Sunteți pe pagina 1din 9

Journal of Intelligent & Fuzzy Systems 29 (2015) 955–962 955

DOI:10.3233/IFS-151624
IOS Press

Soil compaction optimization with soft


constrain
Primoz Jelusic∗
Faculty of Civil Engineering, University of Maribor, Maribor, Slovenia

Abstract. The primary benefit of fuzzy systems theory is to approximate system behavior where analytic functions or numerical
relations do not exist. In soil mechanics relationship between the causes and effects can be observed with laboratory tests but is
difficult to develop analytical functions or numerical relations between input and output. In geotechnical optimization the most
usual constraints represent state variables of structural response for each loading case. The aim of this paper is to define the soft
constrain with adaptive network-based fuzzy inference system (ANFIS) in the soil mechanics. The developed soft constrain is
than applied in non-linear programming (NLP) to obtain optimal solution. In the case of soil compaction the performance of the
proposed optimization algorithm is evaluated. The main aim of soil compaction is to define optimal water content at which soil
can be compacted to a densest state that improve their mechanical and physical properties.

Keywords: Soil compaction, fuzzy relation equations, adaptive network-based fuzzy inference system, non-linear programming

1. Introduction max-average composition was proposed by Khorram


and Ghodousian [4]. Loetamonphong et al. [5] used
For many geotechnical projects, soils have to be com- the genetic algorithm and solved the multi-objective
pacted to a denser state to improve their mechanical and optimization problem with fuzzy relation equation con-
physical properties [1]. They are compacted with ram- straints with max-min composition. Multi-objective
ming, rolling and vibrating. The water content has the optimization problems with Fuzzy relation equation
most important effect on soil compaction. The dens- constraints regarding max-average composition was
est soil is obtained at the optimum water content. The improved by Khorram and Zarei [6]. Zhang et al.
laboratory compaction tests are used to determine the [7] proposed a hybrid multiobjective fireworks opti-
relation between water content and dry unit weight and mization algorithm (MOFOA) that evolves a set of
optimum water content. The upper limit of compaction solutions to the Pareto optimal front by mimicking the
is defined with saturation lines [2]. explosion of fireworks. The experimental tests and real-
The majority of current optimization schemes require world applications in oil crop production in east China
designers to define crisp-valued constrains even when demonstrate the effectiveness and practicality of the
in the presence of subjective “soft” constraints, thereby algorithm. Solving high dimensional bilevel multiob-
ignoring the uncertainty that is inherent during the jective programming problem using a hybrid particle
definition of early stage design constraints [3]. The swarm optimization (PSO) algorithm with crossover
optimization problem with a linear objective function operator was presented in [8, 9].
subjected to the fuzzy relational equations based on the Pasdarpour and et al. [10] described the incorporation
of genetic algorithm methodology using fuzzy system
∗ Corresponding author. Primoz Jelusic, Faculty of Civil
for determining the optimum design of dynamic com-
Engineering, University of Maribor, Maribor, Slovenia, Tel.: +386 paction. The results show that the genetic algorithm has
2 2294 331; Fax: +386 2 2524 179; E-mail: primoz.jelusic@um.si. abilities to optimize dynamic compaction design which

1064-1246/15/$35.00 © 2015 – IOS Press and the authors. All rights reserved
956 P. Jelusic / Soil compaction optimization with soft constrain

can increase the depth of improvement about 20–30% are filled by water, the soil is saturated. Otherwise, the
more than traditional design. Soil compaction is also an soil is unsaturated. If all voids are filled by gas, the soil is
important component of the land degradation syndrome fully dry. Commonly known idealization of three phase
which is an issue for soil management throughout the soil is shown in Fig. 1. The physical and mechanical
world [11]. Qin et al. [12] showed with application and properties of soil are dependent on the relative propor-
evaluation that fuzzy slope positions effectively predict tions of each three phases.
the spatial distribution of soil properties and can provide
useful information for digital soil mapping. 2.1. Soil phase relationships
The Adaptive network-based fuzzy inference system
(ANFIS) is considered to be one of the intelligent tools The total volume of the soil is the sum of the volume
to understand complex problems [13]. ANFIS is being of solids (Vs ) and volume of the voids (Vv ), that are
successfully used in many industrial areas as well as in volume of water (Vw ) and volume of the air (Va ).
research [14–20].
In this paper two optimization models are presented. V = Vs + Vv = Vs + Vw + Va . (1)
The objective function is obtained from laboratory com- The weight of the soil is the sum of the weight of
paction test and constrain function is defined with soil solids (Ws ) and the weight of water (Ww ). The weight
phase relationship. First the optimization model with a of air is negligible.
nonlinear objective function and crisp-valued constrain
is developed. Next the optimization model with fuzzy W = Ws + Ww (2)
relation equation constrain is presented. The ANFIS
was used to develop soft constrain function. The com- In the interpretation of compaction terms, given
parison of optimization models is possible because both below, are used for soil phase relationships. Water con-
crisp and soft constrain are based on soil phase rela- tent (w) is the ratio of the weight of water (Ww ) to the
tionship. Therefore the performance of the proposed weight of solids (Ws ).
optimization algorithm is evaluated. The present article Ww
aims: (1) to analyze the optimization model subjected w= · 100%. (3)
Ws
to crisp and soft constraints, (2) to evaluate the perfor-
mance of the proposed algorithm for selected laboratory Void ratio (e) is the ratio of the volume of void space
compaction test and (3) to present the benefit of ANFIS (Vv ) to the volume of solids (Vs ).
where analytic functions do not exist, but relationship Vv
between the causes and effects can be observed with e= . (4)
Vs
laboratory tests
Specific gravity (Gs ) is the ratio of the weight of the
solids to the weight of water of equal volume
2. Physical soil properties Ws
Gs = . (5)
Vs · γ w
Soil is three phase material, composed of solid, liq-
The dry unit weight (γd ), specific gravity (Gs ), water
uid and gas. The solid phase is creating from particles,
content (w), degree of saturation (Sr ) and water unit
which are different size, shape and from different min-
weight (γw ) are related through the analytical equation
erals. The spaces between soil particles are called voids,
mostly filled by water (fluid) and air (gas). If all voids (1 + wGs ) · γd
Sr = . (6)
Gs · γ w
When Sr , γw and Gs are given the saturation lines
are defined.

2.2. Soil compaction

Soil compaction is the densification of soil by reduc-


tion of void ratio. It is the expulsion of air, if water
Fig. 1. Soil phases and principle of soil compaction. content remains constant. In saturated soil densification
P. Jelusic / Soil compaction optimization with soft constrain 957

The results are presented in the form of a compaction


curve: dry unit weight (γd ) versus water content (w).
This curve is obtained using data points for each com-
pacted sample and connecting these points by smooth
curve. The curve is expressed as third order polyno-
mial function. The saturation lines are also plotted next
to the compaction curve. Usually all the experimen-
tal data points fall under the 100% saturation line.
Lower saturation lines define limits of possible max-
imum dry unit weight at select degrees of saturation
(γd,max (Sr )).

Fig. 2. Theoretical maximum dry unit weight vs. water content for
different degrees of saturation.
3. Optimization model

The comparison of optimization models is presented.


is possible only by reduction of water content. Max- The first optimization model is subjected to crisp con-
imum dry unit weight (γd,max ) is the maximum unit strains and the second optimization model is subjected
weight that a soil can attain using a specified means of to soft constrains. As an interface for mathemati-
compaction. cal modeling and data inputs/outputs GAMS (General
Using (6) dry unit weight of soil is given as a function Algebraic Modeling System), a high level language
of water content (w) and degree of saturation (Sr ). The [22], was used.
theoretical maximum dry unit weight is obtained when
Sr = 1. The theoretical maximum dry unit weight (γd ) 3.1. Optimization model with crisp constrains
decreases with reducing of degree of saturation (Sr ).
Figure 2 shows relationships maximum dry unit weight As the problem of the soil compaction is non-linear,
(γd ) and water content (w) for different degrees of sat- e.g. the objective function and (in)equality constraints
uration (Sr ). Optimum water content (wopt ) is the water are non-linear, the non-linear programming (NLP)
content allowing a soil to attain its maximum dry unit optimization approach is applied. The general NLP
weight following a specified means of compaction. optimization problem can be formulated as follows:

2.3. Proctor compaction test Maxz = f (x)


subjected to: (NLP)
A laboratory test called Proctor test was developed
to determine the maximum dry unit weight (γd,max ) and g(x) ≤ 0
corresponding optimum water content (wopt ). Using h(x) = 0
standard ASTM D1557 [21] a dry soil specimen is  
mixed with water and compacted in a cylindrical mold, x ∈ X = x| x ∈ Rn , xLo ≤ x ≤ xUp
internal dimension 101.6 mm in diameter and 116.4 mm where x is a vector of the continuous variables defined
high. Specimen is compacted by repeated blows from within the compact set X. Functions f (x), g(x) and h(x)
the mass of hammer, 2.5 kg, falling freely from 305 mm. are nonlinear functions involved in the objective func-
The soil is compacted in three layers, each of which tion z, inequality and equality constraints, respectively.
is subjected to 25 blows. For projects involving heavy All functions f (x), g(x) and h(x) must be continuous
loads a modified proctor test was developed. In this and differentiable. In the context of optimal water con-
test, the hammer has a mass 4.54 kg and falls freely tent and maximum dry unit weight determination, the
from a height of 457 mm. Four or more tests need to objective function is determined. The objective function
be performed on the soil using different water content. is expressed as third order polynomial function obtained
The last test is identified when additional water causes from compaction test:
the unit weight of the soil is decrease. The results are
expressed as dry unit weight (γd ) versus water content γd = 0.0014w3 − 0.1003w2 − 2.166w + 2.3961.
(w). Figure 2 show an example of a compaction test. (7)
958 P. Jelusic / Soil compaction optimization with soft constrain

Table 1
Dataset for ANFIS-COMP model
Input Output
Water Degree of Specific Dry unit weight
content w (-) saturation Sr (-) gravity Gs (-) ␥d (kN/m3 )
0.10 1.00 2.7 20.85
0.15 1.00 2.7 18.85
0.20 1.00 2.7 17.19
0.25 1.00 2.7 15.81
0.30 1.00 2.7 14.63
0.10 0.95 2.7 19.81
0.15 0.95 2.7 17.90
0.20 0.95 2.7 16.33
0.25 0.95 2.7 15.02
Fig. 3. Example of compaction curve with 100% (Sr = 1.0), 95%
0.30 0.95 2.7 13.90
(Sr = 0.95) and 90% (Sr = 0.90) saturation lines.
0.10 0.90 2.7 18.76
0.15 0.90 2.7 16.96
The constrain function is defined with degree of sat- 0.20 0.90 2.7 15.47
uration: 0.25 0.90 2.7 14.23
0.30 0.90 2.7 13.17
Sr · Gs γw
γd ≤ . (8) 0.10 1.00 2.6 20.24
1 + w · Gs 0.15 1.00 2.6 18.34
0.20 1.00 2.6 16.78
With NLP optimization the optimum water content 0.25 1.00 2.6 15.45
(wopt ) and maximum dry unit weight (γd,max ) are 0.30 1.00 2.6 14.32
0.10 0.95 2.6 19.22
obtained for various degree of saturation (Sr ). The
0.15 0.95 2.6 17.43
graphical presentation of optimization problem is given 0.20 0.95 2.6 15.94
in Fig. 3. 0.25 0.95 2.6 14.68
0.30 0.95 2.6 13.61
0.10 0.90 2.6 18.21
3.2. Optimization model with soft constrains 0.15 0.90 2.6 16.51
0.20 0.90 2.6 15.10
In order to predict the dry unit weight, the ANFIS- 0.25 0.90 2.6 13.91
COMP model was build, based on the data set presented 0.30 0.90 2.6 12.89
0.10 1.00 2.5 19.61
in Table 1. The inputs of a model are water content (w), 0.15 1.00 2.5 17.83
degree of saturation (Sr ) and specific gravity (Gs ). The 0.20 1.00 2.5 16.35
output is the dry unit weight (γd ). The data presented 0.25 1.00 2.5 15.09
in Table 1 are obtained with (6) for water unit weight 0.30 1.00 2.5 14.01
0.10 0.95 2.5 18.63
γw = 9.801 kN/m3 . In this paper 45 data sets were used 0.15 0.95 2.5 16.94
to develop ANFIS-COMP model with fuzzy toolbox 0.20 0.95 2.5 15.53
[18]. The developed model is then applied in NLP opti- 0.25 0.95 2.5 14.33
0.30 0.95 2.5 13.31
mization approach as a constrain function and present
0.10 0.90 2.5 17.65
the saturation lines. 0.15 0.90 2.5 16.05
The structure of the ANFIS-COMP model is shown 0.20 0.90 2.5 14.71
in Fig. 4. While the nodes on the left side represent the 0.25 0.90 2.5 13.58
0.30 0.90 2.5 12.61
input data, the right node stands for the output. In a con-
ventional fuzzy inference system, a number of rules is
decided by an researcher/engineer who is familiar with
the system to be modeled. There are no simple ways to The calculation procedure of ANFIS-COMP model
determine in advance the minimal number of member- is as follows: first the membership grade of a fuzzy set is
ship functions to achieve a desired performance level. In calculated; second the product of membership function
this attempt, the number of membership functions as- for each rule is calculated; third the ratio between the
signed to each input variable was chosen empirically i-th rule’s firing strength and the sum of all rule’s firing
by examining the desired input-output data and by trial strengths is calculated; next the output of each rule is
and error. For model ANFIS-COMP the, we choose two calculated; and final the weighted average of each rule’s
membership functions in each input (Fig. 5). output is calculated.
P. Jelusic / Soil compaction optimization with soft constrain 959

Table 2
Permise and consequent parameters of ANFIS-COMP model
Permise parameters Consequent parameters
cA,1 0.2080 a0 1 −2.9448
σ A,1 0.0819 a1 1 −21.4099
cA,2 0.0934 a2 1 14.6978
σ A,2 0.0767 a3 1 3.4736
cB,1 0.9659 a0 2 2.9893
σ B,1 0.0581 a1 2 −35.7980
cB,2 0.9867 a2 2 19.5293
σ B,2 0.0581 a3 2 2.9893
cC,1 2.6068 – –
σ C,1 0.0868 – –
Fig. 4. Fuzzy Inference System implemented in the framework of cC,2 2.6929 – –
adaptive networks ANFIS-COMP. σ C,2 0.1029 – –

where ui represents the firing strength of the i-th rule.


The weighted average of each rule’s output is defined
as a ratio between the i-th rule’s firing strength and the
sum of all rule’s firing strengths, see (14) and (15):
u1
u1 = (14)
u1 + u 2

u2
u2 = (15)
u1 + u 2
The output of each rule was finally determined as a
Fig. 5. Membership function of ANFIS-COMP model. sum of products between the weighted average of each
rule’s output and a linear combination between input
The first membership grade of a fuzzy set variables and consequence parameters:
(Ai , Bi , Ci ) was calculated with (9), (10) and (11):

2


⎡  2 ⎤ γd = ui · a0i + a1i · w + a2i · Sr + a3i · Gs (16)
w − c
µAi (w) = exp ⎣− √ ⎦
A i i=1
(9)
2σAi Once the dry unit weight is defined the con-
strain function could be applied in NLP optimization
⎡  2 ⎤ approach, see (17).
S − c
µBi (Sr ) = exp ⎣− √ ⎦
r Bi
(10)
2


2σBi γd ≤ ui · a0i + a1i · w + a2i · Sr + a3i · Gs (17)
i=1
⎡  2 ⎤
G − c
µCi (Gs ) = exp ⎣− ⎦ 4. Comparison of optimization models
Bi

s
(11)
2σBi
The result of NLP optimization with crisp and soft
Where w, Sr and Gs are inputs to Gaussian member- constrain is given in Table 3. In the optimization with
ship functions and parameters cAi , cBi σAi , σBi , σCi , crisp constrain the Equation (8) was used. In optimiza-
σCi are premise parameters (Table 2). In addition, the tion with soft constrain the (17) was used. In both cases
products between membership functions for every rule the objective function is same and is defined in (7). With
were calculated, see (12) and (13): both optimization approaches the optimal water content
w (%) and maximum dry unit weight γd,max (kN/m3 ) are
u1 = µA1 (w) · µB1 (Sr ) · µC1 (Gs ) (12) obtained. The maximum error of optimization model
with soft constrain depends on the accuracy of the
u1 = µA1 (w) · µB1 (Sr ) · µC1 (Gs ) (13) ANFIS-COMP model. In the case of soil compaction
960 P. Jelusic / Soil compaction optimization with soft constrain

Table 3
Comparison of optimization model with crisp and soft constrain
Sr Gs Crisp constrain ANFIS constrain Error wopt (%) Error ␥max (%)
wopt ␥max wopt ␥max
1.00 2.7 16.4976 17.1224 16.4976 17.1224 0.00 0.00
0.95 2.7 16.4976 17.1224 16.4976 17.1224 0.00 0.00
0.90 2.7 14.8020 17.0264 14.8639 17.0335 0.42 0.04
1.00 2.6 16.4976 17.1224 16.4976 17.1224 0.00 0.00
0.95 2.6 15.9779 17.1138 15.8262 17.1079 0.95 0.03
0.90 2.6 13.8376 16.8766 13.7977 16.8688 0.29 0.05
1.00 2.5 16.4976 17.1224 16.4976 17.1224 0.00 0.00
0.95 2.5 14.7420 17.0192 14.6605 17.0090 0.55 0.06
0.90 2.5 12.9492 16.6694 12.9486 16.6692 0.01 0.00

problem the maximum error is 0.95%. Prediction accu- 20.0


Soil type: Clayey sand (SC)
racy of ANFIS model can be expected to improve as Dense soil
more data sets are utilized. Higher capacity of ANFIS 19.5
model is important to obtain the exact optimal solution.
Dry unit weight (kN/m3)

19.0

5. Laboratory soil testing 18.5


METHOD B: ANFIS

In the practice of geotechnical engineering, soil test- Experiment


18.0
ing addresses real problems. The experiment data for Optimum METHOD A
three different soil types are presented in the form of Optimum METHOD B
17.5
a compaction curve: dry unit weigh γd,max (kN/m3 )
METHOD A: cubic
versus water content w (%). According to ASTM stan- polynomial regression
dard [21], for each soil type five test points should be 17.0
5.0 7.0 9.0 11.0 13.0 15.0 17.0 19.0
obtained with standard compaction test. To obtain five Water content (%)
well-placed points on the compaction curve, the water
content is selected low for the first test point and is grad- Fig. 7. Example of compaction curve for Clayey sand (SC).
ually increased for the other points. It should be about
4 to 5% below the optimum water content for the first
20.0
point, and 4 to 5% above the optimum water content for Soil type: Low plasticity clay
(CL)
Consistency stiff
19.5
Dry unit weight (kN/m3)

19.5
Soil type: Gravelly silt (MG) 19.0
Firm consistency (hard to deform by hand squeezing)

19.0 18.5 METHOD B: ANFIS

Experiment
Dry unit weight (kN/m3)

18.0 Optimum METHOD A


18.5
Optimum METHOD B
17.5
METHOD B: ANFIS METHOD A: cubic
18.0 polynomial regression
Experiment 17.0
Optimum METHOD A 5.0 7.0 9.0 11.0 13.0 15.0 17.0 19.0
17.5 Water content (%)
Optimum METHOD B

METHOD A: cubic Fig. 8. Example of compaction curve for Low plasticity clay (CL).
polynomial regression
17.0
5.0 7.0 9.0 11.0 13.0 15.0 17.0
Water content (%) the fifth and last point. Figure 6 show an example of a
compaction test for Gravelly silt (MG). The maximum
Fig. 6. Example of compaction curve for Gravelly silt (MG). unit weight and optimum water content are determined
P. Jelusic / Soil compaction optimization with soft constrain 961

by using two different methods referred to as method [3] A.W. Gray, A.S. Daniels and D.J. Singer, Impacts of fuzzy
A and method B. Method A uses a conventional cubic logic modeling for constraints optimization, Naval Engineers
Journal 122(2) (2010), 121–132.
polynomial regression fit to an experimental data set [1]. [4] E. Khorram and A. Ghodousian, Linear objective function opti-
The optimum of method A is plotted as a solid circle. mization with fuzzy relation equation constraints re-garding
Method B uses ANFIS to obtain the best-fit curve to the max-av composition, Applied Mathematics and Computation
experimental data. The optimum of method B is plot- 173 (2006), 872–886.
[5] J. Loetamonphong, S.C. Fang and R.E. Young, Mul-
ted as a solid triangle in Fig. 6. In this example, method ti-objective optimization problems with fuzzy relation
B gives better optimum water content than method A. equation constraints, Fuzzy Sets and Systems 127 (2002),
Figures 7 and 8 shows an example of a compaction test 141–164.
for Clayey sand (SC) and Low plasticity clay (CL). [6] E. Khorram and H. Zarei, Multi-objective optimization prob-
lems with Fuzzy relation equation constraints re-garding
max-average composition, Mathematical and Computer Mod-
elling 49 (2009), 856–867.
6. Conclusion [7] Y.J. Zhang, Q. Song and S.Y. Chan, Multiobjective fire-
works optimization for variable-rate fertilization in oil
crop production, Applied Soft Computing 13(11) (2013),
The most important thing in the soil mechanics are 4253–4263.
laboratory test. With laboratory tests the relationship [8] T. Zhang, T. Hu, X. Guo, Z. Chen and Y. Zheng, Solving
between the causes and effects can be observed. Usu- high dimensional bilevel multiobjective programming prob-
lem using a hybrid particle swarm optimization algorithm
ally the results of laboratory test are difficult to describe with crossover operator, Knowledge-Based Systems 53 (2013),
with analytic functions. In laboratory test the inputs 13–19.
are also vague and ambiguous. The primary benefit of [9] M. Fontan, A. Ndiaye, D. Breysse and C. Fernandez, Soil-
fuzzy systems theory is to approximate system behav- structure interaction: Parameters identification using particle
swarm optimization, Computers and Structures 89 (2011),
ior where analytic functions or numerical relations do 1602–1614.
not exist. Since the solution set of the fuzzy relation [10] M. Pasdarpour, M. Ghazavi, M. Teshnehlab and S.A. Sadrne-
equations is in general a non-convex set, when it is not jad, Optimal design of soil dynamic compaction using genetic
empty, conventional nonlinear programming methods algorithm and fuzzy system, Soil Dynamics and Earthquake
Engineering 29 (2009), 1103–1112.
are not ideal for solving such a problem. In proposed [11] T. Batey, Soil compaction and soil management-a review, Soil
optimization algorithm the ANFIS model is used to Use and Management 25 (2009), 335–345.
develop soft constrain. The model contains symmetric [12] C.Z. Qin, A.X. Zhu, W.L. Qiu, Y.J. Lu, B.L. Li and T. Pei,
Mapping soil organic matter in small low-relief catchments
Gaussian membership functions which are continuous
using fuzzy slope position information, Geoderma 171-172
and differentiable. To evaluate the performance of the (2012), 64–74.
optimization with soft constrain the ANFIS model is [13] J.S.R. Jang, ANFIS: Adaptive-network-based fuzzy infer-
based on the soil phase relationship and results are ence system, IEEE Trans Syst, Man Cybernet 23(3) (1993),
665–685.
than compared with current optimization schemes with [14] S. Akkoç, An empirical comparison of conventional tech-
crisp-valued constrains. The comparison is made on niques, neural networks and the three stage hybrid Adaptive
selected laboratory compaction test. Since the devia- Neuro Fuzzy Inference System (ANFIS) model for credit
tion between the optimization models is very small, scoring analysis: The case of Turkish credit card data,
European Journal of Operational Research 222(1) (2012),
less than 1%, the optimization with soft constrain was 168–178.
successfully applied in the case of soil compaction. The [15] E. Shekarian and A.A. Gholizadeh, Application of adap-
fuzzy system is very useful in soil mechanic, because tive network based fuzzy inference system method in
the numerical relations based on experimental measure- economic welfare, Knowledge-Based Systems 39 (2013),
151–158.
ments could be effectively developed. Those numerical [16] L. Faravelli and T. Yao, Use of adaptive networks in fuzzy
relations are than used in the optimization approach as control of civil structures, Comput, Aided Civ Infrastruct Eng
constrain function. 11(1) (1996), 67–76.
[17] C. Kayadelen, O. Gunaydin, M. Fener, A. Demir and A.
Ozvan, Modeling of the angle of shearing resistance of soils
using soft computing systems, Expert Syst Appl 36 (2009),
11814–11826.
References [18] B. Zlender, P. Jelusic and D. Boumezerane, Planning geotech-
nical investigation using ANFIS, Geotech Geol Eng 30 (2012),
[1] J.P. Bardet, “Experimental Soil Mechanics,” 1st Edition, Pren- 975–989.
tice Hall, New Jersey, 1997. [19] B. Zlender, P. Jelusic and D. Boumezerane, The feasibility
[2] M.D. Braja, “Advanced Soil Mechanics,” 4th Edition, Taylor analysis of underground gas storage caverns, Eng Struct 55
& Francis Group, London, 2014. (2013), 16–25.
962 P. Jelusic / Soil compaction optimization with soft constrain

[20] T.J. Ross, Fuzzy Logic with Engineering Applications, 2nd [22] A. Brooke, D. Kendrick and A. Meeraus, GAMS—a user’s
edn, Wiley, Chichester, 2014, pp. 223–228. guide, Redwood City, CA, Scientific Press, 1988.
[21] ASTM Standard D1557, Standard Test Methods for Labora- [23] MATLAB version 7.10.0., The MathWorks Inc., Natick, MA,
tory Compaction Characteristics of Soil Using Modified Effort, 2010
ASTM International, West Con-shohocken, PA, 2009, DOI:
10.1520/D1557-09
Copyright of Journal of Intelligent & Fuzzy Systems is the property of IOS Press and its
content may not be copied or emailed to multiple sites or posted to a listserv without the
copyright holder's express written permission. However, users may print, download, or email
articles for individual use.

S-ar putea să vă placă și