The soil consolidation problem under linear behaviour is governed by the diffusion equation of excess pore pressure, with well-known analytical solutions for the usual boundary conditions. However, under the hypotheses of non-linear dependencies of hydraulic conductivity and void ratio with the effective stress, as well as under the consideration of variable volume elements in thickness, numerical solutions are used. Among the most common non-linear models used in literature are those of Davis and Raymond [1, 2], Juárez-Badillo [3, 4, 5] and Cornetti and Battaglio [6, 7]. The interest of the non-linear models against the linear ones is consequence of the important deviations that emerge from their solutions, above 100% [2] both in relation to the time characteristic of the process and the evolution of the average degree of consolidation.
In this paper, we address, on the one hand, the deduction and verification of the dimensionless groups that govern the solutions of the Davis and Raymond model and the extensions derived from eliminating one or more of its restrictive hypotheses (constant 1+e, cv and dz) and, on the other hand, based on these groups, construction of universal curves that allow the engineer to obtain the solutions of the most important unknowns of the problem in a direct way. In order to verify the results obtained, a set of significant cases has been simulated, widely covering the range of values of the physical and geometric parameters and the loading ratio (final and initial effective stress) that can take place in real problems. In each case, the value of one or more of the parameters involved has been significantly altered, but the numerical value of the dimensionless groups has been retained so that it is verified that the same pattern of solutions is obtained for the whole set of cases; or the value of the groups has been modified appropriately to confirm that the form of the solutions changes.
As a deduction technique for these groups, the nondimensionalization of governing equations [8, 9, 10] has been used, a form of application of the pi theorem [11], thus allowing the dimensionless parameters involved in the problem (void ratio and compression index), and that they would form independent dimensionless groups, to be included in the inferred groups. In the nondimensionalization process, which has been carried out in terms of both excess pore pressure dissipation and settlement, the characteristic time of consolidation is introduced as a reference, an unknown that is incorporated into one of the resulting monomials and whose dependency with the rest of the groups is established by the pi theorem. Despite suppressing the restrictive original hypotheses of the authors [1], the resulting number of dimensionless groups is small enough to be able to represent their dependencies by means of universal curves obtained through numerical simulations using the network simulation method [12, 13].
For the most general and precise case (non-constant 1+e, cv and dz), the characteristic time in terms of settlement is defined by a single group, so it has a direct relationship with the parameters of the problem and a single test is sufficient to obtain the proportionality factor. However, in terms of pressure, the characteristic time is a function of the final and initial effective stress ratio. On the other hand, the average degree of settlement is a direct function of the dimensionless characteristic time, while the average degree of pressure dissipation depends, additionally, on the loading ratio, so its universal representation needs to be done by means of an abacus.
The general consolidation equation is obtained by matching, in a soil element, the temporal change in water volume with the temporal change in void volume [14]. This balance, together with the constitutive equation that associates the variables flow and pressure gradient (Darcy’s law) and the empirical expressions that relate the parameters of the problem with the dependent variable (excess pore pressure, u, or effective pressure,
Using Darcy’s Law,
expression (1) is finally written in the form
where k (hydraulic conductivity) and e (void ratio) are, in general, parameters that depend on the effective stress (
In their model [1], Davis and Raymond assumed the following classical hypotheses: i) secondary consolidation is ignored, ii) soil particles and pore water are incompressible, iii) soil is saturated, iv) soil weight is negligible and v) the thickness change is considered negligible compared to the initial thickness, this is, 1+e constant. In addition, they assumed a constant consolidation coefficient based on the fact that, in a real mass of soil, compressibility and hydraulic conductivity vary during the consolidation process, both decreasing while increasing the effective pressure but in such a way that the changes in both are compensated so that the consolidation coefficient cv =
where I
From the definition of the volumetric compressibility coefficient,
which gives us the volumetric compressibility coefficient as a function of the effective stress.
Finally, the assumption of a constant value for both the consolidation coefficient and the factor 1+e allows writing
an equation equivalent to assuming that k and
The Davis and Raymond consolidation equation [1] is derived from expression (1) and considering constant the factor 1+e, resulting in
an expression very similar to that of equation (2), with the exception that the factor 1+e does not appear squared (note that, in this case, e = eo). Since k and
which, after mathematical manipulation, results in
and, from expression (5), since c
Expression (9a), a clearly non-linear equation, is the Davis and Raymond consolidation equation [1] in terms of the effective stress (
The authors, for the usual boundary conditions [14], obtained the analytical solution for the variable (u), which besides depending on position (z) and time (t) also depends on the ratio
On the other hand, looking for a simplification (by means of a change in variable) that reduces the consolidation equation (9) to a linear one, the authors proposed the introduction of a new variable (w), to which they did not assign any explicit physical meaning, whose relation with the effective stress is in the form
The introduction of this new variable, through its spatial and temporal derivatives, into equation (9) leads to a pure diffusion equation
What is diffused in this linear equation, then, is a new local magnitude (w) that, although the authors did not mention, is directly related to the settlements (changes in the void ratio) of the problem. Thus, from equation (3), it is easy to obtain that
It follows from equation (11) that the solution for the variable w depends on position and time, and therefore, as the authors conclude [1], the average degree of settlement in the domain (Ūs) is only function of time.
Within this section, we will consider two models that eliminate some restrictive hypotheses considered by Davis and Raymond, giving rise to more general and, therefore, more precise models: i) the first one in which the value of 1+e is assumed to be non-constant but keeping the consolidation coefficient cv constant and ii) a more general second model in which both 1+e and cv are considered non-constant.
In this case, we start from expression (2), which represents the general consolidation equation under the hypothesis of 1+e non-constant. For the changes in the void ratio with the effective stress, we maintain the constitutive relation (3). On the other hand, if we consider that cv is constant throughout the process, on this criterion it is true that
so that the relation between the hydraulic conductivity and the effective pressure, necessarily, becomes now
that is, the changes in k are inversely proportional to those in
After mathematical operation, it is deduced that
which, from the deduction of expression (13), finally leads to:
For this extended model, we again begin from expressions (2) and (3). Maintaining the Davis and Raymond philosophy, we return to the relation k
With all this, equation (2) can be written as
which expressed in terms of the consolidation coefficient takes the form
or, in terms of the initial consolidation coefficient (a constant parameter), results in
In this section, we proceed to obtain the dimensionless groups that govern the solution of the non-linear consolidation scenarios presented in the previous section to subsequently address a universal representation of the main variables of interest: average degree of consolidation and characteristic time of the duration of the process. By defining the variables of the problem (effective stress, medium depth and characteristic time) in a dimensionless form, the non-linear equation is nondimensionalized. Given the different hypotheses assumed in the different models, the nondimensionalization process deduces different dependencies in each case and, at the same time, different solutions.
It will also be analyzed, for the models in which 1+e is non-constant, the influence of the elimination of another restrictive hypothesis, consequence of this, which is the consideration of the variation in the volume element size dz. This last question, which implies addressing a moving boundary problem (and even closer to the real consolidation problem), involves changes in the governing equations, which will result in different solutions for the variables of interest, as we will see in the following.
The consolidation problem describes an interesting phenomenon whereby as the excess pore pressure is relaxed (allowing water to escape from the system), the ground settlement takes place. Two processes closely linked but, due to the non-linearity of the problem, develop in a different way, with dependencies that are not always equal. Therefore, the characterization of the problem will be carried out, for all scenarios, both in terms of pressure and settlement.
For the nondimensionalization of this model (9a), we will adopt the following dimensionless variables:
normalized to the interval [0, 1];
As they are normalized variables, all derivative terms of (
from which, dividing by the first, monomials result:
that, by means of the pi theorem[11], provide the solution for the characteristic time (
where Ψ is an arbitrary and unknown function of its argument. Adopting for
which means that the characteristic time taken by the excess pore pressure to relax to approximately the value of zero is a function of the ratio between the final and initial effective stresses, in addition to being directly proportional to
In short, the monomials that govern the solution of the characteristic time associated with the dissipation of interstitial pressure throughout the soil are two:
and, therefore, the average degree of pressure dissipation (Ū
At the local level, the characteristic time will depend, additionally, on the depth z, which when expressed in dimensionless form is
where Ψ, an arbitrary function of its arguments, can also be written in the form
a result that coincides with the analytical solution (Davis and Raymond [1]) in terms of argument dependencies.
Davis and Raymond did not seem to notice that the new local magnitude (w), equation (10), is proportional to the differential void ratio ‘ef – e’, equation (12) and, under this dependency, also the local settlement. It would have seemed more illustrative and didactic, although in essence it is the same approach, starting from the empirical relation e = e(
where e
while its partial derivatives with respect to position and time are
By calculating the partial derivatives of
expressions from which it is deduced that
Now, starting from (9a), it is immediate to reach
A pure diffusion equation of the local variable differential void ratio directly related to the local degree of settlement. For its nondimensionalization, the dimensionless variables are defined as
normalized to the interval [0, 1], with
The coefficients of this equation, once simplified, result in
that, by division, provides a single monomial solution
and a value for the characteristic time of settlement (
In this way, and based on the result obtained, the average degree of settlement (Ū
a result also coinciding with that obtained by Davis and Raymond in terms of argument dependencies.
In this case, starting from equation (17), which we repeat here for convenience
and with the dimensionless variables of the expression (22), we reach
from which the following coefficients result in:
where the values of
that provide the solution for the characteristic time (
Adopting for
allowing to write now more comfortably (
On this occasion, the evolution of the average degree of pressure dissipation (Ū
If, in addition, we take into account the variation of the volume element size
resulting in the coefficients
and, dividing by the third coefficient, the monomials
Once we adopt the values
In this case, starting from (17) and through the expressions (37–39), it is immediate to reach
and with the dimensionless variables of the expression (41), we obtain
The coefficients of this equation, once simplified, result in
that, dividing by the second, provide the monomials
of which it is easy to reach
what allows, finally, to write
For the case of variable
providing the monomials
from which it is easy to reach the same monomials (61) as for the case of constant dz, allowing to write now for the characteristic time of settlement (
whereas the evolution of the average degree of settlement (Ū
In this case, starting from equation (21), which we repeat here for convenience
and with the dimensionless variables of expression (22), we have
from which the following coefficients result:
Dividing these coefficients by the first, we get the groups
Adopting for
or since
Now we can write for the characteristic time in terms of pressure (
and for the evolution of the average degree of pressure dissipation, Ū
For the variant dz, proceeding as in the previous sections, we have the coefficients
which, dividing by the first, provide the monomials
that can be simplified to
being able to write for the characteristic time in terms of pressure (
and for the evolution of the average degree of pressure dissipation, Ū
Thus, in view of the expressions (69–71) and (74–76), the fact of adopting the more general (and more precise) hypothesis, dz variable, decreases in this case the number of argument dependencies of the problem, which is reduced to only two monomials.
In this case, starting from (21) and through expressions (37–39), it is immediate to reach
and with the dimensionless variables of expression (41), we have
The coefficients of this equation, once simplified, result in
that, by dividing, provide the monomial
and, proceeding as in the previous sections, we reach
So the expressions for the characteristic time (
Finally, adding the condition of variable
These coefficients provide the monomial
and the expressions for the characteristic time (
Thus, in view of the expressions (81–83) and (85–87), the fact of adopting the more general (and more precise) hypothesis, dz variable, decreases in this case the number of argument dependencies of the problem, which is reduced to a single monomial.
Table 1 summarizes the expressions that govern each of the models addressed in this section, whereas Table 2 shows the monomials that rule their solution patterns.
Governing equations for the different variants of the Davis and Raymond model the assumption of variable dz adds the assumption of variable dz adds
Pressure
Settlement
Davis and Raymond
Davis and Raymond 1+e ≠ constant
Davis and Raymond 1+e ≠ constant
Dimensionless groups that characterize the solutions for the different variants of the Davis and Raymond model
Pressure
Settlement
Davis and Raymond
Davis and Raymond 1+e ≠ constant c
Davis and Raymond 1+e ≠ constant c
Davis and Raymond 1+e ≠ constant c
The objective of this section is the verification of the deduced dimensionless groups that govern the consolidation problem of Davis and Raymond and, once tested, the obtention of universal curves based on these groups for the variables of greatest interest to geotechnical engineers: characteristic time and average degree of consolidation.
In order not to make this section too extensive, we exclusively stick to the most general and precise Davis and Raymond model, that is, the one that considers both 1+e and cv non-constant, with the added hypothesis of variable dz. For the rest of the models, of less interest, universal solutions can be found (based on the monomials obtained ) in the works of Davis and Raymond [1] (for the original model of the authors with constant cv and 1+e) and García-Ros [16] (for the extended models with constant cv and non-constant 1+e).
Table 3 shows a series of nine simulations for the most general Davis and Raymond model (non-constant 1+e and cv, with variable dz) in which several parameters or initial values are modified in order to check and verify the dependencies of the different solutions of the models with respect to the monomials of Table 2:
Verification of the dimensionless groups for the extended Davis and Raymond model with both non-constant c
Case
Ko (m/yr)
eo
Ic
Ho (m)
cvo (m2/yr)
01
0.02
1.5
0.45
30000
1
60000
0.783
0.4941
0.4328
0.967
2.0
0.847
02
0.04
1.5
0.45
15000
1
30000
0.783
0.4941
0.4328
0.967
2.0
0.847
03
0.02
0.25
0.1125
30000
1
60000
1.566
0.4941
0.4328
0.967
2.0
0.847
04
0.04
1.5
0.45
60000
2
120000
3.133
0.4941
0.4328
0.967
2.0
0.847
05
0.03
1
0.3
25000
1.5
50000
1.175
0.926
0.811
0.967
2.0
0.847
06
0.02
1.5
0.45
30000
1
120000
0.783
0.5501
0.4328
1.077
4.0
0.847
07
0.02
1.5
0.45
30000
2
120000
0.783
2.2004
1.7312
1.077
4.0
0.847
08
0.02
1.5
0.45
30000
1
240000
0.783
0.6001
0.4328
1.175
8.0
0.847
09
0.02
1.5
0.45
30000
1
480000
0.783
0.6444
0.4328
262
16.0
0.847
For this purpose, a first reference case is established, on the basis of which we will modify the different parameters or initial values to define the other cases. For all models, the potentially variable physical and geometric characteristics are ko (m/yr), eo, Ic,
In Table 3, cases 01–05 represent different consolidation scenarios, but the monomial remains
Regarding the dimensionless form
Once the expression for the characteristic time in terms of settlement (
Regarding the problem in terms of pressure, as can be deduced from the expressions (74–76), the dimensionless form of the characteristic time is a function of
Finally, once the value for the characteristic time in terms of pressure (
The search for the dimensionless groups that govern the non-linear consolidation problem based on the Davis and Raymond original and extended models, by means of the nondimensionalization technique for governing equations, has led to simple solutions despite the enormous set of physical and geometrical parameters, in addition to those referred to the boundary conditions, involved in the problem.
By introducing as reference different characteristic times of consolidation (parameters of great interest in ground engineering) in order to nondimensionalize the real time, the groups have been deduced by means of the dimensional coefficients derived from the mathematical treatment of the governing equations. In this way, these same characteristic times can be expressed as a function of the emerging dimensionless groups.
The comparison between the most complex model (non-constant 1+e and cv and variable dz) and the original (whose groups can be deduced from the analytical expressions reported by Davis and Raymond) has given rise, curiously, to the same number of monomials, one for settlement and two for pressure (despite having introduced two new parameters in the extended model: initial void ratio and initial consolidation coefficient), which can be considered a contribution of great interest given the higher precision of the extended model.
It is worth mentioning that in the less general extended models, case i) non-constant 1+e and constant cv and
The application of the pi theorem has allowed to represent the results as a function of the dimensionless characteristic time, for both the average degree of settlement and the average degree of pressure dissipation, in the second case by means of an abacus that used the loading ratio as a parameter. It has been observed that the characteristic time of settlement is always lower than the characteristic time in terms of pressure and that this decreases depending on the loading ratio. This is, undoubtedly, due to the non-linear nature of the e∼