You are on page 1of 7

Computers and Geotechnics 109 (2019) 82–88

Contents lists available at ScienceDirect

Computers and Geotechnics


journal homepage: www.elsevier.com/locate/compgeo

Technical Communication

Application of improved radial movement optimization for calculating the T


upper bound of ultimate bearing capacity of shallow foundation on
unsaturated soil

Liangxing Jin , Hanchao Zhang, Qixuan Feng
School of Civil Engineering, Central South Univ., Changsha 410075, Hunan, China

A R T I C LE I N FO A B S T R A C T

Keywords: Considering engineering practices in terms of decrease in the matric suction of unsaturated soil with increasing
Critical slip surface embedded depth, this study, employing a new global optimization algorithm on the basis of radial movement
Improved radial movement optimization optimization (RMO), focused on investigating the critical slip surface and ultimate bearing capacity of a rough
Matric suction strip footing on unsaturated soil subjected to vertical load. Based on upper-bound limit analysis theorem, an
Shallow foundation
equation for calculating the ultimate bearing capacity of unsaturated soil has been established. The data
Ultimate bearing capacity
structure of RMO was improved by considering self-feedback of the particles. The results show that the improved
Unsaturated soil
radial movement optimization can be successfully applied to accurately and stably search the critical slip surface
and subsequently calculate the ultimate bearing capacity on unsaturated soil. It also found that, compared with
no matric suction, the critical slip surface determined considering the matric suction of unsaturated soil is larger
and deeper, and the corresponding ultimate bearing capacity is also significantly improved.

1. Introduction incorporated parameters of unsaturated soil such as the effective co-


hesion, the modified angle of friction considering dilatancy, and the soil
The determination of ultimate bearing capacity of shallow founda- and water characteristic curve (SWCC) into Terzaghi’s equations. In the
tions is key to foundation design. Many studies have focused on the previous analyses on bearing capacities, the foundation soil was typi-
ultimate bearing capacity of shallow foundations on the basis of several cally treated as an isotropic body and the matric suction was considered
theories such as the limit equilibrium theory [20,12,8,22], slip line to be uniformly distributed along depths (e.g., [16,5]. The influence of
theory [1], and limit analysis theory [3,13,19,24]. Among these the- unevenly varied distribution of matric suction with depth on the ulti-
ories, the upper-bound limit analysis theorem has been adopted by mate bearing capacity has been typically ignored in order to simplify
many scholars owing to its elegance and strict solution. In the past, the calculations. In practical engineering, however, the matric suction
many scholars have focused on the ultimate bearing capacity of shallow of the foundation soil typically decreases with depth and eventually
foundations on dry and saturated soils; however, in engineering prac- reduces to zero at the groundwater level. This uneven variation of
tice, shallow foundations are often located above the groundwater matric suction may significantly influence the ultimate bearing capacity
level, where the foundation soil is often unsaturated. Therefore, the and thus, it is of significant importance to incorporate it into more
study of the ultimate bearing capacity of shallow foundations in- accurate evaluation of bearing capacities.
corporating unsaturated soil mechanics is of great practical sig- With the advancement of computational power, many scholars have
nificance. introduced artificial intelligence algorithms into the field of geo-
Through a series of numerical analysis, experimental studies, and technical engineering, among which their application in slope stability
field investigation, scholars have demonstrated that the matric suction analysis is relatively mature [11,4]. However, none of these algorithms
of unsaturated soil increases the soil’s shear strength and hence the are perfect. For example, the genetic algorithm is relatively weak in
bearing capacity [15,25]. Fredlund and Rahardjo [6] studied the determining the local optimum solution [23]; particle swarm optimi-
bearing capacity of shallow foundations on unsaturated soil by using zation relies heavily on the values of inertial factors, resulting in a large
Terzaghi’s equations and considering the shear strength induced by amount of required calculation; and in ant colony optimization, it is
suction as a part of cohesion. Mohamed and Vanapalli [14] comparatively more difficult for the algorithm to converge, resulting in


Corresponding author.
E-mail address: jlx871162@csu.edu.cn (L. Jin).

https://doi.org/10.1016/j.compgeo.2019.01.016
Received 20 November 2018; Received in revised form 29 December 2018; Accepted 20 January 2019
0266-352X/ © 2019 Elsevier Ltd. All rights reserved.
L. Jin et al. Computers and Geotechnics 109 (2019) 82–88

poor quality of results [7]. Recently, Rahmani and Yusof [17] proposed
a new type of heuristic global optimization method, namely, the radial
movement optimization (RMO). Compared to other heuristic algo-
rithms, RMO has the advantages of fast convergence, small memory
storage, and simple calculation flow. Mahrami et al. [10], Seyedmah-
moudian et al. [18], Vanithasri et al. [21], and Jin and Feng [9] applied
RMO to the system optimization of a power load forecast engine, the
detection of global maximum power point of a photovoltaic (PV)
system, the estimation of coefficients of the fuel cost function in thermal
power plants, and slope stability analysis, respectively; all these studies
achieved satisfactory results. However, RMO also has certain limita-
tions. For example, it neglects the self-feedback of particles and dis-
regards the related information of the particles themselves, which in-
creases the randomness of the results and thus, the desired search
accuracy and stability requirements cannot be satisfied.
In the present study, an improved radial movement optimization
(IRMO), which was established by optimizing the data structure of
Fig. 1. Principle of radial movement optimization.
RMO, was applied to calculate the ultimate bearing capacity of shallow
foundations on unsaturated soil. The critical slip surface beneath the
foundation was identified based on the rigid multi-block upper-bound made between this value and the fitness value of the ith particle for the
limit analysis method. This study could provide theoretical reference previous generation so as to determine whether it is necessary to update
for the determination of the ultimate bearing capacity of shallow the fitness function value of ith particle of this generation. And choose
foundations on unsaturated soil and the optimization of foundation the optimal fitness value of all particles in this generation to update
design in the future. Rbest. If Rbest is better than Gbest, then update Gbest. The above im-
provements makes it possible for the algorithm to carry forward the
2. Improved radial movement optimization (IRMO) information of good locations among the previous generation of parti-
cles, so that the results will approach the global optimal solution in a
2.1. Basic principle of radial movement optimization (RMO) quick and stable way.
The specific steps to implement IRMO are as follows:
Radial movement optimization (referred to as RMO hereinafter) is a
global optimization algorithm designed to quickly yield the optimal (1) Initialization
solution of a multi-dimensional objective function. RMO simulates a
group of particles Xi,j moving in a gradually shrinking living space, and Randomly generate N initial particle group vectors within dimen-
the variable of each particle in the jth dimension has its value range, sion M, represented as Xi1, as shown in Eq. (1). The jth variables of each
namely, [minxj, maxxj]. The new generation of particles is produced by particle can be obtained from Eq. (2), and the value range for variables
randomly generating offsprings near the center of the living space. of each particle in the jth dimension is [minxj, maxxj](1 ≤ j ≤ M).
In RMO, a particle's location in the space represents a solution Calculate the fitness function value corresponding to the initial particle
vector (in this study, it refers to a critical slip surface), and the space f (Xi1), take the location of the best fitness as the initial center location
dimension to be solved M stands for the dimension of the solution Center1, and take Center1 as the global optimal location Gbestl1.
vector (in this study, it refers to the number of variables required to Xi1 = [x i1,1, x i1,2, …, x i1, M − 1, x i1, M ] (1 ⩽ i ⩽ N ) (1)
characterize a critical slip surface). Each solution vector is evaluated by
the fitness function (in this study, it refers to the ultimate bearing ca- x i1, j = min x j + rand (0, 1) × (max x j − min x j ) (2)
pacity). After a comparison of the fitness values of all particles in the
current generation, the optimal solution of this generation Rbest is ob-
tained, and the best solution ever obtained among all generations, (2) Generate a Pre-location
namely, the global optimal solution Gbest is updated. The center point
of the living space, that is, the Center is determined by Rbest and Gbest Randomly generate N new location points as the pre-location Yik
together and changes as the evolutionary algebra varies. The principle based on the center location, and the pre-location of the kth generation
of RMO is shown in Fig. 1. can be obtained from Eq. (3). Wk (it can help to narrow the solution
space down to a point) is a coefficient that decreases with the algebra
2.2. Improved radial movement optimization (IRMO) and Its Basic Steps and can be calculated by Eq. (4), where G represents the maximum
number of iterations.
Like other algorithms, the RMO algorithm also has its limitations. 1 1
Particles of each generation in RMO are generated based on the center yik, j = Center jk + rand ⎛− , ⎞ × (max x j − min x j ) × W k
⎝ 2 2⎠ (3)
location, which makes the algorithm dependent on the center location
and potentially misses out optimum solutions from the previous gen- k
Wk = 1 −
eration. Therefore, the solution may deviate from the global optimal G (4)
solution, leading to unstable results. Hence, in order to make up for
instability exposed in searches conducted by RMO, an improved radial (3) Update the Optimal Location
movement optimization (hereinafter referred to as IRMO) is proposed
in this study based on certain improvements made to the data structure Calculate the fitness function value of the ith pre-location fi(Yk),
of RMO. Compared with the RMO algorithm, the IRMO algorithm compare it with the corresponding fitness value of the previous gen-
mainly has the following improvements. eration fi(Xk−1) according to Eq. (5), and take the minimum value of
In IRMO, when a new generation of particles is generated, a new the two as the fitness value of the ith particle of this generation. Com-
location point is randomly generated as the pre-location at first, then pare the fitness values of all particles in this generation, and choose the
the fitness value of the ith pre-location is evaluated and a comparison is minimum value to update the optimal solution of this generation Rbestk.

83
L. Jin et al. Computers and Geotechnics 109 (2019) 82–88

If Rbestk is superior to the global optimal solution Gbestk, then Gbestk


also needs to be updated.

fi (X k ) = min(fi (X k − 1), fi (Y k )) (5)

(4) Generate a New Center Location

A new center location is generated according to Eq. (6), where C1


and C2 are proportional coefficients which affect the speed and accu-
racy of calculation convergence. Rahmani and Yusof [17] showed that
the C1 and C2 should satisfy the relation 0.4 ≤ C1 < C2 ≤ 0.9. When
the fitness function is unknown, C1 and C2 prefer lower values. During
the present study, the parameters adopted are C1 = 0.4 and C2 = 0.5. If
the algorithm does not reach the maximum number of iterations G, then
return to step (2) to continue the calculation until the algorithm reaches
the maximum number of iterations G, and the algorithm stops. At this
time, the global optimal solution GbestG is the minimum value of the
multi-dimensional function we seek to obtain.

Center k + 1 = Center k + C1 × (Gbestl k − Center k )


+ C2 × (Rbestl k − Center k ) (6)

3. Ultimate bearing capacity of strip footing on unsaturated soils


Fig. 3. Compatible velocity field.
3.1. Rigid multi-block failure mechanism

Zhao et al. [27] proposed a two-dimensional kinematic failure me- sin(αi − 1 + βi − 1 − βi )


Vi - 1, i = Vi − 1 (i = 2, 3, ⋯, n)
chanism for unsaturated soil. It assumed that (2n + 1) rigid blocks are sin(βi − 2φ′) (8)
involved in the translational failure of a rigid block beneath a strip
foundation as shown in Fig. 2. And half of the failure surface is used for sin(π /2 + φ′ − θ)
V1 = V0
analysis due to symmetry. φ′ is the effective internal friction angle. B0 is sin(β1 − 2φ′) (9)
the width of foundation. δi represents the angular increment of the base
sin(π + 2φ′ − αi − 1 − βi − 1)
of the i-th rigid block relative to the base of the (i − 1)-th rigid block. Vi = Vi − 1 (i = 2, 3, ⋯, n)
The variable parameters θ, αi, and βi govern the shape of rigid blocks sin(βi − 2φ′) (10)
and subsequently determine the failure mechanism. The rigid block
ABC is labeled as block No. 0 and the remaining rigid blocks are
3.2. Equations for calculating the ultimate bearing capacity on unsaturated
numbered 1, 2, … , n from the left to right.
soils
According to Fig. 2, the lengths of the rigid blocks, namely,
Li(i =1,2, … ,n), Di(i = 1,2, … ,n), and B1 as well as the area SABC,
Zhao et al. [27] derived the ultimate bearing capacity formula of
Si(i = 1,2, … ,n) can be calculated from the triangle geometry. The
unsaturated soils under vertical load based on the upper-bound limit
compatible velocity field is shown in Fig. 3. V0 is the speed vector value
analysis theorem. According to the principle of virtual power, energy
of rigid block No. 0, Vi−1,i (i = 1,2, … ,n) and Vi (i = 1,2, … ,n) are the
consumption mainly includes external force work and internal energy
velocity vector values at the junction Li, Di of the i-th rigid block. Vi−1,i
dissipation. The external force work for the failure mechanism includes
(i = 1,2, … ,n) and Vi (i = 1,2, … ,n) can be obtained according to
WPu, Wsoil and Wq. They are the work of the ultimate bearing capacity
Fig. 3, as shown in Eqs. (7)–(10), respectively.
Pu, the gravity work of (2n + 1) rigid blocks on both sides of the
sin(π /2 + φ′ + θ − β1) foundation and the work of surcharge q, respectively. And expressions
V0,1 = V0 for all of them are as follows in Eqs. (11–13), where ϕi stands for the
sin(β1 − 2φ′) (7)
angle between the direction of gravity and the direction of velocity for
any rigid block in the failure mechanism and is expressed in Eq. (14).
WPu = B0·Pu·V0 (11)
n
Wsoil = SABC·γ ·V0 + 2 ∑ Si·γ ·Vi ·cosϕi
i=1 (12)


Wq = 2·q·B1·Vn·cos ⎛ − α n − βn + φ′⎞
⎝ 2 ⎠ (13)
i−1
π
ϕi =
2
+ φ′ + θ + ∑ αj − βi (i = 1, 2, ⋯, n; j = 1, 2, ⋯, n − 1)
j=1

(14)
The internal energy dissipation only occurs along the velocity dis-
continuity lines Li, Di (i = 1,2, … ,n). In the calculation of internal en-
Fig. 2. Rigid multi-block translation failure mechanism under vertical load. ergy dissipation, the increased cohesion caused by suction is

84
L. Jin et al. Computers and Geotechnics 109 (2019) 82–88

rigid block. One set of variables for determining the critical slip surface
corresponds to a solution vector in IRMO, thus making IRMO relevant
to the determination of the critical slip surface. Therefore, when IRMO
is applied to calculate the ultimate bearing capacity of the foundation,
θ, αi, and δi can be used as independent variables to formulate a fun-
damental variable matrix Xi,j as shown in Eq. (22). In order to ensure
that the critical slip surface is searched within a reasonable range, the
independent variables θ, αi, δi are required to satisfy the velocity field
constraints and geometric constraints for rigid block motion coordina-
tion in the rigid multi-block failure algorithm and the constraint
boundary conditions for each variable are given in Eq. (23).
1 1
⎡ θ1,1 α1,2 δ1,3 ⋯ α1,nM − 1 δ1,nM ⎤
⎢θ 1 1
α 2,2 δ2,3 ⋯ α 2,n M − 1 δ2,n M ⎥
Xi, j = ⎢ 2,1 ⎥
Fig. 4. Different forms of matric suction distribution in Zhao et al. [27]. ⎢⋮ ⋮ ⋮ ⋱ ⋮ ⋮ ⎥

θ α1 δ1 ⋯ αNn , M − 1 δNn , M ⎥ (22)
⎣ N ,1 N ,2 N ,3 ⎦
considered. Fredlund and Rahardjo [6] proposed a double stress state
variable shear strength formula, as shown in Eq. (15). It can be seen 0 < θ < π /2 ⎫
n
from Eq. (16) that the total cohesion consists of two parts: the effective θ + ∑i = 1 αi = π ⎪
cohesion c' and the increased cohesion caused by the suction (ua-uw) ⎪
0 < αi < π − 2φ′ ⎪
tanφb. The total cohesion at the midpoint depth of the velocity dis- (i
0 < δi ⎬
continuity lines Li and Di (i = 1,2, … ,n) is regarded as the average total ⎪
α1 + 2θ − π < δ1 < 2θ − 2φ′
cohesion cLi , cDi (i = 1,2, … ,n) on the discontinuous line. In this study, ⎪
αi + αi − 1 + βi − 1 − π < δi < αi − 1 + βi − 1 − 2φ (i = 2, 3, ⋯, n) ⎪


three forms of matric suction distributions are considered: no matric
suction (Case 1), homogeneous variation of matric suction with depth = 1, 2, ⋯, n) (23)
(Case 2), and linearly decreasing matric suction until zero at the
groundwater level (Case 3), as shown in Fig. 4. The average cohesion Based on the above fundamental variable matrix Xi,j and boundary
cLi , cDi (i = 1,2, … ,n) under Case 3 are as shown in Eqs. (17) and (18), constraints, the IRMO for calculating the ultimate bearing capacity of a
respectively. shallow foundation can be realized on the MATLAB platform. In specific
engineering examples, when the basic parameters of IRMO, such as N,
τf = c′ + (ua − u w ) tan φb + (σ − ua )tanφ′ (15) M, G, and other relevant parameters for the calculation of the ultimate
bearing capacity are set, IRMO can be applied to search for the location
c = c′ + (ua − u w ) tan φb (16)
of the critical slip surface and to calculate the corresponding ultimate
Z bearing capacity under three cases.
cLi = c′ + (ua − u w )0 ⎛1 − Li ⎞ tan φb
⎜ ⎟

⎝ Hw ⎠ (17)
5. Case studies
Z
cDi = c′ + (ua − u w )0 ⎛1 − Di ⎞ tan φb
⎜ ⎟

⎝ H w⎠ (18) 5.1. Comparative analysis on ultimate bearing capacity for saturated soil

where Hw represents the depth of the water table; (ua − uw)0 represents In previous studies, extensive research has been conducted on the
the suction at the surface; ZLi, ZDi represent the depth at the midpoint of ultimate bearing capacity of foundation without considering matric
the velocity discontinuity Li, Di (i = 1,2, … ,n); φb indicates the internal suction based on different theories and methods [20,12,1,3,13,19]. In
friction angle associated with the suction. this study, the bearing capacity factors Nγ, Nq, and Nc of a strip footing
Considering the effects of different distributions of suction, the in- under Case 1 are calculated using IRMO and compared with those ob-
ternal energy dissipation are shown in Eqs. (19) and (20), respectively. tained from previous studies. The results of the comparison are pre-
According to the principle of virtual power, the expression of the ulti- sented in Tables 1 and 2. The basic parameters of IRMO are taken as
mate bearing capacity Pu can be derived as shown in Eq. (21). N = 50, M = 51, and G = 250.
n
xi − yi
WL = 2 ∑ cLi·Li ·Vi − 1, i·cos φ′ n
∑i = 1 yi
i=1 (19) er = × 100%
n (24)
n
WD = 2 ∑ cDi·Di ·Vi ·cos φ′ where er is the average relative error, xi and yi are the IRMO-calculated
i=1 (20) and that from other scholar’s bearing capacity factors, respectively.
WL + WD − Wsoil − Wq From Table 1, we can see that the value of Nγ calculated by IRMO is
Pu = close to the results obtained using other upper bound limit analysis
B0·V0 (21)
methods [19,24,13,3]. Moreover, compared to the results reported in
Chen [3], the average relative error er of the result obtained in this
4. Implementation of IRMO study is only 8.8%; further, the results are smaller than those from Chen
[3]. In addition, the results obtained herein are also similar to that
Based on the rigid multi-block failure mechanism established by calculated using the slip line theory, with the average relative error er
Zhao et al. [27], IRMO is used to further optimize the calculation of being only 2.8% compared to the results of Bolton and Lau [1]. The
ultimate bearing capacity of strip foundations on unsaturated soil. To results obtained using Terzaghi’s equations are relatively small, which
determine the critical slip surface, only three variables are required, may be owing to the shear strength of soil within the embedment depth
namely, the base angle θ of rigid block No. 0, vertex angle αi of the i-th of foundation being neglected. Table 2 shows that the values of Nc and
triangular rigid block, and angular increment δi of the base of the i-th Nq calculated by IRMO and those by other upper bound limit analysis
triangular rigid block relative to the base of the (i − 1)-th triangular methods [19,24] are nearly identical, and the average relative errors er

85
L. Jin et al. Computers and Geotechnics 109 (2019) 82–88

Table 1
Comparison of Nγ values calculated from different studies.
φ′ (°) Upper bound limit analysis theorem Limit equilibrium theory Slip line theory

This study Soubra [19] Wang et al. [24] Michalowski [13] Chen [3] Terzaghi [20] Meyerhof [12] Vesic [22] Caquot and Kerisel Bolton and Lau
method1 [2] [1]

20 5.81 4.49 8.01 4.47 5.87 5.00 2.87 5.39 4.97 5.97
25 11.74 9.81 14.12 9.97 12.40 9.70 6.77 10.88 10.40 11.60
30 24.68 21.51 25.97 21.4 26.70 19.70 15.67 22.40 21.80 23.60
35 53.22 49.00 53.60 48.7 60.20 42.40 37.15 48.03 48.00 51.00
40 119.51 119.84 129.66 119 147.00 100.40 93.69 109.41 113.00 121.00

Table 2
Comparison of Nc and Nq values calculated in different studies.
φ′ (°) Nc Nq

This study Terzaghi [20] Meyerhof [12] Soubra [19] Wang et al. [24] This study Terzaghi [20] Meyerhof [12] Soubra [19] Wang et al. [24]
method1 method1

20 15.13 17.3 18.00 14.86 16.13 6.53 7.4 6.4 6.47 5.81
25 21.29 – – 20.77 21.83 10.93 – – 10.69 10.74
30 31.24 36.4 30.00 30.24 31.70 19.06 22.5 18.4 18.46 18.75
35 49.10 – – 46.33 50.18 35.38 – – 33.44 35.66
40 93.98 91.2 100.00 75.77 85.56 78.47 81.3 64.1 64.58 70.62

Table 3
Basic Parameters of example of Zhao et al. [27].
Width B0 (m) Unit weight γ (kN/m3) Effective cohesion c′ (kPa) Effective friction angle φ′ (°) Suction angle φb (°) Depth of ground water table Hw Overload q (kPa)
(m)

0.5 18.3 5 20 15 4 9.15

which further validates the capability of the IRMO in calculating the


ultimate bearing capacity of shallow foundations on saturated soil.

5.2. Comparative analysis on ultimate bearing capacity for unsaturated soil

Based on the unified solution of shear strength for unsaturated soil


under plane strain conditions, the formulation of Terzaghi’s ultimate
bearing capacity for a strip footing under the linear distribution of
matrix suction was established by Zhang et al. [26]. Based on the Mohr-
Coulomb strength criterion, Zhao et al. [27] obtained a better upper
bound solution of the ultimate bearing capacity for a rough strip footing
on unsaturated soil under linear suction by using a sequential quadratic
programming (SQP) method. The basic parameters of example of Zhao
et al. [27] are shown in Table 3. In this study, the ultimate bearing
capacity of the rough strip foundation is calculated by the IRMO al-
gorithm under Case 3. The basic parameters of IRMO are also taken as
N = 50, M = 51, and G = 250. A comparison of these three sets of re-
sults for different matrix suction levels are shown in Fig. 5.
Fig. 5. Comparison of bearing capacity of strip foundation on unsaturated soil
Fig. 5 show that with an increase of the matric suction, the ultimate
calculated using IRMO and those from Zhao et al. [27] and Zhang et al. [26] bearing capacity of the rough strip footing calculated by the IRMO al-
under Case 3. gorithm is consistent with the results of Zhao et al. [27] and Zhang et al.
[26]. However, overall, the results from this study are relatively
smaller, thereby representing more conservative solution for en-
of Nc and Nq calculated by Wang et al. [24] and in this work are only
gineering practice. The increase in matrix suction contributes to the
4.4% and 5.5%, respectively; these results demonstrate the accuracy of
shear strength of the soil, resulting in an increase in the ultimate
IRMO and the analysis model in this study.
bearing capacity of strip footing. Nevertheless, this result further con-
In the work by Soubra [19], in which the symmetric rigid multi-
firms the correctness and effectiveness of the IRMO in calculating the
block failure mechanism is adopted, the basic parameters are as fol-
ultimate bearing capacity of shallow foundations on unsaturated soil.
lows: B0 = 1 m, γ = 18 kN/m3, c′ = 10 kPa, and q = 10 kPa. When
The stability of the IRMO algorithm for the search of critical slip
φ′ = 30°, the ultimate bearing capacity Pdirect under the static force
surface and the calculation of the corresponding ultimate bearing ca-
obtained by Soubra [19] is 726.13 kN/m2, whereas that calculated
pacity are verified in this section. When the basic calculation para-
value using IRMO is 726.01 kN/m2, which is an excellent agreement.
meters of foundation are taken as the following values: B0 = 1 m,
Thus, this value can be considered an optimized upper bound result,
γ = 18 kN/m3, c′ = 10 kPa, φ′ = 30°, q = 10 kPa, Hw = 4 m, φb = 15°,

86
L. Jin et al. Computers and Geotechnics 109 (2019) 82–88

Fig. 6. Search results for the critical slip surface using IRMO 20 times.

Fig. 7. Calculation results for the ultimate foundation bearing capacity using IRMO 20 times.

Fig. 8. Search results of the critical slip surface under different distributions of matric suction.

matrix suction at the surface (ua − uw)0 = 100 kPa under Case 3, the IRMO is subsequently used to optimize the search for the critical slip
IRMO algorithm was applied continuously for 20 times to search for the surface of a shallow foundation under three different distributions (see
critical slip surface and calculate the corresponding ultimate bearing Fig. 4) of matric suction and the results are shown in Fig. 8. It can be
capacities in order to investigate the convergence of the algorithm. The seen from Fig. 8 that compared with the search results under Case 1,
resulted critical slip surfaces and the corresponding ultimate bearing wider and deeper critical slip surfaces are noted when matrix suction is
capacities are shown in Figs. 6 and 7, respectively. Figs. 6 and 7 in- taken into account, resulting in a significant increase in the ultimate
dicate that the results of the 20 searches for the critical slip surface and bearing capacity. It is worth noting that the critical slip surface de-
ultimate bearing capacity under Case 3 are nearly identical. The range termined under Case 3 is wider and deeper than that under Case 2;
of the calculated 20 ultimate bearing capacity values is within 0.18% however, the corresponding ultimate bearing capacity under Case 3 is
around the averaged value. Figs. 6 and 7 thus demonstrate that IRMO slightly smaller than that under Case 2. This may be due to the total
can be applied to search the critical slip surface and calculate the cor- equivalent matric suction in Case 2 is higher than the one in Case 3,
responding ultimate bearing capacity accurately with efficient con- which may be one explanation why the ultimate bearing capacity in
vergence. Case 2 is higher than Case 3, since the matric suction can enhance the

87
L. Jin et al. Computers and Geotechnics 109 (2019) 82–88

bearing capacity. 1993.


[7] Gao W. Stability analysis of rock slope based on an abstraction ant colony clustering
algorithm. Environ Earth Sci 2015;73(12):7969–82.
6. Conclusion [8] Hansen JB. A revised and extended formula for bearing capacity. Bull Danish
Geotech Inst 1970;28:5–11.
In this study, the improved radial movement optimization (IRMO) [9] Jin LX, Feng QX. Improved radial movement optimization to determine the critical
failure surface for slope stability analysis. Environ Earth Sci 2018;77:564. https://
was used to investigate the critical slip surface and ultimate bearing doi.org/10.1007/s12665-018-7742-6.
capacity of rough strip foundations on unsaturated soil subjected to [10] Mahrami M, Rahmani R, Seyedmahmoudian M, Mashayekhi R, Karimi H, Hosseini
vertical load based on the upper-bound limit analysis theorem and rigid E. A hybrid metaheuritic technique developed for hourly load forecasting.
Complexity 2016;21(S1):521–32.
multi-block method. Based on the results of this study, the following [11] Meier J, Schaedler W, Borgatti L, Corsini A, Schanz T. Inverse parameter identifi-
conclusions are drawn: cation technique using pso algorithm applied to geotechnical modeling. J Artif Evol
Appl 2008;2008. https://doi.org/10.1155/2008/574613.
[12] Meyerhof GG. The ultimate bearing capacity of foundations. Géotechnique
(1) By considering the self-feedback of particles, the data structure of
1951;2(4):301–32.
RMO was improved, and its stability was enhanced. IRMO can be [13] Michalowski R. An estimate of the influence of soil weight on bearing capacity
applied to accurately and stably search the critical slip surface and using limit analysis. Soils Found 1997;37(4):57–64.
calculate the corresponding ultimate bearing capacity of shallow [14] Mohamed FMO, Vanapalli SK. Laboratory investigations for the measurement of the
bearing capacity of an unsaturated coarse-grained soil. Proceedings of the 59th
foundations on unsaturated soil. Therefore, IRMO could be applied Canadian geotechnical conference, Vancouver. 2006. p. 219–26.
to correctly estimate the stability of foundations and provide a re- [15] Oh WT, Vanapalli SK. Interpretation of the bearing capacity of unsaturated fine-
ference for engineering practice. grained soil using the modified effective and the modified total stress approaches.
Int J Geomech 2013;13(6):769–78.
(2) Different distributions of matric suction significantly influence cri- [16] Oloo SY, Fredlund DG, Gan JK-M. Bearing capacity of unpaved roads. Can Geotech
tical foundation slip surface and ultimate bearing capacity. For the J 1997;34(3):398–407.
same foundation model, compared with that in Case 1, the critical [17] Rahmani R, Yusof R. A new simple, fast and efficient algorithm for global optimi-
zation over continuous search-space problems. Appl Math Comput
slip surface determined considering the matric suction of un- 2014;248(1):287–300.
saturated soil is larger and deeper, and the corresponding ultimate [18] Seyedmahmoudian M, Horan B, Rahmani R, Oo AMT, Stojcevski A. Efficient pho-
bearing capacity is also significantly increased. Compared with the tovoltaic system maximum power point tracking using a new technique. Energies
2016;9(3):147.
ultimate bearing capacity calculated under Case 2, in which the [19] Soubra AH. Upper-bound solutions for bearing capacity of foundations. J Geotech
matric suction is evenly distributed over depth, the result obtained Geoenviron Eng 1999;125(1):59–68.
under Case 3 is relatively conservative but is more relevant to en- [20] Terzaghi K. Theoretical soil mechanics. New York (America): Wiley; 1943.
[21] Vanithasri M, Balamurugan R, Lakshminarasimman L. Modified radial movement
gineering practice.
optimization (MRMO) technique for estimating the parameters of fuel cost function
in thermal power plants. Eng Sci Technol Int J 2016;19(4):2035–42.
References [22] Vesic AS. Analysis of ultimate loads of shallow foundations. J Soil Mech Found Div
1973;99(1):45–73.
[23] Wan W, Cao P, Feng T, Yuan HP. Improved genetic algorithm freely searching for
[1] Bolton MD, Lau CK. Vertical bearing capacity factors for circular and strip footings dangerous slip surface of slope. J Centr South Univ Technol 2005;12(6):749–52.
on Mohr-Coulomb soil. Can Geotech J 1993;30(6):1024–33. [24] Wang YJ, Yin JH, Chen ZY. Calculation of bearing capacity of a strip footing using
[2] Caquot A, Kerisel J. Sur le terme de surface dans le calcul des foundations en milieu an upper bound method. Int J Num Anal Meth Geomech 2001;25(8):841–51.
pulverulent. Proceedings of the 3rd international conference on soil mechanics and [25] Wuttke F, Kafle B, Lins Y, Schanz T. Macroelement for statically loaded shallow
foundation engineering, Zürich, Switzerland, vol. 1. 1953. p. 336–7. strip foundation resting on unsaturated soil. Int J Geomech 2013;13(5):557–64.
[3] Chen WF. Limit analysis and soil plasticity. Amsterdam (Holland): Elsevier; 1975. [26] Zhang CG, Yan Q, Zhou F, Zhang CL. Formulation of ultimate bearing capacity for a
[4] Cheng YM, Li L, Sun YJ, Au SK. A coupled particle swarm and harmony search strip foundation resting on unsaturated soils with different suction distributions.
optimization algorithm for difficult geotechnical problems. Struct Multidiscip Chin J Rock Mech Eng 2018;37(05):1237–51. [in Chinese].
Optim 2012;45(4):489–501. [27] Zhao LH, Li L, Yang F, et al. Ultimate bearing capacity calcution of strip foundation
[5] Costa YD, Cintra JC, Zornberg JG. Influence of matric suction on the results of plate on unsaturated soil with upper bound theorem and SQP method. Chin J Rock Mech
load tests performed on a lateritic soil deposit. Geotech Test J 2003;26(2):219–26. Eng 2009;28(suppl. 1):3021–8. [in Chinese].
[6] Fredlund DG, Rahardjo H. Soil mechanics for unsaturated soils. NewYork: Wiley;

88

You might also like