Clay minerals consist of two crystalline elements: a tetrahedral element (the silica, SiO₂) and an octahedral element (aluminium hydroxide, Al(OH)₃) [1] [2] . These elements are piled up to constitute the mineral layers. Their stacking determines the variety of clays.

Figure 1(a) shows the main clay minerals according to elements stacking (kaolinite, illite and smectites).

The clay paste rheological behavior is also related to the clay elements stacking. Due to the presence of water between mineral layers, these layers are sliding and the clay paste has plastic properties (Figure 1(b)) [1] - [3] .

Clay pastes are not purely viscous and have a much more complex rheological behavior which is similar to that of viscoelastic fluids of Bingham with a sliding on walls [4] .

- Below a certain stress threshold, the viscoelastic paste does not undergo a deformation.

- Above this stress threshold, it becomes deformed and the value of the stress is related to the paste shearing or flowing velocity. This stress threshold depends on the water content of the clay paste.

Indeed, due to the foliar structure of clay minerals, the paste flows by the sliding of layers between which there is water (Figure 2(a)). The stress threshold depends on the thickness of water layers.

Figure 2(b) shows a clay paste which is deformed between two plates. The down plate is fixed and the upper plate moves at a velocity V by an applied force F. The clay structure undergoes a shearing according to the de-

formation direction. The shearing stress is expressed by Equation [5] :

where S is the surface of each plate and F the force applied on the upper plate.

The gradient of the flowing velocity in the clay paste is expressed as [5] :

where H is the clay paste thickness and V, the moving velocity of the upper plate.

In these conditions, the apparent viscosity of the clay paste is given by the expression [5] :

The apparent viscosity of the studied clays has been determined experimentally (Table 1) for the need of the finite element simulations. These average values reveal a major difference in the viscosity of studied clays.

The paper deals with six clay varieties which raw materials are processed and compressed to perform con- struction materials structures density. These raw materials coming from Togo are named according to their respective deposit name and their natural colours:

- ABB: White clay of Bangéli which is a kaolin;

- ANT: Black clay of Togblékopé;

- ARA: Red clay of Albi;

- ARG: Red clay of Guérin-kouka;

- ARK: Red clay of Kouvé;

- AVK: Green clay from Kouvé.

Clay raw materials are generally a natural mixture of the following main constituents [2] [6] : kaolin clay (which consists of aluminum oxide Al₂O₃, silica 2SiO₂, water 2H₂O), feldspar which provides sodium ion (Na⁺) and potassium ion (K⁺) that allow the formation of the vitreous phase when fired, a neutral solid constituent that reduces the shrinkage but does not act on the reactions during burning. The mineral elements contained in these clays are identified by x-rays. Table 2 gives the average contents values of three of the studied clays.

Due to their visco-plastic likely behavior under load, clay pastes can be modeled by the law [7] [8] :

where σ is the normal stress induced by the loading; k defines the paste consistency; ε is the generalized defor- mation of the paste; is the generalized deformation velocity; m is the sensitivity coefficient to the velocity; n

is the coefficient of the hardening.

By fitting the experimental curves of stress-strain σ(ε) with the law of Equation (4), the obtained parameters of this equation have the magnitude order of those typically given by the literature, namely:

This law is based on the properties of the class of so called viscometric flows. In this case and for these flows, the stress tensor can be written, considering the axes in cylindrical coordinates, in the following form [9] [10] :

Then, the measurement of the strain time rate permits to define three functions called viscometric functions and that define the behavior of the clay pastes [9] [10] :

Due to the cylindrical shape of the molded clay samples, the equations are defined according to the cylindrical coordinates (r, q, z). Figure 3 illustrates these geometrical considerations. Due to the symmetrical and revolution shape, the non null stresses applied to a cylindrical element are s_rr, s_qq, s_zz, t_rz and t_zr (Figure 3(b)).

In this analysis of clay pastes flowing, we neglect the terms related to gravity, because of the high rigidity of the pastes when completely filled in the shaping cylinder. In the same way we consider only the case of sufficiently slow flows in order to neglect the inertial terms.

Under these conditions, the balance equations are written as following [6] [11] [12] :

where u_r, u_q and u_z are the velocity components according to the radial, tangent and axial directions, and p the loading pressure during the paste compaction.

With the assumption that the clay paste is incompressible, the continuity equation is expressed as follows:

It is assumed that the law governing the fluid behavior may be written as:

where is a function of invariants of the strain time rate tensor that is independent of the deformation history.

The compression were also carried out on dried samples, in the condition of a low rate of deformation (about 0, 5 mm/min, to avoid an abrupt fracture of the sample) in order to locate and follow the materials cracking.

The problem is axi-symmetric due to the samples cylindrical shape. The clay structure is supposed to be iso- tropic when compressed at a low displacement speed. The paste compression is simulated with the code ANSYS. To assess the efficiency of the simulation compared to the experimental approach, a numerical three dimension- al example was studied on a cylindrical shape sample loaded by a compression force F_z equal to 5 kN and 35 kN or a uniform pressure respectively of 7 MPa and 50 MPa. A 30 mm diameter sample is meshed with a 3D solid triangular element. The bottom of the sample is clamped and its top is considered as a rigid surface [13] [14] . The clay sample is compressed in a hollow solid cylinder which surfaces are clamped. Figure 4(a) shows the meshed sample under compression loading in the hallow cylinder and Figure 4(b) shows the corresponding ex- perimental device.

The calculation code doesn’t allow simulating the compression of viscoelasic pastes. In the present case, the simulation concerns the compression of clay structure that is considered to be an elasto-plastic and isotropic material. The modelling parameters are the Young modulus and the coefficient of Poisson associated with an asymmetric option and an incremental loading.

The raw material aggregates size plays an important role in the mechanical properties of the obtained paste. A paste with smaller particles sizes has a more high plasticity because the particles react more intensively between each other. Thus, studied clay raw materials were grinded. The resulted powders were kept at 60˚C during 24 hours in an oven until a complete dehydration. For the samples characterization, each clay paste is processed by mixing a mass m₀ of anhydrous clay powder with a mass me of water. The resulting clay paste has an average water content v expressed as:

where m_ws is the mass of the wetted sample and m_0s the mass of the same sample when dried.

The masses were measured using a METTLE PJ 360 Delta Range balance, with an accuracy of 1/100 g.

For samples processing, an average water content v of about 18% was chosen in relation with the assessment of pastes fluidity, the consistency of pastes and the external appearance of samples after shaping. The water content

of the six varieties of studied clays ranges from 15% to 20%. These water content values lie in the limits set by ATTERBERG [12] [15] [16] .

To get a clay paste having satisfactory plasticity properties, the powder-water mixtures were kneaded for about 2 hours until getting a homogeneous paste. The paste is then stored in a hermetically climatic chamber for more than 24 hours, to insure a uniform moisture content while preventing evaporation and increasing the plas- ticity of the paste under the effect of micro-organisms that play an important role in the process.

Clay pastes were compacted by compression in a duralumin hallow cylinder shown in Figure 4(b). This cy- linder has an interior diameter of about f30 mm. The loading is made by a f30 piston. To meet the manufactur- ing requirements in tiles industrial factories, the clay structure must be compressed to present a density of about 2 g/cm³. For this purpose, to obtain samples having a height of 30 mm and a diameter of 30 mm, the mass of the paste to be introduced into the mould is calculated so to obtain a structure density of 2 g/cm³ after compression.

Compression is performed by imposing the piston motion at a low speed of 3 mm/min to avoid resistance re- lated to the viscosity of the paste. Clay pastes are subject to adhesion to the die walls in industrial processes, and this is also the case for the moulding carried out in this study. In most cases, friction is below the threshold at which shear occurs. Therefore, pastes slip instead of warping. This promotes good extrusion without lubricants [17] [18] . Several samples were processed (about 24 samples). Samples were loaded at 5 kN, 20 kN and 35 kN or respectively at compression stress of about 7 MPa, 28 MPa and 50 MPa.