stress level, which indicates a development of micro cracks during loading. The drop of static Young’s modulus for soilcrete is about two times greater than that of ordinary concrete, probably because the absence of coarse aggregates results in a faster propagation/ extension of cracking. However, this loss can be limited by increasing of cement dosage. In this research work, experimental data are correlated with each other and mathematical relations are proposed in order to help dimensioning the structures made by Deep Soil Mixing. The compressive strength is estimated with very good accuracy from P-wave velocity and considering the bulk density of the soilcrete. The EC2 and ACI formulas for lightweight concretes are more adapted to evaluate the soilcrete Young’s modulus than these for normal weight concretes. In this case, the Young’s modulus is evaluated from UCS with considering the bulk density of material. The main findings and relationships of this paper may vary depending on the soil and cement characteristics and on the different curing conditions. The sand used in this study is uniformly graded which certainly have a negative impact on physical and

O. Helson et al. / Construction and Building Materials 166 (2018) 657–667

mechanical performances. Compared to field mixed sample, more control was exercised over mixing process in this laboratory study. So the influence of unmixed clay lumps on soilcrete behaviour should be investigated in further studies. Acknowledgments Thanks to ‘‘SPIE fondations” company and ‘‘FNTP” federation who both financially supported and supervised the progression of the research carried out. The authors express their gratitude to these organizations. References [1] G. Cartier, E. Waschkowski, M. Poitout, H. Dunoyer, Expérimentation d’une technique innovante de stabilisation de remblais instables par colonnes de sol traité en place, ENPC’s Central Laboratory Newsletter 144 (1986) 5–18. [2] A. Porbaha, S. Shibuya, T. Kishida, State of the art in deep mixing technology. Part I: basic concepts and overview, Gr. Improv. 4 (2000) 91–110. [3] D. Bellato, P. Simonini, M. Grisolia, E. Leder, I.P. Marzano, Quality control of Cutter Soil Mixing (CSM) technology – a case study, in: Proceedings of the 18th International Conference on Soil Mechanics and Geotechnical Engineering, 2013, pp. 495–498. [4] N. Denies, N. Huybrechts, F. De Cock, B. Lameire, J. Maertens, A. Vervoort, A. Guimond-Barrett, Thoughts on the durability of the soil mix material, in: Proceedings of the 16th European Conference on Soil Mechanics and Geotechnical Engineering, Edinburg, 2015, pp. 1404–1408. [5] A. Guimond-Barrett, Influence of mixing and curing conditions on the characteristics and durability of soils stabilised by deep mixing, PhD thesis in civil engineering, University of ‘‘Le Havre”, 2013. [6] K.G. Kolovos, P.G. Asteris, D.M. Cotsovos, E. Badogiannis, S. Tsivilis, Mechanical properties of soilcrete mixtures modified with metakaolin, Constr. Build. Mater. 47 (2013) 1026–1036. [7] J. Tinoco, A.G. Correia, Prediction of soil-cement mechanical properties by EC2 modified approach based on early age test results, in: Geotechnical Special Publication, vol. 254, ASCE, 2014, pp. 25–32. [8] M. Topolnicki, Geotechnical design and performance of road and railway viaducts supported on DSM columns – A summary of practice, in: Deep Mixing Conference, Deep Foundations Institute, 2015, pp. 131–150. [9] D. Bellato, Experimental study on the hydro-mechanical behaviour of soils improved using the CSM technology (Ph.D. thesis), University of padua, 2013. [10] F.M. Leoni, G. Pianezze, Elliptical jet grouting: an innovative, viable, and effective solution: the example of the bottom plugs for the SELA projects in New Orleans, LA, in: 5th International Conference on Grouting, Deep Mixing, and Diaphragm Walls (grouting 2017), Honolulu, 2017, pp. 11–20. [11] N. Denies, N. Huybrechts, F. De Cock, B. Lameire, A. Vervoort, J. Maertens, SOIL MIX WALLS as retaining structures – mechanical characterization, in: ISSMGETC 211 International Symposium on Ground Improvement IS-GI, Brussels, 2012, pp. 99–117. [12] K. Boussaid, Sols intermédiaires pour la modélisation physique : application aux fondations superficielles, PhD thesis in civil engineering, ENPC’s central laboratory, 2005. [13] AFNOR, NF EN 197-1, Standard. Cement – Part 1: Composition, Specifications and Conformity Criteria for common Cements. French Association for Standardization, 2012. [14] Y.-N. Sheen, L.-H. Zhang, D.-H. Le, Engineering properties of soil-based controlled low-strength materials as slag partially substitutes to Portland cement, Constr. Build. Mater. 48 (2013) 822–829. [15] AFNOR, NF P 18-459, Standard. Concrete - Testing hardened. Testing porosity and density. French Association for Standardization, 2010.

667

[16] AFNOR, NF P15-433, Standard. Methods of cement testing – determination of shrinkage and swelling. French Association for Standardization, 1994. [17] H. Åhnberg, Strength of Stabilised Soils – A laboratory study on clays and organic soils stabilised with different types of binder, PhD thesis in civil engineering, Swedish Geotechnical Institute, Linköping, 2006. [18] M. Grisolia, E. Leder, I.P. Marzano, T. Mizutani, Y. Morikawa, Influence of tire chips on the mechanical properties of cement treated soil, in: ISSMGE-TC 211 International Symposium on Ground Improvement IS-GI, Brussels, 2012, pp. 325–333. [19] M. Kitazume, M. Terashi, The deep mixing method. (ISBN: 978-0-203-589632), CRC Press/Balkema, 2013. [20] K.H. Head, Manual of Soil Laboratory Testing: Volume 3 – Effective stress tests, John Wiley, England, 1998. [21] J. Daïan, ‘‘Equilibre et transferts en milieux poreux I- Etats d’équilibre,” p. 183 pages, 2010. [22] S. Mindess, J.F. Young, D. Darwin, Concrete, Edition 2. Hydration of Portland Cement, Book chapter 4, pp. 57–91 (IBSN: 0-13-064632-6), 2002. [23] O. Helson, A.-L. Beaucour, J. Eslami, A. Noumowe, P. Gotteland, Physical and mechanical properties of soilcrete mixtures: soil clay content and formulation parameters, Constr. Build. Mater. 31 (2017) 775–783. [24] T.-H. Vu, Caractérisation de la phase solide et transferts de gaz dans les milieux poreux insaturés. Etude expérimentale et modélisation appliquées à la diffusion de l’hydrogène dans les matériaux cimentaires., PhD thesis in civil engineering, University of Toulouse, 2009. [25] A. Costa, Permeability-porosity relationship: a reexamination of the KozenyCarman equation based on a fractal pore-space geometry assumption, Geophys. Res. Lett. 33 (2) (2006) 1–5. [26] Y. Deng, X. Yue, S. Liu, Y. Chen, D. Zhang, Hydraulic conductivity of cementstabilized marine clay with metakaolin and its correlation with pore size distribution, Eng. Geol. 193 (2015) 146–152. [27] J.-P. Ollivier, A. Vichot, La durabilité des bétons. (ISBN: 978-2-85978-434-8) Presses de l’école nationnale des ponts et chaussées, 2008. [28] N. Burlion, F. Bourgeois, J.F. Shao, Effects of desiccation on mechanical behaviour of concrete, Cem. Concr. Compos. 27 (3) (2005) 367–379. [29] H. Samouh, E. Rozière, A. Loukili, Interprétation des mesures du retrait de dessiccation des bétons autoplaçants (BAP), in XXXe Rencontres AUGC-IBPSA, 2012, pp. 1–10. [30] O. Helson, Comportement thermo-hydro-mécanique et durabilité des bétons de sol: influence des paramètres de formulation et conditions d’exposition, PhD thesis in civil engineering, University of Cergy-Pontoise, 2017. [31] O. Helson, J. Eslami, A.-L. Beaucour, A. Noumowe, P. Gotteland, Mechanical behavior of soilcrete mixtures”: an experimental parametric study, in: 5th International Conference on Grouting, Deep Mixing, and Diaphragm Walls (grouting 2017), 2017, pp. 1–10. [32] M. Terashi, M. Kitazume, Deep mixing – four decades of experience, research and development, in: Deep Mixing Conference, Deep Foundations Institute, 2015, pp. 781–800. [33] G. Blanc, G. Escadeillas, A. Turatsinze, A. Abdou-Chakra, S. Brûle, Durabilité de mortiers à bas module d’élasticité, in: NoMaD international Francophone conference, 2015, pp. 1–10. [34] M. Breccolotti, M.F. Bonfigli, A. D’Alessandro, A.L. Materazzi, Constitutive modeling of plain concrete subjected to cyclic uniaxial compressive loading, Constr. Build. Mater. 94 (2015) 172–180. [35] S. Larsson, State of Practice Report – Execution, monitoring and quality control, in: Proceedings of International Conference on Deep Mixing: Best Practice and Recent Advances. Deep Mixing ’05, vol. 2, 2005, pp. 732–786. [36] D.K. Panesar, B. Shindman, Elastic properties of self consolidating concrete, Constr. Build. Mater. 25 (8) (2011) 3334–3344. [37] D. Bellato, A.D. Coste, F.-W. Gerressen, P. Simonini, Long-term performance of CSM walls in slightly overconsolidated clays, in: ISSMGE-TC 211 Interntional Symposium on Ground Improvement IS-GI, vol. 3, 2012, pp. 23–33. [38] A. Le Kouby, L. Saussaye, Y. Fargier, Y. Boussafir, C. Chevalier, Y. Ananfouet, S. Patouillard, L. Gervais, N. Auger, Renforcement des digues de loire par la technique du Deep Soil Mixing; Deux méthodes proposées - suivi sur deux ans, National Days Geotech. Eng. Eng. Geol. 45 (2016) 229–236.