**3. Results and discussion**

As shown in **Figure 3**, the investigated process is mainly composed of six stages. The adopted flow sheet is principally supported by the previous works on natural brines (Janecké, 1907; Berthon, 1962; Cohen-Adad et al., 2002; M'nif and Rokbani, 2004; Hammi, 2004) usu-

graphic-tool is helpful in natural brines exploitation or valorization. In fact, it defines, during the system's evolution, the number, the nature, the composition and the relative quantity of different condensed phases that crystallize or disappear. The first treatment step consists in evaporating at 35°C the raw brine to precipitate the maximum of sodium chloride (halite).

**Figure 3.** Flow sheet of the process for the bischofite salt recovery from Sebkha El Melah natural brine.

, K+

, Mg2+/Cl<sup>−</sup>

, SO4

2−/H2

O. This useful

ally described using the oceanic quinary diagram Na<sup>+</sup>

178 Cement Based Materials

### **3.1. Experimental procedure**

Magnesium oxide powder was mixed with magnesium chloride solution mechanically to form homogenous MOC pastes. The weight of MgO is fixed and the weight of MgCl<sup>2</sup> .6H2 O has been varied. Mixtures were cast in cylindrical molds (26 mm in diameter, 50 mm high) and stored for 24 h, then unmolded and air-cured for 28 days.

The X-ray diffraction (XRD) analysis was carried out on the powdered sample using X-ray powder diffractometer (XRD PHILIPS) with Cu K radiation (λ K = 1.54 Å).

Differential thermograms were obtained using the Netzsch 449 STA F1 Jupiter thermal analysis system. The rate of heating was 15°C/min.

The microstructure of the samples was examined using scanning electron microscope, the Carl ZEIIS LEICA S430i model.

Measurement of thermal conductivity was performed in dry state using the photothermal deflection technique. Setting time was determined by using the Vicat Apparatus.

Porosity accessible to water of MOC is determined according to EN 12390-7 norm. The measurement of porosity in water under a vacuum of 0.1 bar quantifies the volume of open pores (accessible to water) using the following protocol:

Cement samples are placed in sealed desiccators and kept under vacuum of 0.1 bar for 12 h.

Previously degassed water is introduced progressively in desiccators to fill all the pores of samples, without introducing air bubbles.

Once the samples are saturated, they are kept immersed in water for 24 h, and finally we determined hydrostatic mass *msss imm* and saturated dry surface mass mss.

The porosity is calculated by Eq. (4):

$$\frac{m\_{\rm sSS} - m\_{\rm dry}}{m\_{\rm sSS} - m\_{\rm sSS}} \tag{4}$$

where

msss: the saturated dry surface mass of the sample;

mdry: the mass of sample before saturation; and.

*msss imm*: mass of sample measured in water.

#### **3.2. Studied factors and experimental domains**

According to the preparation of MOC, three quantitative factors are chosen: mass ratio of MgCl2 /MgO, stirring speed and mixing time. The corresponding variables and their levels (set according to the data of preliminary experiments and the equipment abilities) are given in **Table 2**. The two experimental responses tracked were compressive strength (Y<sup>1</sup> ) and the setting time (Y<sup>2</sup> ).

To test the direct influence of the three studied factors as well as their possible interaction effects on the measured experimental responses, we have realized a two-level complete factorial design 23 which is expected to provide excellent information concerning not only the main effects but also the double interaction effects.

The experimental design and the measured responses are summarized in **Table 2**.

Comparing MOC and Portland cement (setting time between 2 and 3 h), it is found that MOC has a faster setting. It also has better mechanical strength.

For a very short setting time (6 min), MOC has a high strength (75.48 MPa): in this case the cement is recommended for applications that require fast setting (decoration use, restoration of monuments, damaged marble, etc.).

For a longer setting time (64 min), it has a good mechanical strength (46.59 MPa): in this case the cement is recommended for applications which require a longer setting time (floor covering).

Considering that the interaction effects between three or more factors are negligible, the factor effect estimation is computed by Mathieu et al. [26]; according to Goupy [27]:

$$b\_i = \frac{\sum\_{j}^{N} \pm Y\_j}{N} \tag{5}$$

where *Sa* 2

**No. exp.**

setting time (Y<sup>2</sup>

is the pooled experimental variance, *Si*

.

**4. Identification of the influential factors**

**Coefficient Y1 Y2**

degree of freedom i, and n <sup>=</sup> <sup>∑</sup>*ν<sup>i</sup>*

**Table 2.** Factorial matrix 23

**Mass ratio of MgCl2**

**MgO**

**/**

**Mixing time (min)**

 1.42 5 650 49.47 20 2.22 5 650 46.59 64 1.42 15 650 4.55 17 2.22 15 650 21.20 37 1.42 5 1600 41.38 14 2.22 5 1600 76.40 41 1.42 15 1600 75.48 6 2.22 15 1600 20.50 32 1.82 10 1125 67.00 30 1.82 10 1125 60.54 31 1.82 10 1125 59.00 28

).

**Table 3.** Factor signification for the two responses Y<sup>1</sup>

2

**Stirring speed (rpm)**

Based on check student for an error risk *α* = 5%, it was found that tabulated = 4.303. **Table 3** summarizes the factor effects estimation for the two responses: compressive strength (Y<sup>1</sup>

Value SD t.exp P Value SD t.exp P

b0 47.464 1.279 37.087 0.000726 29.090 0.460 63.1634 0.000251 b1 −0.773 1.500 −0.515 0.657478 14.625 0.540 27.0802 0.001361 b2 −11.513 1.500 7.672 0.016568 −5.875 0.540 −10.878 0.008345 b3 11.493 1.500 7.658 0.016624 −5.625 0.540 −10.415 0.009093 b12 −8.808 1.500 −5.869 0.027819 −3.125 0.540 −5.7864 0.028592 b13 −4.216 1.500 −2.809 0.106780 −1.375 0.540 −2.5460 0.125809 b23 6.063 1.500 4.040 0.056143 1.625 0.540 3.0089 0.094979 b123 −13.691 1.500 −9.123 0.011802 2.875 0.540 5.3235 0.033522

> and Y<sup>2</sup> .

is the experimental variance estimation i, *ν<sup>i</sup>*

**Compressive strength** 

**Setting time (min)**

181

Sorel Cements from Tunisian Natural Brines http://dx.doi.org/10.5772/intechopen.74315

**(MPa)**

is the degree of freedom of the pooled experimental variance.

is the

) and

where bi is the effect estimation of the factor i, Yj is the response j, and N is the number of experiences.

The pooled variance estimation used to determine the significant factors is computed as

$$\mathbf{S}\_a^2 = \frac{\sum\_{i}^{\mathbb{R}} \mathbf{v}\_i \mathbf{S}\_i^2}{n} \tag{6}$$


**Table 2.** Factorial matrix 23 .

(4)

) and the

where

180 Cement Based Materials

*msss*

MgCl2

setting time (Y<sup>2</sup>

rial design 23

experiences.

).

msss: the saturated dry surface mass of the sample;

mdry: the mass of sample before saturation; and.

**3.2. Studied factors and experimental domains**

effects but also the double interaction effects.

of monuments, damaged marble, etc.).

*bi* <sup>=</sup> <sup>∑</sup>*<sup>j</sup>*

*Sa*

has a faster setting. It also has better mechanical strength.

According to the preparation of MOC, three quantitative factors are chosen: mass ratio of

To test the direct influence of the three studied factors as well as their possible interaction effects on the measured experimental responses, we have realized a two-level complete facto-

Comparing MOC and Portland cement (setting time between 2 and 3 h), it is found that MOC

For a very short setting time (6 min), MOC has a high strength (75.48 MPa): in this case the cement is recommended for applications that require fast setting (decoration use, restoration

For a longer setting time (64 min), it has a good mechanical strength (46.59 MPa): in this case the cement is recommended for applications which require a longer setting time (floor covering).

Considering that the interaction effects between three or more factors are negligible, the factor

*<sup>N</sup>* <sup>±</sup>*<sup>Y</sup>* \_\_\_\_\_\_*<sup>j</sup>*

where bi is the effect estimation of the factor i, Yj is the response j, and N is the number of

The pooled variance estimation used to determine the significant factors is computed as

<sup>2</sup> = ∑ *i n ν<sup>i</sup> Si* 2

*<sup>N</sup>* (5)

\_\_\_\_\_ *<sup>n</sup>* (6)

effect estimation is computed by Mathieu et al. [26]; according to Goupy [27]:

in **Table 2**. The two experimental responses tracked were compressive strength (Y<sup>1</sup>

The experimental design and the measured responses are summarized in **Table 2**.

/MgO, stirring speed and mixing time. The corresponding variables and their levels (set according to the data of preliminary experiments and the equipment abilities) are given

which is expected to provide excellent information concerning not only the main

*imm*: mass of sample measured in water.

where *Sa* 2 is the pooled experimental variance, *Si* 2 is the experimental variance estimation i, *ν<sup>i</sup>* is the degree of freedom i, and n <sup>=</sup> <sup>∑</sup>*ν<sup>i</sup>* is the degree of freedom of the pooled experimental variance.
