Chapter 1

INTRODUCTION
1.1

General

The earthquake occurred in Sikkim (18 September 2011) of magnitude 6.9 was felt up to
Allahabad, Delhi, Jaipur and North Central Region (NCR). Earthquake occurring in different
countries, conjointly cause severe damages. An earthquake of 8.9 on Richter scale struck
Indonesia in 2012 whose tremors were felt in Kolkata, Chennai, Bangalore, Patna, Kochi,
Thiruvananthapuram, Cuttack, Bhubaneswar and several other cities on the eastern coast of
India. Metrological department of India issued the tidal wave warning for coastal regions of
Tamilnadu, Andhra Pradesh, Andaman and Nicobar Islands. The quake was felt as far away as
Singapore, Thailand and India. Therefore earthquake occurring in higher seismic zones could
end in damage as a result of liquefaction effects in lower seismic zones also.

Structures are laid on soil / rock is a well-known fact and most people rarely study the
concerning changes in engineering behavior of soil / rock and its sequent effect on structures in
an earthquake condition. Different soil properties will encounter different effect on seismic
waves as they pass through a soil layer. In some areas, there could also be many varieties of soils
bedded one upon the other as nonhomogeneous material before hard rock is encountered.
Sometimes, ground shaking is amplified which influences the preventive measures undertaken to
shield the structures in an earthquake condition. Hence, it's doable to shield the structures from
ground shaking up to certain magnitude of earthquake. However it's nearly not possible to shield
the structures undergoing differential settlement as a result of a phenomenon known as
liquefaction which occur in moderate to major earthquakes especially in weak soil like loose

sand. Liquefaction is not likely to occur in the soil possessing higher strength (for example hard
soil that may support greater than 3.72 tons/m2 load) doesnt seem to be likely for liquefaction
[1].

Due to buildup of dynamic pore water by sufficient shaking time of an earthquake having
sufficient magnitude soil loses its strength and it behaves like a liquid. This phenomenon is
termed as liquefaction. To assess the liquefaction for a zone or region is always an important task
for engineers and researchers to safeguard the structures from this phenomenon.

Fig. (1.1): Building damages due to liquefaction during the 1964 Niigata earthquake.
1.2

Purpose of the Study

Liquefaction is one of the most destructive hazards due to an earthquake. This can be as a result
of loose sand features a tendency to compress once a load is applied on the other hand dense

sands tends to expand in volume or 'dilate' when exposed to high water content. If the water (a
condition that usually exists particularly when soil is below the ground water table or sea level)
saturates the soil then water fills the gaps between soil grains ('pore spaces'). In response to the
soil compression, the pressure of this pore water will increase and makes an attempt to flow out
of the soil from the zones of low pressure (usually upward towards the ground surface).
However, if loading is dynamic in nature and high in magnitude (for example earthquake
shaking and storm wave loading etc.) such pore water doesn't flow out in time before the next
cycle of load is applied. The water pressures could build to that extent where it exceeds the
contact stresses between the grains of soil which keep them in contact with each other. These
contacts between grains are the means by which the weight from buildings and overlying soil
layers are transferred from the ground surface to layers of soil or rock at greater depths.
Therefore, pore water pressure decreases the shear strength of soil. This loss of strength (the
ability to transfer shear stress) of soil causes it to flow like a liquid (hence referred to as
'liquefaction').

Although, the seismicity status of Allahabad is low and it is placed in the zone-II as per Indian
standard code IS-1893 (2002) [2] but it is pertinent to know that status of liquefaction potential
of Allahabad. It has been seen in past that liquefaction had also occured in area of low seismicity
viz. Latur (Magnitude-6.2), Jabalpur (Magnitude-6.0) etc. Reason of Jabalpur earthquake was the
presence of Narmada fault whereas Latur (Maharashtra) earthquake happened due to release of
pressure along fault lines. Before this earthquake, Latur was in zone-I but, now it is in zone-III
[3]. Allahabad city is also situated on the bank of river Ganga and Yamuna and these river beds
are situated near the fault plane. This study will lead to assess and analyse of liquefaction
potential to densify the liquefaction prone zone in Allahabad city. The stream of Ganga and
3

Yamuna carries alluvial soil consist of sand and silt having varied structures. As indicated earlier
that loose sandy soil is prone to liquefaction, therefore it is necessary to study the occurrence and
non-occurrence of liquefaction in these areas. As liquefaction has occurred in varied earthquakes
and has left its mark within the geologic and historical record.

Evidence of past liquefaction, termed as paleo-liquefaction, has been used to evaluate seismic
hazards in areas where instrumental and historical data were sparse. The topic of liquefaction
came to the forefront of geotechnical earthquake engineering with the 1964 earthquakes that
occurred in Niigata, Japan and Alaska. In Niigata, liquefaction caused lateral spreading
pertaining to loss of bearing capability (Seed, 1979) [4], whereas strong earthquakes in other part
of the world like Loma Prieta (1989), Northridge (1994), California and Japan (1995), Turkey
(1999) and Taiwan (1999) have provided additional evidence of the damaging effects of
liquefaction.

1.3 Objectives and Scope of Study

1.3.1 Objectives:

To fulfill the purpose of research work, the following objectives of the present study has been
achieved:

i.

Geotechnical data collection, geotechnical investigation and generation of missing

geotechnical data of Allahabad city.

ii.

Evaluation of liquefaction potential and parametric study by conventional methods.

iii.

Development of liquefaction potential models for Allahabad city using artificial neural
network, adaptive neuro fuzzy inference system and multi linear regression techniques.
4

iv.

Comparative and critical analysis of conventional and computational approach

considering various statistical parameters.

v.

Evaluation of liquefaction prone areas to avoid the future destruction by considering

suitable anti-liquefaction measures in such areas.

1.3.2. Scope of Work

The determination of liquefaction potential in the Allahabad city is based upon in situ standard
penetration test (N-value) and subsurface condition. Subsurface data were obtained from State
and central government agencies firms and by the field investigation that was performed as a part
of this research work.

Identification of the regions where soil present with >15% clayey particles and the seismicity
potential after reviewing some empirical formulae. In this research work the best empirical
formulae should be validated which can help in the development of soft computing models.
Geographically, the study area was limited to the more urbanized regions where infrastructure is
growing rapidly.

1.4

Outline of the study:

Chapter 1: provides an overview and background of this study and describes the importance of
research work for the Allahabad region. The major argument made in this research work is the
The aim and objectives of the study are set out.
The theoretical frameworks are outlined.
Some of the key concepts used in this study are defined. and
The foundation for the study is presented.
5

Chapter 2: presents the work done in the area of the liquefaction potential with suitable research
gap w.r.t. national & international aspects. This chapter also describes the basic concept about
liquefaction potential behaviour.
Chapter 3: outlines the geotechnical data collection, investigation with geological features in
Allahabad city and its surroundings.
Chapter 4: outlines and justifies the use of methodological frameworks in this study. These
frameworks are considered highly compatible. The various methods used in the participatory
evaluation are described along with identification and assessment process of liquefaction
potential. The theoretical and methodological assumptions are discussed as well as the processes
used to undertake and validate. The strengths and limitations of each methodological framework
are discussed and the various validation processes are outlined.
Chapter 5: presents results and discussions as well as comparative studies based on statistical
parameters for the predictive.
Chapter 6: highlights the conclusions and scope of future work in this area.
-------

Chapter 2

LITERATURE REVIEW
2.1 General

In soil mechanics the term "liquefied" was first used by Hazen in reference to the failure of the
Calaveras Dam in California in the year 1918[5]. He had described the mechanism of flow
liquefaction of the embankment dam as follows:
If the pressure of the water in the pores is great enough to carry all the load, it will have the
effect of holding the particles apart and of producing a condition that is practically equivalent to
that of quicksand the initial movement of some part of the material might result in accumulating
pressure, first on one point, and then on another, successively, as the early points of
concentration were liquefied.
The ground motion during earthquake influences the properties of soil. Therefore, under dynamic
load, soil exhibits plastic deformation due to combination of permanent slip of soil particles
relative to one another. In any type of soil, when shearing stress induced, the soil grains tend to
rearrange into a more dense packing that results decrease in volume and increase in pore water
pressure which create less space in voids and water in pore spaces comes out with forced. If
drain of pore water is obstructed, pore water pressures increase progressively with the shear load.
This leads to transfer of stress from the soil skeleton to the pore water precipitating a decrease in
effective stress and shear resistance of soil. If the shear resistance of the soil becomes less than
the driving shear stress, the soil will undergo large deformations and is said to liquefy [6].
The bearing capability of soil to sustain foundation load is directly associated with strength,
liquefaction poses a serious threat to structures and should be assessed in urban areas where

loose soil deposit, liquefaction prone exist. Eight varieties of failure unremarkably associated
with soil liquefaction in earthquakes are [7]:
Buoyant rise of buried structures such as tanks.
Failure of retaining walls because of increased lateral loads from liquefied backfill soil or
loss of support from liquefied foundation soils.
Flow failures of slopes involving very large down-slope movements of a soil mass.
Ground oscillation where liquefaction of a soil deposit at a lower place level site leads to
back and forth movements of intact blocks of surface soil.
Ground settlement, usually related to another failure mechanism.
Lateral spreads ensuing from the lateral displacement of gently sloping ground.
Loss of bearing capability inflicting foundations failures.
Sand boils, occur when water under pressure wells up through a bed of sand and
comparatively minor damage.
2.2 Definition

When there is ground water less than 30 feet from the surface in soils that contain layers of sand,
the pressures generated by vibrations of the earth by several seconds of seismic wave that causes
the ground water to flow up and down. Once this happen, the sand grains that have no strength
except when they are in contact to. The soil turns up to the properties of a semi-solid state. If this
happens over a large area, houses and buildings with inadequate foundations may actually have
differential settlement. Whereas, if it happens in a small area liquefied sand may be ejected to the
surface through fissures within the overlying layers. These soil failures have a larger impact on
pipelines, pile foundations and other subsurface structures.

Soil liquefaction describes a phenomenon, when saturated or partly saturated soil considerably
loses their strength and stiffness because of an applied stress during earthquake shaking or
sudden change in stress condition, inflicting it to behave like a liquid and might be expressed as:

(2.1)

(2.2)
Where, = Shear strength of soil

= Effective vertical stress

= Total vertical stress

u = Pore water pressure

= Internal angle friction of soil.

Following factors generally influences the liquefaction phenomenon during earthquakes.

Grain size of the soil

Duration of the earthquake, amplitude and frequency of shaking

Distance from the epicenter

Location of the water table

Cohesiveness of the soil

Permeability of the layer

2.3 Factors affecting Soil Liquefaction

The susceptibility of soil deposits to liquefaction is determined by a combination of various

factors to which they may be subjected, such as:

i.

Ground motion and surrounding faults

ii.

Soil properties

iii.

Geological conditions

2.3.1 Soil properties

2.3.1.1 Soil Type

The type of soil that are most susceptible to liquefaction had been described by Ishihara (1985)
[8] which state: "The hazard associated with soil liquefaction during earthquakes has been
known to be encountered in deposits consisting of fine to medium sand and sands containing
low-plasticity fines. Occasionally, however, cases are reported where liquefaction apparently
occurred in gravelly soils."

Thus, the soil types susceptible to liquefaction are in nonplastic (cohesionless) soils. An
approximate listing of cohesionless soils from least to most resistant to liquefaction is clean
sands, nonplastic silty sands, nonplastic silt, and gravels. There could be numerous exceptions to
this sequence. For example, Ishihara (1985, 1993) [8, 9] describes the case of tailings derived
from the mining industry that were essentially composed of ground-up rocks and were classified
as rock flour. Further, the rock flour in a water-saturated state did not possess significant
cohesion and behaved as if it were clean sand. These tailings were shown to exhibit as low a
resistance to liquefaction as clean sand.

Seed et al. (1982, 1983) [6, 10] stated that based on both laboratories testing and field
performance, the great majority of cohesive soils that will not liquefy during earthquakes.

10

Subsequently Youd and Gilstrap (1999) [11] confirmed hence that a cohesive soil to liquefy,
must meet all the following three criteria:

The soil must have less than 15 percent of the particles, based on dry weight, that are
finer than 0.005 mm (i.e., percent finer at 0.005 mm 15 percent).

The soil must have a liquid limit (LL) that is less than 35 (that is, LL 35).

The water content w of the soil must be greater than 0.9 of the liquid limit (that is, w 0.9
(LL)).

If the cohesive soil does not meet all three criteria, then it is generally considered to be not
susceptible to liquefaction. Although the cohesive soil may not liquefy, there could still be a
significant undrained shear strength loss due to the seismic shaking.

2.3.1.2 Soil Relative Density

Based on field studies, loose cohesionless soils will contract during the earthquake which will
cause the development of excess pore water pressures [12]. For dense sands, the state of initial
liquefaction does not produce large deformations because of the dilation tendency of the sand
upon reversal of the cyclic shear stress. Poulos et al. (1985) [13] state that if the in situ soil can
be shown to be dilative, then it need not be evaluated because it will not be susceptible to
liquefaction. In other words we can say that dilative soils are not susceptible due to liquefaction
because their undrained shear strength is greater than their drained shear strength.

2.3.1.3 Particle Size Gradation

11

Soil gradation plays an important role in liquefaction. Uniformly graded nonplastic soils tend to
form more unstable particle arrangements and are more susceptible to liquefaction than wellgraded soils. Kramer (1996) [14] states that field evidence shows that most liquefaction failures
have involved uniformly graded granular soils. Well-graded soils will also have small particles
that fill in the void spaces between the large particles. This tends to reduce the potential
contraction of the soil, resulting in less excess pore water pressures being generated during the
earthquake.

The soil particle shape can also influence liquefaction potential. For example, soils having
rounded particles tend to densify more easily than angular-shape soil particles. Hence a soil
containing rounded soil particles is more susceptible to liquefaction than a soil containing
angular soil particles.

2.3.1.4 Placement Conditions or Depositional Environment

Hydraulic fills (fill placed under water) tend to be more susceptible to liquefaction because of the
loose and segregated soil structure created by the soil particles falling through water. Natural soil
deposits formed in lakes, rivers, or the ocean also tend to form a loose and segregated soil
structure and are more susceptible to liquefaction. Soils that are especially susceptible to
liquefaction are formed in lacustrine, alluvial, and marine depositional environments [7]. If the
excess pore water pressure can quickly dissipate, the soil may not liquefy. Thus highly
permeable gravel drains or gravel layers can reduce the liquefaction potential of adjacent soil.

2.3.1.5 Confining Pressure

12

The confining pressure possess inversely proportional to the liquefaction potential. The greater
the confining pressure, the soil is less susceptible to liquefaction. The higher confining pressure
can be created in conditions of deeper groundwater table, soil that is located at a deeper depth
below ground surface, and a surcharge pressure applied at ground surface [7].

2.3.1.6 Aging and Cementation

Newly deposited soils tend to be more prone to liquefaction than older deposits of soil. It has
been shown that the longer a soil is subjected to a confining pressure, the greater the liquefaction
resistance (Ohsaki 1969, Seed 1979, Yoshimi et al. 1989) [15, 4, 16]. The increase in
liquefaction resistance with time could be due to the deformation or compression of soil particles
into more stable arrangements. Cementation between soil particles increases as time passes.

2.3.1.9 Ground Shaking

It has also been determined that the historical environment of the soil can affect its liquefaction
potential. Liquefaction resistance also increases with an increase in the over consolidation ratio
(OCR) and the coefficient of lateral earth pressure. The underlying soil has been preloaded, it
will have a higher over consolidation ratio and it will have a higher coefficient of lateral earth
pressure. Such a soil that has been preloaded will be more resistant to liquefaction than the same
soil that has not been preloaded [17].

2.3.2 Geological Condition

2.3.2.1 Groundwater Table

13

The condition most conducive to liquefaction is a near-surface groundwater table. Unsaturated

soil located above the groundwater table will not liquefy. If it can be demonstrated that the soils
are currently above the groundwater table and are highly unlikely to become saturated for given
foreseeable changes in the hydrologic regime, then such soils generally do not need to be
evaluated for liquefaction potential.

At sites where the groundwater table significantly fluctuates, the liquefaction potential will also
fluctuate. Generally, the historic high groundwater level should be used in the liquefaction
analysis unless other information indicates a higher or lower level is appropriate (Division of
Mines and Geology 1997) [18].

Poulos et al. (1985)[13] state that liquefaction can also occur in very large masses of sands or
silts that are dry and loose and loaded so rapidly that the escape of air from the voids is
restricted. Such movement of dry and loose sands is often referred to as running soil or running
ground. Although such soil may flow as liquefied soil does, in this text, such soil deformation
will not be termed liquefaction. It is best to consider that liquefaction only occurs for soils that
are located below the groundwater table.

2.3.2.2 Porous Medium

2.3.3 Ground motion characteristics

2.3.3.1 Peak Ground Acceleration (amax)

2.3.3.2 Earthquake Intensity and Duration

14

For Liquefaction of soil, ground shaking duration and magnitude is one of the most important
parameters. The character of the ground motion, such as acceleration and duration of shaking,
determines the shear strains that cause the contraction of the soil particles and the development
of excess pore water pressures leading to liquefaction. The most common cause of liquefaction is
due to the seismic energy released during an earthquake. Those earthquakes which possess high
value of ground acceleration and ground shaking time is large are most vulnerable.

Although data are sparse, there would appear to be a shaking threshold that is needed to produce
liquefaction. These threshold values are a peak ground acceleration amax of about 0.10g and local
magnitude M of about 5 [7, 8]. Thus, a liquefaction analysis would be needed for those sites
having a peak ground acceleration amax greater than 0.10g or an earthquake magnitude M is also
greater than 5.0 on Ritcher scale. Besides earthquakes, other conditions can cause liquefaction,
such as subsurface blasting, pile driving, and vibrations from train traffic [2].

In summary, the site conditions and soil type most susceptible to liquefaction are as follows:

Site Conditions

Site that is close to the epicenter or location of fault rupture of a major earthquake

Site that has a groundwater table close to ground surface

2.4 Importance
Seismic events can cause a number of dangerous ground conditions which may lead to structural
damage and failure resulting in loss of life. Severe ground shaking enables lateral spreading
(such as land sliding and deposit movement/shifting), as well as incoherence means
propagation of a wave in a structure causing foundation or bridge piers to experience movement
15

that is out of sync with the rest of the structure (Fig. 2.1(a) 1964 Earthquake in Niigata, Japan
(left) [20] and Fig. 2.1(b) upper deck collapse of Bay Bridge from San Francisco, California
1989, (right) [19]). Liquefaction arises due to earthquake further amplify the potential for those
conditions and the damage they cause.

Fig. 2.1: (a) Bridge failure in 1964 Earthquake occurred in Niigata, Japan (right), and (b) upper deck collapse of Bay
Bridge from San Francisco, California 1989 (left).

It has been observed that liquefaction is also responsible for extreme property damage and loss
of life due to a several variations of failure potential during earthquake. Liquefied ground is no
longer stable to withstand the stresses it is subject to from structural foundations or even its own
weight, leading to a variety of potential failures. The witnessed effect on structures with their
foundations in a liquefied deposit resembles quicksand with a bearing capacity failure occurring
beneath the foundations. The building structures will lean and fall; or at times even split open
under the strains (Fig. 2.2) [21].

16

Fig. 2.2: structural damage caused by 1999 earthquake in Izmit Turkey.

Dams and retaining walls are common boundaries to many major bodies of water and their
adjacent shores. Both rely on the strength and stiffness properties of soil for their stability. The
failure of the soil around them not only can cause the structure itself to weaken, lean, and
possibly even fall; but also can result in subsurface landslides, during which the supporting soil
at the base of the dam or wall loosens and slides out. Dam and retaining wall failure are
especially problematic concerns due to the additional potential for flooding. The damaging
effects of liquefied soils are not only visible in the structural chaos left behind. Often, erosion
from rivers and streams cuts into the soil along their banks, leaving behind scoured ground and
gullies. The stresses produced during liquefaction can cause tension cracks to form in the soil
near the embankment, or it can collapse altogether, commonly known as lateral spreading or land
sliding. Soils on or near slopes, hills, or mountains can experience the same effects.
Lateral spreading during the Kobe earthquake in Japan in 1995 caused the paved surface to fall
1-1.2 meters below original grade as shown in Fig. 2.3 done to which local flooding occurred.

17

Fig. 2.3: Kobe, Japan earthquake occurred in 1995 (Source: http://theconstructor.org)

2.5 Liquefaction Occurrence in National Scenario

India has experienced some of the strong earthquakes, like Assam 1897 (M = 8.7), Kangara 1905
(M = 8.6), Bihar-Nepal 1934 (M = 8.4), Latur 1993 (M= 6.4), Chamoli 1999 (M = 6.8),
Muzaffarabad-Kashmir 2005 (M = 7.6, Fig.2). In recent past some frequent earthquakes jolted
the India viz Delhi NCR 2011 (M = 4.2), Sikkim India 2011 (M = 6.9), New Delhi 2012 (M =
4.9) and Koyna Nagar 2012 (M = 4.9). Five major earthquake causing liquefaction have been
discussed below.
(i) Bhuj Earthquake (26th Jan, 2002)

An earthquake originated in Bhuj (India) on 26th January 2001with maximum horizontal

acceleration 0.35 g (M = 7.7, Fig.2.4), damaged many medium and high rise buildings in and
around Bhuj city. The city buildings experienced differential settlement by violent shaking, this
happened due to liquefaction in which soil losses its strength and behaves like liquid. Soil
18

liquefaction occurred in loose, saturated cohesionless soil units (sands and silts) and sensitive
clays. A large number of water-retaining structures like earthen dams were affected by the
earthquake. Three large dams like Chang Dam, underwent severe slumping, whereas Fatehgadh
Dam and Kaswati Dam were also affected. The free-field ground motion with a PGA of 0.52 has
been recorded for these dams [22].

Fig. 2.4: Damage during Bhuj Earthquake 2001.

ii. Bihar earthquake (1934)

The 1934 Bihar earthquake is considered to be one of the worst quakes in Indian history. The
quake occurred on January 15, 1934 and was recorded 8.1 magnitude on the Richter scale. Over
30,000 people were killed in the disaster. The epicenter of the earthquake was located in eastern
Nepal. Extensive damage to life and property was observed. The intensity of the earthquake was
so strong that its effect was felt in Kolkata which in lower 650 km from epicenter. Purnea,
Munger, Muzaffarpur and Champaran were among the worst affected areas. In Muzzafarpur,
most of the buildings were damaged due to liquefaction because of Muzaffarpur city contains
mainly alluvial soil. All the Kutcha buildings collapsed while other buildings suffered damage
due to sinking and cracking of the ground [23].
19

iii. Maharashtra earthquake (1993)

Over 20,000 people were killed in the earthquake that occurred on September 30, 1993 in the
state of Maharashtra. The quake measured 6.4 on the Richter scale with epicenter at Killari
village in Latur district. Osmanabad and Latur were the worst affected areas. The destruction was
massive with over 52 villages being raised completely to the ground. The liquefaction, which
resulted from the earthquake, destroyed the foundations of the houses and caused them to
crumble. More than 60% of the deaths were a result of this [24].

Fig. 2.5: Damage due to liquefaction during Latur Earthquake (1993).

iv. Assam earthquake (1950)

The Assam earthquake also known as the Medog earthquake occurred on August 15, 1950 and
had a magnitude of 8.6 on the Richter scale. The epicenter was located at Rima in Tibet. The

20

quake caused widespread destruction in both Assam and Tibet. Over 1,500 people were killed in
Assam alone. It was considered to be among the ten largest earthquakes of 20th century [25].

v. Uttarkashi earthquake (1991)

On October 20, 1991 an earthquake measuring 6.1on the Richter scale shook the districts of
Uttarkashi,Chamoli and Tehrilocated in the state of Uttarakhand. Over a thousand people were
killed in the quake and extensive damage was caused to property. The tremors of the quake were
felt up to Delhi [26].

2.6 Liquefaction occurrence in International scenario

Some empirical approach to determine liquefaction potential
After the Loma Prieta Earthquake of October 17, 1989, Bardet et al. (1993)[27] observed and
analysed the sand boils that emerged in the Marina District of San Francisco. The sand boils
were located within the boundary of an old lagoon, the periphery of which was severely damaged
in the 1006 San Francisco earthquake. The sand boils left behind by liquefaction were useful to
delineate the liquefied area and to understand the ground displacement in the Marina District.
Holzer et al. (1999)[28] explained the heterogeneous nature of alluvial fan sediments in
combination with variations in the ground-water table can be responsible for complex patterns of
permanent ground deformation after Northridge devasting earthquake occurred on January 17,
1994. This was the most costly earthquake ever to strike the United States whose magnitude is
greater than 6.0 but horizontal acceleration measured approximately 16.7 m/s2 (1.6g). To
evaluate the mechanism of soil failure, Holzer et al. conducted at four sites for detailed
subsurface investigations in which three of them are underlain by saturated sandy silts with low
standard penetration test and cone penetration test values. Properties of soils were similar to that
San Fernando earthquake.
Prominent works are given in more detail below.
21

2.6.1 Semi Empirical Approach to Determine Liquefaction Potentials:

Standard Penetration Test (SPT) is one of the most suitable and familiar Site exploration
techniques. The SPT based methods are commonly used for determining liquefaction potential,
which is generally used for various methods. However in-situ testing is tedious, involving skilled
labour, high cost and extra time. Advantages and disadvantages are always associated with these
methods. Most of the assessment charts used Seeds method as the basis for determination of
necessary factors. Gradual improvements in these methods made it more precise and viable for
almost all kind of sandy soils.
The concept of critical SPT-N value (based on Nigata earthquake of 1964) for liquefaction
assessment of sandy soil was first proposed by Koizumi (1966)[29] and Kishida (1966)[30].
Owing to devastation from Japan earthquake Seed and Idriss (1971)[31] proposed frameworks
for SPT-N-based assessments of liquefaction potential in simplified way. This procedure is time
to time modified and improved by the researchers (Seed, 1979; Seed and Idriss, 1982; Seed et al.,
1985; Youd et al., 2001; Cetin et al., 2004; Idriss and Boulanger, 2006)[4, 6, 32, 33, 34, 35]. An
expert committee from National Research Council (NRC) of United States under the headship of
Professor Robert V. Whitman evaluated the then existing methods of liquefaction assessment in
1985. Later National Centre for Earthquake Engineering Research (NCEER) issued a report in
1997 but review continued till Youd et al. 2001[33] published final recommendation on behalf of
the committee which then became standard for liquefaction assessment. Other than councils,
individuals efforts on state of art for evaluating liquefaction by own methods or for appraising
limitations of existing methods continued, some of them are discussed in ongoing literature.

22

Numerous additional researchers have made subsequent improvement, and these types of SPTbased methods continue to evolve today.
In the year 1971, Seed et al. [31] published a literature for evaluation of liquefaction potential
based on in situ test i.e. standard penetration test on cohesionless soil. At the current time that
formulae was used mainly in United States and throughout much of the world considering as
simplified procedure. This method for evaluating liquefaction potential of soil were modified by
various researchers like Seed, 1979; Seed & Idriss, 1982; Seed et al., 1983; Seed et al., 1985 etc.
[4, 6, 10, 32].
Tokimatsu and Yoshimi (1983)[36] carried out a study on sandy soil on past Niigata earthquake
based upon SPT-N values and fines content. Along with the developments of own charts
separating liquefiable and non-liquefiable conditions is presented in terms of dynamic shear
stress ratio, SPT N values, fines content, and shear strain amplitude by using the basic approach
of Seeds method.
The 126 case history data were employed by Seed et al. (1984)[37] for re-evaluated in detail and
clarified the meaning of the values of standard penetration resistance used in correlations of field
observations of soil liquefaction with values of N, measured in SPT tests. Liquefaction resistance
curves for sands with different (Ni)6o values and with different fines contents are proposed. It is
believed that these curves are more reliable than previous curves expressed in terms of mean
grain size.
Tokimatsu and Seed (1987) [38] have established procedures to estimate strain and ground
settlement for dry and saturated sands, too. Charts are presented for estimating settlements using
these parameters, and the results are shown to compare favourably with settlements observed at
six sites for which good data on settlements have been observed.
23

Kramer S. (1989) [39] studied the effects of uncertainty and use steady-state technique for
evaluation of liquefaction potential on saturated sand. For many liquefiable soils, significant
uncertainty exists in the steady-state strength predicted by the steady-state method. For such
soils, this steady-state strength must be significantly reduced in order to reduce the probability of
misclassifying a liquefiable soil as a nonliquefiable soil to an acceptably low value.
Trifunac (1995) [40] carried out a study on fully saturated sand on the basis of five empirical
equations developed from 90 case histories of liquefaction. Developed equation helped relating
earthquake magnitude. Epicentral distance, site based motion of energy, peak ground velocity,
Fourier amplitude of velocity and duration of motion with pore pressure.
Arango I. (1996) [41] applied energy concepts to the conditions that are likely to have existed at
distant liquefaction sites in past earthquakes. The factors are independent of the field acceleration
assumed to have existed at the sites, and are only dependent on the magnitude-equivalent number
of cycles relationship. These factors are compared with others based on laboratory cyclic
strengths from Seed and Idriss and on statistical regression of data from field case histories of
liquefaction. The factors derived are based on energy concepts. From this, magnitude scaling
factors are derived that reflect field cyclic strength conditions.
Kayabali (1996) [42]; Andrus and Stokoe (2000) [43] carried out study on granular soil and soil
ranging from fine sand to sandy gravel from 26 earthquakes on more than 70 sites respectively.
They used Seeds and Seeds &Idriss method respectively, in addition new charts based on shear
wave velocity data were developed for various earthquake magnitudes.
Robertson et al. (1998)[44] described a method to estimate grain characteristics directly from the
CPT and to incorporate this into one of the methods for evaluating resistance to cyclic loading
and also provides an update method to evaluate cyclic liquefaction using the cone penetration

24

test (CPT). This study was based on sandy soil. According to Robertson for low risk project the
liquefaction potential can be estimated using penetration tests such as CPT. The CPT provides
continuous profiles of penetration resistance, For the CPT, the function of soil behaviour type
index, Ic, which is affected by a variety of grain characteristics can also be expressed. For
medium to high risk projects, the CPT can also be useful for providing a preliminary estimate of
liquefaction potential in sandy soils.
Firat et al. (2009) [45] analyzed four well known methods namely simplified procedure,
Tokimatsu and Yoshimi method, Seed-Dealba method and Japan Road Association method for
Marmara earthquake in Turkey occurred on 17 august 1999 measuring 7.4 magnitudes.
An update for Seed-Idriss simplified procedure had been laso published in the year 1999 by
Idriss [46]. In this publication Idriss incorporated rd and magnitude scaling factor and the
derived formulae is to be checked by more than one procedure like CPT, SPT, Becker test and
shear velocity techniques. He used CPT techniques surrounding followed by SPT .
Lee et al. (2001)[47] studied the liquefaction performance of soils at the site of a partially
completed ground improvement project at the Chang-Hwa Coastal Industrial Park during the
1999 Chi-Chi earthquake in Taiwan. After the earthquake, additional site exploration was carried
out using SPTs and CPTs. The data from these in situ tests carried out before and after the
earthquake and in areas with and without ground improvement are analyzed and the results are
reported.
Hosseini et al. (2001)[48] selected southern part if Iran site to evaluate liquefaction potential of
soil. This site contains cohesionless soil varies from silty sand to sandy silt. The water table is
relatively high. He found the correlation between liquefaction results based on SPT-N value and
CPT value. Robertson and Wride method was used for estimation of liquefaction potential. He
25

found that seismicity of this region is under high risk zone. He also calculated factor of safety
based on SPT-N value and CPT value.
Kostadinov et al. (2001)[49] studied on the earthquake occurred in Japan and U.S.A. and found
the capability of different types of ground motion parameters to indicate alone soil liquefaction.
They also proposed a new liquefaction detection method that simultaneously analyses
instantaneous frequency content of the horizontal and the vertical ground acceleration. Ground
motion parameters were studied and several techniques developed for detection of soil
liquefaction from seismic records.
Moaeyed et al. (2002) [50] studied to evaluate the influence of silt content on cone tip resistance
in loose silty sand mixtures and then verify the existing methods to determine liquefaction
potential. Twenty-seven cone penetration tests are performed in saturated silty sand (silt contents
ranging from 10% to 50%) samples. Consolidated tests are performed at three overburden
stresses like 100, 200 and 300 Kpa. It concludes that, as the silt content increases, the cone tip
resistance decreases.
Finn W. D. L. (2002) [51] studied the state of the art for evaluating the potential occurred in
saturated sands and silts under sufficient earthquake shaking and produce latest liquefaction
potential evaluation chart based on in situ shear wave velocity.
After Chi-Chi earthquake National Science Council conducted subsurface exploration by Chern
et al. (2003) [52]. Obtained data had been studied and three methods were employed to evaluate
liquefaction these are Seed method, T-Y method and JRA method. Multiple Additive Decision
Making method (MADM) was applied to determine the suitability between these three methods.
On the basis of results obtained, the T-Y method is the most suitable one, followed by Seed
method. The JRA method which is used in Taiwans Building Code is the worst one.
26

Lee et al. (2003) [53] gave an index, called the Liquefaction Potential Index (LPI), is calculated
based on an integration of the calculated factor of safety (FS) over depth with a weighting
function. The calculated indexes are then used to construct the failure potential maps and these
maps are checked with the field observations.
Cetin et al. (2004) [34] also define stress reduction factor in the literature and different
formulations were proposed to calculate the variation of cyclic stress ratio. In these formulae the
effect of other factors such as peak ground acceleration, magnitude of the design earthquake and
soil stiffness up to the depth of 12.0 m from ground surface.
Typhoon induced according to Chang et al. (2004) [54] water wave induced at sea bed erode
sandy shore due to cyclic pressure on ocean floor causes changes in geotechnical properties
which may cause severe damages in coastal areas. To calculate liquefaction potential on sea bed
he used Nataraja and Gills analysis (1983) [55], Ishihara and Yamasakis analysis. Collected YiLan sea bed sand and the sand is tested (Localized Testing procedure were performed) and
provide relationship between liquefaction resistance of seabed sand and wave loading period.
In the year 2004 Idriss and Boulanger [56] established a semi empirical procedure to calculate
liquefaction potential of soil on saturated cohesionless soils during earthquake. He modified
stress reduction factor, magnitude scaling factor (MSF), overburden correction factor and
overburden normalized factor of penetration resistance. He reevaluates these modified relations
with SPT and CPT-based liquefaction correlations are recommended for use in practice.
Ku et al. (2004) [58] study is also based on Chi-Chi earthquake. He presented results of the cone
penetration test (CPT) exploration and post-earthquake liquefaction analysis. Two hundred and
seventy five (275) CPT data were collected from the liquefaction-affected areas, and 46
liquefaction case histories and 88 non-liquefaction case histories were derived that can be used to
27

evaluate the accuracy of existing liquefaction evaluation models. In addition, the strength of the
liquefied soils after earthquake and the implication of its liquefaction potential in the future event
are discussed.

Lai et al. (2005) [59] developed discriminant models for evaluating liquefaction potential of soil
by using 592 datasets occurrences of liquefaction and nonliquefaction. The discriminant model
used as a multivariate statistical method and established two models through the SPT-N value are
also established in this literature, which allows calculated results to be compared to the empirical
curves. He suggested that soil liquefaction resistance (CRR)7.5 linearly proportional to fines
content for any magnitude of (N1)60 when we use simplified suggested curve by Youd et al.
(2001)[33] & Tokimatsu and Yoshimi (1983)[36].
Andrianopolous et al. (2006)[60] published a new numerical methodology for evaluation of
liquefaction phenomenon considering non-cohesive soils under small, medium and large
response. This methodology is based on Finite Difference Code (FLAC) via its under-definedmodel capability.
Idriss et al. (2006)[35]
Al-karni A. (2007)[61] assessed the liquefaction potential of the saturated soil up to depth of 25
m below the ground surface by the use of cyclic stress approach in conjunction the standard
penetration test (SPT) procedures at the location of the University of Jazan. He collected 41
borehole datasets and SPT was conducted at every 2.0 m depth. He found that wide range of soil
i.e. from sandy to silty-clayey soil. On the basis of liquefaction potential value by simplified
procedure he found that the top at10 m of soil located near and around of University of Jazan has
high liability against liquefaction.
Hasancebi et al. (2007)[62] evaluated existing correlations between shear wave velocity and
penetration resistance and compared with correlations with SPT values obtained in between
28

geotechnical and geoseismic data collected from a first-degree earthquake zone in Turkey. He
found that SPT value is a major parameter in these correlations while soil type has no
significance. The regression equations developed in this research compare well with most of the
existed equations and show good prediction capability.
Kramer et al. (2007) [63] describes a performance-based approach to the evaluation of
liquefaction potential, and showed that it can be used to account for the entire range of potential
ground shaking. Kramer also define the liquefaction dependent parameters which are position
and slope of the peak acceleration hazard curve and on the distribution of earthquake magnitudes
contributing to the ground motion hazard. The results also show that the consistent use of
conventional procedures for the evaluation of liquefaction potential produces inconsistent actual
likelihoods of liquefaction.
In the year, 1999 Marmara earthquake hits Izmit Bay and Sapanca Lake, as well as the city of
Adapazari, Turkey by a magnitude i.e. M>7.0 by which major portion of these city were
destroyed. Bol et al. (2008) [64] performed major investigation of soil i.e. approximate 700
borehole standard penetration test were performed and soil collected are tested in standard
laboratory as well as 300 cone penetration test. Researcher established database for the city
based on surface

observations as well as field and laboratory test results using GIS

methods. Adapazari criteria are similar to the Chinese Criteria. Then plot soil classification,
groundwater level, SPN-N60, organic content, undrained shear strength, allowable bearing
capacity and liquefaction potential for the

top 15m depth map. Research is continuing to

include this feature to finalise the Adapazar liquefaction map.

Due to Marmara earthquake large scale damage to structures and a death toll of more than
20,000 in the city Kocaeli, Sonmez et al. (2008)[65] studied the soil liquefaction and lateral
spreading along the southern coast of Izmit Bay. The assessment suggests that the amount of
liquefaction-susceptible layers in the soil profile reduces away from the coastal zone. A
microzonation map was created using the data from liquefaction susceptibility analyses and a
29

liquefaction severity index. The predicted liquefiable zones showed good agreement with site
observations.
Chen et al. (2008)[66] determine liquefaction potential based on situ tests like standard
penetration test and cone penetration test and developed a particular device modified from the
conventional triaxial compression test apparatus, namely Triaxial Cone Penetration Test, was
developed to obtain the peak values of cone resistance in soils so as to correlate the liquefaction
resistance of the reclaimed soils evaluated by cyclic triaxial tests.
Due to lack of shear resistance of soil liquefaction affects mostly shallow foundations and road
surfaces. Vessia et al. (2008)[67] proposed a new approach to liquefaction potential estimation
based on finite element dynamic analyses. He evaluate liquefaction potential as stress influence
factor using Westergaaard equation.
Vakili et al. (2009)[68], in this study focused on two clean sand boundary curves proposed
recently and evaluated their level of conservatism using comprehensive databases of laboratory
data and field case histories. Researchers found that how equivalent clean sand relations for
sand-silt mixtures having up to 15% of fine content may lead to un-conservative results. It is
demonstrated that there is a reasonable consistency between the prediction of the existing
relationships boundaries and laboratory evidences. By implementing cyclic triaxial and standard
penetration test, the percent of confidence to Youd et al. (2001)[33] and Cetin et al. (2004)[34]
recommendations for determining (N1)60cs have been calculated. Percent of confidence to this
relation is about 87% and 96% based on laboratory and field study respectively.
Brandenberg et al. (2010) [69] selected total 21 bridges of 79 boring logs were performed in
California and find out SPT blow count (N60), and vertical effective stress, 'v and defined Shear
wave velocity in both terms. Regression analysis was used to derive statistical relations for sand,
silt, and clay soil types.
Chao et al. (2010) [70] summarized an evaluation of the soil liquefaction potential in the Ilan
County of Taiwan whose subsurface conditions are characterized by loose uniform grained soils
30

with high groundwater table. The SPT approach was utilized in this study for the characterization
of liquefaction resistance is based on 685 bolelog datasets throughout Lanyang plain. This paper
describes the procedure for constructing the soil liquefaction potential maps of the densely
populated Ilan City and Lotung Town within the Lanyang Plain using two scenario earthquakes.
These maps can provide information for formulating disaster reduction strategy in order to
mitigate damages and losses predicted by the liquefaction hazards.
Choobbasti et al. (2010)[71] adopt Semi empirical methods based on N-SPT to evaluate
liquefaction resistance of Babolsar sandy soil. SPT tests were performed at 16 locations to
subsurface investigations out of which most of them are sandy. Water table is also very close to
ground surface. Different semi empirical methods based on blow counts by standard penetration
tests, N-SPT, liquefaction resistance of Babolsar sandy soil considering its geotechnical
characteristics was evaluated by each method and then different graphs based on CRR and
significant parameters have been drawn.
Stamatopoulos C. A., (2010)[72] had performed cyclic triaxial laboratory tests on mixtures of
sand and silt with fines content 0%, 15% and 25% and found the effect of density, consolidation
stress and non-plastic fines on the liquefaction strength. All specimens prepared for the triaxial
device had diameter 38.2 mm and height 84.9 mm. Specimens were prepared at different fines
content, void ratio and consolidation stress. The rate of applied axial strain was about 2% per
hour. The obtained results showed that the relationship between the state parameter and the
cyclic strength.
Prakash et al. (2010)[73] studied the liquefaction behavior of silts and silt clay mixers was
investigated over a range of plasticity index values of interest by conducting cyclic triaxial tests
on reconstituted samples and their behavior was compared with that of sand. The results showed
that liquefaction susceptibility of silts shows a marked change with change in the values of
plasticity index.

31

22 boreholes were conducted by Neupane et al. (2010)[74] over Kathmandu valley who studied
the relevancy of empirical approach over semi empirical approach for site specific liquefaction
investigations including preparation of more reliable liquefaction hazard map. He used only six
parameters like water table depth, grain size, depth of burial, capping layer thickness, age of
deposits and liquefiable layer thickness which are pertinent factors for liquefaction. Liquefaction
vulnerability were classified as high, moderate, low and very low on the basis of these six
factors. Seismic parameters were also considered as earthquake magnitude around 8.0 and peak
ground acceleration 0.3 g. He concluded that factor of safety increased with increase of fine
contents and Empirical approach might be more relevant than semi-empirical approach while
studying small geographic location.
On May 12, 2008 a devastating earthquake (M=8.0), with an epicenter in Wenchuan County,
struck Sichuan Province, in southwestern China, killing more than 100,000 people. After this
devastating earthquake Cao et al. (2011)[75] did field investigations following the 2008
Wenchuan earthquake (M=8.0) identified 118 liquefaction sites nearly all of which are under lain
by gravelly sediment in the Chengdu Plain and adjacent Mianyangarea. Field studies ,including
core drilling ,dynamic penetration tests (DPT),and multiple channel analysis of surface wave
velocity tests (MASW) for measurement of shear wave velocities.
Chang et al. (2011)[76] study is also based on Chi-Chi earthquake in Taiwan. In this study,
predicted the accuracy of several SPT-N-based methods liquefaction and non-liquefaction
incidents observed during the earthquake. On the basis of result they states that most sensitive
parameters are SPT blow count (N) and peak ground acceleration (amax) whereas hammer energy
ratio (ER), earthquake magnitude (M), fines content (FC), and ground water depth (GWT) follow
in sensitivity and Stress reduction factor (rd) and overburden pressure correction factor (CN)
appear least sensitive in the computed liquefaction potential.
Huang et al. (2011)[77] studied the dynamic characteristics of a liquefiable silt substratum within
the foundation soil of a reservoir dam in the Tianjin area and investigated by means of standard
penetration resistance and dynamic triaxial tests. A comprehensive program based on the
32

Chinese code and standard for geological investigation and Seeds simplified procedure was
carried out to evaluate the potential of liquefaction within the reservoir dam foundation. The
evaluation showed that saturated surface silt in the reservoir dam foundations is vulnerable to
liquefaction at seismic intensities of 7.0 and above. The two assessment methods are in good
agreement with each other, and the research results can provide useful information for the safe
construction and normal operation of the reservoir.
Eskisehir is situated within the earthquake region on the seismic hazard zonation map of Turkey
and is surrounded by several fault planes. Geotechnical datasets were collected by Tosun et al.
(2011)[78] in two ways: field and laboratory. He collected Field Geotechnical investigations data
were carried out by Standard Penetration Test (SPT). In the first stage, 232 boreholes in different
locations were drilled and SPT were performed. Test pits at 106 different locations were also
excavated to support geotechnical data obtained from field tests. They analyzed the results
obtained which indicated that presence of high ground water level and alluvial soil increase the
liquefaction potential with the seismic features of the region.
Park et al. (2012)[79] study was based on modified disturbed state concept model. Based on test
results, a classification of liquefaction phases was proposed, in terms of the dynamic effective
stress path and the excess pore pressure development. The liquefaction assessment method is
also proposed, using the disturbed state concept model based on the deviatoric plastic strain
trajectory. Factors of safety, calculated from the equivalent cyclic stress concept, were compared
with the proposed method using the original and modified disturbed state concept models. This
was proposed by using examples with different soil and earthquake conditions.
Johari et al. (2013)[80] used probability liquefaction based on standard penetration tests using
the jointly distributed random variables (JDRV) method. For developing the model, total 227 site
case histories taken, collected by Idriss and Boulanger was used. The database is composed of
115 non-liquefied cases and 112 liquefied cases. A number of models have been developed.
Among these methods, the approaches presented by Idriss and Boulanger (2010) and Juang et al.
(2012)[81,82] were selected as recent models for comparison.
33

One of the most important and familiar empirical relation i. e. Simplified procedure for
evaluation of liquefaction potential was over again revised in the year 2001 by Youd et al. [33]
along with 20 experts after 10 years of hard work in National Centre of Earthquake Engineering
Research (NCEER). According to Youd et al. the cone penetration test (CPT), the standard
penetration test (SPT), shear-wave velocity (Vs) measurements and for gravelly sites the Becker
penetration test (BPT) i.e. four tests were recommended for routine evaluation of liquefaction
potential. Seed and Idriss (1982) [6] modify the simplified procedure for earthquake magnitude
<7.5 but in this paper he define the MSF value for earthquake magnitude with >7.5 also which
are more conservative than the original Seed et al. (1982)[6] factors. This paper also considers
the peak ground acceleration to become relationships more compatible with soil conditions.
2.6.2 Soft Computing Methods:
To estimate liquefaction potential by soft computing method A. T. C. Goh in 1995 [83]
developed a back propagation artificial neural network models with a typical transfer function
i.e. sigmoid transfer function. Consequently in 2002, A. T. C. Goh [84] worked on probabilistic
neural network (PNN) approach based on the well-established Bayesian classifier method, to
evaluate seismic liquefaction potential. This paper demonstrate the usefulness of the PNN to
model the complex relationship between the seismic and soil parameters, and the liquefaction
potential using in situ measurements based on the CPT and the shear wave velocity.
Though, soft computing methods have been applied in various field of civil engineering but
limited applications are available in the area of liquefaction assessment (Goh, 1995, 2002; Wang
et al., 2010; Moradi et al, 2011; Wang and Rahman, 1999; Hanna et al., 2007a, 2007b; Hsu et al.
2006; kayabah, 1996; Sitharam, et al., 2004; Rao et al., 2007; Juang et al. 2000, 2001; Hsu et al.
2006; Ramakrishnan et al., 2008; Gracia et al., 2008; kayadelen et al., 2009).
Wang and Rahman (1999) [85] developed fuzzy artificial neural networks (FANN) for
evaluation of liquefaction based on SPT-N value. They used two different databases in which M,
34

amax, cyclic shear stress ratio, median grain diameter of the soil D50; critical depth of liquefaction
Dcr, depth of water table Dw. were common parameters whereas SPT & CPT was exceptions,
rather first database included fine content. Though numerous literatures are available but few
significant parametric studies are discussed.
Juang et al. (2000) [86] performed standard penetration test and collected 243 datasets for
evaluating liquefaction potential. He defines a boundary that separates liquefaction from noliquefaction occurrence. Firstly he train an artificial neural network from these datasets again test
for predicting the occurrence of liquefaction or no liquefaction. The successfully trained neural
network is then used to establish a liquefaction limit state function. Based on the developed limit
state function, mapping functions that relate calculated factors of safety to probability of
liquefaction are established. The established mapping functions form a basis for the development
of a risk-based chart for liquefaction potential evaluation.
Rahman et al. (2002) [87] took 205 field liquefaction datasets from 20 earthquakes and triangular
membership function to developed integrated fuzzy neural network models for the assessment of
liquefaction potential. Five datasets were wrongly predicted out of 27 and 28 cases tested in first
and second model respectively (about 18%-19%).
Baziar et al. (2003) [88] selected different sets of effective parameters for the neural network
analyses such that to reduce the noise and to obtain more accurate results. A reliable Cone
Penetration Test data set was gathered with a wide range of parameters. This data was
incorporated in Neural-Networks computer software called STATISTICA Neural-Networks. The
back propagation algorithm with a multilayer perceptron network is utilized to analyze the
liquefaction occurrence in different sites, which was further developed by him in the year 2005.
Jeng et al. (2004) [89] gave an alternative approach for the prediction of the maximum
liquefaction depth, based on neural network (NN). Unlike previous engineering mechanics
approaches, the proposed NN model is based on data learning knowledge, rather than on
knowledge of mechanisms.
35

Chen et al. (2005) [90] developed a seismic wave energy-based method with back-propagation
neural networks to assess the liquefaction probability. The proposed method shows capability in
evaluating the probability of soil liquefaction based on the boundary curve and a logarithm
normal distribution.
Hsu et al (2006) [91] reported that high fines content (FC) and high cyclic stress ratio is the two
main characteristics of the liquefied and non-liquefied cases in Taiwan. Field performance data
generated from several earthquakes in Taiwan and gathered with previous records were used for
SPT based models of neural networks.
Su et al. (2006)[92] developed a backpropagation artificial neural network (ANN) model to
predict the liquefaction cyclic resistance ratio (CRR) of sands using data from several laboratory
studies involving undrained cyclic triaxial and cyclic simple shear testing.
Hanna et al (2007) [93, 94] explored GRNN methods to address collective knowledge from
simplified procedures to assess nonlinear liquefaction potential. To meet this objective SPT and
CPT results from 1999 Turkey and Taiwan earthquake were used. Further, liquefaction decision
was validated by the SPT, confirming the viability of the SPT to CPT data conversion which is
the main limitation of most of the simplified methods.
Gracia et al. (2008) [95] developed a hybrid system named neurofuzzy, which profits from fuzzy
and neural paradigms, is advanced. The resulting model called NEFLAS (NEuro Fuzzy
estimation of liquefaction induced LAteral Spread) is shown to yield a much improved
forecasting than both multiple regression and neural network procedures.
Jha et al. (2009) [96] using a simplified deterministic Seed method, this reliability analysis has
been performed. The probability of liquefaction along with the corresponding factor of safety
have been determined based on a first order second moment (FOSM) method, an advanced
FOSM (HasoferLind) reliability method, a point estimation method (PEM) and a Monte Carlo
simulation (MCS) method. A combined method that uses both FOSM and PEM is presented and
found to be simple and reliable for liquefaction analysis. Based on the FOSM reliability
36

approach, the minimum safety factor value to be adopted for soil liquefaction analysis
(depending on the variability of soil resistance, shear stress parameters and acceptable risk) has
been studied and a new design safety factor based on a reliability approach is proposed.
Lee et al. (2009) [97] presented a multilayer perceptron neural network (MLP model) approach
for recognition of field liquefaction. Firstly a MLP model is developed to discriminate between
the cases of liquefaction and non-liquefaction chances. Performance of MLP model are good
predictive and generalization, with the accuracy rate 98.9% in the training phase, 91.2% in
testing phase and 96.6% on both cases. Using this model, the SFi values are then calculated and
reveal that peak ground acceleration (PGA) is the most sensitive factor in both the liquefaction
and non-liquefaction cases.
Baykasoglu et al. (2009) [98] gave a new approach which is based on mining data to predict
liquefaction prediction. The proposed approach is based on extracting accurate classification
rules from neural networks via ant colony optimization. The proposed algorithm is also
compared with several other data mining algorithms. It is shown that the proposed algorithm is
very effective and accurate in prediction of liquefaction.
Jafarian et al. (2010) [99] employed genetic programming (GP) to develop a new empirical
analytical equation studied the vmax/amax ratio of strong ground motions can be used in seismic
hazard studies as a parameter that captures the influence of frequency content. The suggested
model is a function of earthquake magnitude, closest distance from source to site (Rclstd),
faulting mechanism, and average shear wave velocity over the top 30 m of site. A wide ranging
database of strong ground motion released by Pacific Earthquake Engineering Research Center
(PEER) was utilized. It is demonstrated that residuals of the final equation show insignificant
bias against the variations of the predictive parameters. The results indicate that vmax/amax
increases through increasing earthquake magnitude and source-to-site distance while magnitude
dependency is considerably more than distance dependency. In addition, the proposed model
predicts higher vmax/amax ratio at softer sites that possess higher fundamental periods.
37

Consequently, as an instance for the application of the proposed model, its reasonable
performance in liquefaction potential assessment of sands and silty sands is presented.
Farrokhzad et al. (2010) [100] research work study the liquefaction potential of Babol city with
30 borehole datasets over 7 km2 area i.e. total 2500 samples taken. Seed and Idriss (1983)
empirical method was used to evaluate liquefaction potential of soil. Artificial neural network
(ANN) technique was used to develop the predictive model. These datasets are divided into three
categories namely training (60%), testing (30%) and validation (10%), based on random
selection along with back propagation algorithm. In this analysis, the number of epochs varied
between 500 to 800. There is one hidden layer and layer is activated by tan-sigmoid transfer
function. Three ANN models were developed with varied input parameters. The average
accuracy between ANN predicted and real data in all cases is over 91%. This study showed that
the neural networks are a powerful computational tool which can analyse the complex
relationship between soil liquefaction potential and effective parameters in liquefaction.
Sen et al. (2010) [101] developed a genetic algorithm-based model to determine the liquefaction
potential by confirming Cone Penetration Test datasets derived from case studies of sandy soils.
Cha et al (2011)[102] established and compared single-artificial neural network (SANN) and
multi-artificial neural network (MANN) models, and applied these models to predict wave
induced liquefaction potential in a porous seabed. The results indicated accuracy of MANN
model in the prediction of the wave-induced maximum liquefaction depth.
To model, the soil mass with liquefaction potential, Moradi et al. (2011) [103] used FLAC
software. With FLAC software, the effects of pore water pressure with or without loss of pore
water pressure and the generation of pore water pressure can be calculated with the help of Finn
constructive model.
On 22 June 2002, at 07:28:20 local time, a destructive earthquake struck some vast regions in
Qazvin, Hamedan, and Zanjan provinces in Iran. Shahari et al. (2011) [104] study is based on
this devasting earthquake. The results of this study showed that Abbas Converter is a reliable
38

tool for site response analysis. Comparison of the site response analysis of a proposed profile
agreed good reasonable matching by the known applicable procedures. This study shows that the
proposed method can be used for site response analysis.
Cabalar et al. (2012) [105] gave a review of the Adaptive Neuro-Fuzzy Inference System
(ANFIS) in current use for geotechnical engineering-based studies, as well as some applications
employed in resonant column testing, triaxial testing, and liquefaction triggering.
Khozaghi et al. (2012) [106] predicted the potential of liquefaction through neural network
approach by using data from sounding in the southeast part of Tehran. Collection of data was
based upon 30 km2 area with a high level of underground water. The neural network having one
hidden layer, is trained and tested by some new data, based on standard penetration test, in order
to ensure the efficiency operation of the network. After all, the result of neural network method
can be compared with the result of Seed method for predicting liquefaction and was shown that
the neural network method could predict with 92 percent accuracy in the southeast area of
Tehran.
Bagheripour et al. (2012) [107] developed a new relation to determine liquefaction potential
depend on the estimated probability of liquefaction (PL), cyclic stress ratio (CSR) and
normalized standard penetration test (SPT) blow counts while containing a mean error of less
than 10% from the observational data. Advanced First-Order Second-Moment (AFOSM)
technique associated with genetic algorithm (GA) and its corresponding sophisticated
optimization techniques have been used to calculate the reliability index and the probability of
liquefaction.
Xue et al. (2013) [108] studied the feasibility of using fuzzy comprehensive evaluation model for
predicting soil liquefaction during earthquake. Fuzzy systems are used to handle uncertainty
from the data that cannot be handled by classical methods. It uses the fuzzy set to represent a
suitable mathematical tool for modelling of imprecision and vagueness. The pattern classification
of fuzzy classifiers provides a means to extract fuzzy rules for information mining that leads to
39

comprehensible method for knowledge extraction from various information sources. The model
is trained with large databases of liquefaction case histories. Nine input parameters such as
earthquake magnitude, water table, total vertical stress, effective vertical stress, depth, peak
ground acceleration, cyclic stress ratio, mean grain size and measured cone penetration test tip
resistance were used. ANFIS model is a fairly promising approach for the prediction of the soil
liquefaction potential and capable of representing the complex relationship between seismic
properties of soils and their liquefaction potential.
4.2 National Status
Rao and Satyam (2007) [109] developed a liquefaction hazard map of Delhi using SPT-N based
methods with 1200 borehole datasets at various locations since, Delhi falls in the high seismic
zone with high seismic probability. Sitharam et al. (2004) [110] also work on the determination
of dynamic properties cum liquefaction potential of soil of sands which were collected from
Sabarmati river belt Ahmedabad. A simple procedure for the dynamic properties of layered
ground has been obtained.
Ramakrishnan et al. (2008) [111] developed a backpropagation artificial neural network to
predict CSR with 23 datasets of Bhuj city. They established a relation between liquefaction
severity index (LSI), liquefaction sensitivity index (LSI) and cyclic resistance ratio (CRR) and
cyclic stress ratio.
Jain et al. (2000)[112] studied the developments of earthquake engineering in India during the
last one hundred years, the current status of earthquake risk reduction in India, strengths and
weaknesses of Indian model of earthquake engineering developments, and the future challenges.
Chakrabortty et al. (2004) [113] did an extensive analysis for determination of liquefaction
hazard of Kolkata city at different locations. The areas with river channel deposit are the most
hazardous area for liquefaction. From the study it is also concluded that if acceleration level is

40

increased then more area will be affected due to liquefaction. The plotted contour maps would
assist the designers in taking suitable decision regarding necessary sub-soil treatment at different
locality based on the design Peak Ground Accelerations.
Sitharam et al. (2004)[110] studied the methods of determining the dynamic properties as well as
potential for liquefaction of soils. Parameters affecting the dynamic properties and liquefaction
have been brought out. A simple procedure of obtaining the dynamic properties of layered
ground has been highlighted. Results of a series of cyclic triaxial tests on liquefiable sands
collected from the sites close to the Sabarmati river belt have been presented.
Singh et al. (2005)[114] examined the nature of distress in seven relatively severely affected
damsThe consequences of these problems were not very severe because of the fact that (a) the
reservoirs in question were almost empty at the time of the earthquake and (b) the dams
performed reasonably in spite of being shaken by free-field horizontal peak ground acceleration
(PGA) as high as 0.5g.
Kanth et al. (2010)[115] simulated the liquefaction hazard due to great earthquake events in the
past. The obtained results are in general agreement with the reported damages due to the past
earthquakes.
Maheshwari et al. (2010)[116] studied the effects of fine silts on liquefaction potential of sandy
soil. Tests have been conducted on the vibration table at different accelerations and pore water
pressure is measured. The results of the study performed are used to clarify the effects of nonplastic fines content on the Solani sand.
Vipin et al. (2010)[117] estimated the liquefaction return period for Bangalore, India, through a
probabilistic approachthe entire range of peak ground acceleration (PGA) and earthquake
magnitudes was used in the evaluation of liquefaction return period. The seismic hazard analysis

41

for the study area was done using probabilistic approach to evaluate the peak horizontal
acceleration at bed rock level. Based on the results of the multichannel analysis of surface wave,
it was found that the study area belonged to site class D.
Raghunandan et al. (2011)[118] studied intrusion of air and its dissolution over prolonged period
is first reviewed using centrifuge test. Cyclic triaxial tests were conducted on samples of sand
pre-introduced with air pockets to investigate the cyclic response and resistance to liquefaction.
The cyclic triaxial test setup and method used to introduce air in sand samples are then
described.
Dixit et al. (2012)[119] made an attempt to study the susceptibility of soil liquefaction using
simplified empirical procedure based on number of blow counts (N values) of the soil layers
from standard penetration test.
Muduli et al. (2013)[120] studied the evaluation of liquefaction potential of soil based on
standard penetration test (SPT) dataset using evolutionary artificial intelligence technique, multigene genetic programming (MGGP)
Das et al. (2013)[121] made an attempt to develop a limit state function for assessing the cyclic
resistance ratio (CRR) of soil based on cone penetration test (CPT) data obtained after Chi-Chi
earthquake, Taiwan, 1999, using evolutionary artificial intelligence technique, genetic
programming (GP), and to evaluate the liquefaction potential of soil in a probabilistic approach
through a Bayesian mapping function
Sharma et al. (2013) [122] evaluated the liquefaction potential of saturated cohesionless deposits
in Guwahati city, Assam. The liquefaction potential was evaluated by determining factor of
safety against liquefaction with depth for areas in the city. A soil database from 200 boreholes
covering an area of 262 km2 was used for the purpose. The results show that 48 sites in Guwahati

42

are vulnerable to liquefaction according to the Seed and Idriss method and 49 sites are vulnerable
to liquefaction according to the Idriss and Boulanger method. Results are presented as maps
showing zones of levels of risk of liquefaction.
Satyam et al. (2014) did a detailed assessment of liquefaction hazard, important for evaluating
and reducing the risk through appropriate mitigation techniques. The liquefaction susceptibility
can be mapped using specific, well established geologic and geotechnical criteria. Damages
caused by liquefaction of saturated soil revealed that after liquefaction the ground failed, sand
boiling occurred and the structure subsided unevenly causing tilting, cracking or even collapse.
In India few researchers are working in this area and the work carried out by these researchers is
focused to the particular area. Therefore, liquefaction assessment of new regions should be taken
up extensively for awareness and precaution. This research is emerging in India, which will lead
to basic investigation and research in the Geotechnical Earthquake Engineering at local level.

2.7 Assessment of liquefaction potential by Empirical Methods:

Geotechnical professionals generally investigate subsurface to evaluate the potential for
liquefaction. The most common techniques using Standard Penetration Test (SPT) blow count
(commonly referred as to the N-value) follows certain procedures:

1. The first step in the liquefaction analysis is to determine if the soil has the ability to
liquefy during an earthquake. The vast majority of soils that are susceptible to
liquefaction are cohesionless soils. Cohesive soils should not be considered susceptible to
liquefaction unless they meet all three criteria listed in Sec. 2.3.1.1 .
Estimation of the cyclic stress ratio (CSR) induced at various depths within the soil by
the earthquake.
43

2. By using the standard penetration test, the cyclic resistance ratio (CRR) of the in situ soil
is then determined. If the CSR induced by the earthquake is greater than the CRR
determined from the standard penetration test, then it is likely that liquefaction will occur
during the earthquake, and vice versa.
3. The final step is to determine the factor of safety against liquefaction, which is defined as
FS is equal to CRR/CSR.

The SPT based bore log charts are commonly used for determining liquefaction potential. Most
of the assessment charts uses Seeds method as the basis for determination of necessary factors.
Gradual improvements in these methods made it more precise and viable for almost all type of
sandy soils. Different empirical methods like modified Seeds method, Tokimatsu and Yoshimi
method and Idriss and Boulanger methods will be applied for estimating of liquefaction potential
of soil using field and laboratory datasets.
After the devasting earthquake comes in Alaska and Niigata (Japan) in 1964, compelled the
researchers to study about to estimate liquefaction potential of soil. First of all Seed et al. (1971)
published a literature to calculate liquefaction potential of soil. The method has been updated
from span of time by different researchers. Hence there are so many empirical methods but in
present research works following three empirical methods has been used. The selection of these
methods is based on its wide used by many researchers due to applicability. These are as follows:
i.

Modified Seeds method

ii.

Tokimatsu and Yoshimi method

iii.

Idriss and Boulanger method.

2.7.1 Modified Seeds Method

44

Earthquake in Alaska and Niigata (Japan) in 1964 compelled the researchers to develop the most
common type of analysis to determine the liquefaction potential is to use the standard penetration
test (SPT) (Seed et al. 1985, Stark and Olson 1995). The analysis is based on the simplified
method proposed by Seed and Idriss (1971) is often termed the simplified procedure. The
procedure was modified and improved periodically during the time, primarily through landmark
papers by Seed (1979) and subsequently Seed and Idriss (1982), Seed et al. (1985). After 3
decades Youd et al. (2001) again modified Seeds method in laboratory held by National Center
for Earthquake Engineering Research (NCEER). This is the most commonly used method to
evaluate the liquefaction potential of a site. On the basis of detailed study, simplified procedures
have been discussed in detail.
In Newtons 2nd law of motion, the horizontal earthquake force F acting on the soil column has
a unit width and length i.e.

( )

( )

(2.3)

Where, F = horizontal earthquake force acting on soil column.

m = total mass of soil column i.e. ( ).
= total unit weight of soil
z = depth from the ground level
= acceleration which in this case is maximum horizontal ground acceleration caused by the
earthquake i.e.

max.

= total vertical stress at bottom of soil column.

g = acceleration due to gravity.
45

The force F acting on the rigid soil element is equal to the maximum shear force at the base on
the soil element. Since the element is assumed to have a unit base width and length, the
maximum shear force F is equal to the maximum shear stress as shown in fig 1.

(3.4)

Since the soil column act as a deformable material rather than rigid body during the earthquake
Seed and Idriss [4] incorporated a depth (or stress) reduction factor

in the right side, hence the

equation has been modified as

max
Fig. 3.1: Conditions assumed for evaluation of the CSR.

(3.5)

Average value of Stress reduction factor rd is given as:

rd= 1.0 - 0.00765 z ; for z 9.15 m

(3.6a)

rd= 1.174 - 0.0267 z ; for 9.15 m < z 23 m

(3.6b)

rd= 0.744 - 0.008 z; for 23 m < z 30 m

(3.6c)

rd= 0.50; for z > 30 m

(3.6d)

As depth (z) increases rd also increases. The mean value of rd calculated from above equation is
shown in figure below.
For ease of computation, the mean value curve plotted in Fig 3.2 may be approximated by the
following equation [22]:
46

(3.7)
For simplified method Seed et al [23] considered the soil in the field to undergo by average stress
avg, which is 0.65 of max. Subsequently the average shear stress is normalized by the vertical
effective stress to obtain CSR induced by the earthquake given in Eqn. (3.8):

)(

(3.8)

Where v = total vertical stress

v = total vertical effective stress (v-u)
u = pore water pressure.

Fig. 3.2: rd versus depth curves (Youd and Idriss 2001).

Cyclic Resistance Ratio (CRR):

Youd et al. (2001) approximated the simplified base curve using the following equation:

(3.9)

47

Above equation is valid for (N1)60 >30 where x = (N1)60 > 30and is fixed at 1.20; a = 0.048; b = 0.1248; c = -0.004721; d = 0.009578; e = 0.0006136; f = -0.0003285; g = -1.673E-05 and h =
3.714E-06.
CRR7.5 is the cyclic resistance ratio for magnitude of 7.5 earthquakes, magnitude smaller or
larger than 7.5, introduces a correction factor namely Magnitude Scaling Factor (MSF) defined
by the following equation given by Youd et al. (2001):

(3.10)

The appropriate cyclic strength obtained is:

(3.11)
2.7.2 Tokimatsu & Yoshimi (T-Y) method:
CSR defined by this method is:
CSR =

(3.12)

Where avg= amplitude of uniform shear stress cycles equivalent to actual seismic shear stress
time history.
= The maximum horizontal acceleration at ground surface
= initial effective vertical stress.
= initial vertical stress contribution to the shear stress, defined by
M = magnitude of earthquake.

48

(3.13)

Where = unit weight of soil, z = depth below the ground surface,

CRR specified by Tokimatsu and Yoshimi is;

) ]

(3.14)

Where l = shear stress on horizontal plane; Cr, a & n are correction factors and may be taken as
0.57, 14 & 0.45 respectively.

and subsequently N1, may be calculated from the following

pairs of equations;

(3.15)

(3.16)

Where,

is correction factor for SPT-N value and

is effective vertical stress.

2.7.3 Idriss & Boulangers method:

Modus operandi by Idriss & Boulanger [ ] for evaluation of CSR is same as simplified method.
Right after CSR calculated value of CSR is adjusted for the moment magnitude M = 7.5.
Accordingly the value of CSR is given as
(

(3.17)

Where,
Idriss and Boulanger (2002) adjusted the equation of CRR for clean sands as follows

49

(3.18)

Subsequent expressions describes the way parameters in the above equation is calculated
=

(3.19)
(

))

(3.20)

(3.21)
Where, FC = Fine content
The use of equations in preceding articles provides a convenient means for evaluating the cyclic
stress ratio required to cause liquefaction for cohesion-less soils with varying fines content.
The factor of safety against liquefaction is defined as:

(3.22)
Liquefaction is predicted to occur when FS 1.0, and if FS > 1, the soil will not undergo for
liquefaction. The higher the factor of safety, the more resistant soil is to liquefaction. However,
soil that has a factor of safety slightly higher than 1.0 may still liquefy during the earthquake.
Estimation of CSR, CRR and MSF require necessary assumptions on early stages whereas
alternatively computational models will save time by omitting lengthy and tedious task of
calculation of aforementioned parameters. Some pertinent soil properties along with seismic
characteristics will help in modeling and analyzing liquefaction potential of sites. The major
advantage of computational methods is the ability to associate both SPT and CPT indicator
properties for better engineering judgment.
50

2.8 Soft computing methods:

2.8.1 ANN method:
Artificial neural network (ANN) consists of a large number of interconnected processing units
known as biological neuron (Goh, 1995). It is a soft computing approach that is inspired by the
function and structural aspects of biological neurons. ANN is advanced and standard tools
simulated by the mathematical model and computational model of the human brain and being
used around the globe to find solutions to a wide variety of non-linear statistical data
complications (Hanna et al. 2007). They are usually used to model complex relationship among
inputs and targets to find patterns in datasets. Interconnections among neurons are established by
weights, which are applied to all values passing through one neuron to another. The ANNs are
arranged in three or more layers, one input layer, one or more hidden layers and one target layer.
Each neuron receives an input signals from the previous neurons connected. Each of these
connections has numeric weights associated with it. The signals from each input are then
processed through a weighted sum of the inputs, and the processed output signal is then
transmitted another neuron via a transfer of activation function. Once the network trained with
sufficient number of datasets, it can validate, the trained network required to make predictions
for a new set of data that it has never been introduced during the previous phases (Rao et al.
2007). Due to its multidisciplinary nature, ANN is becoming advanced and standard tool for
accomplishment of their work.
The Feed-forward back propagation technique learning algorithm will be applied in this study to
evaluate liquefaction potential. It has two stages, in first stage, the inputs are forwarded from
input layer to output layer. After computing the errors of each output between computed and
desired output, in second stage information is send backward to the inputs which readjust the
51

connecting weights in the hidden and output layer to minimize this error. The modification of the
weights is carried out by using generalized delta rule.

Feed-forward back propagation technique:

In this section here we used Feed-forward back propagation technique. In this technique learning
algorithm has two stages. In first stage, the inputs are forwarded from input layer to output layer.
After computing the errors of each output between computed and desired output, in second stage
information is send backward to the inputs which readjust the connecting weights in the hidden
and output layer to minimize this error. The modification of the weights is carried out by using
generalized delta rule [23].

Fig. 3.3: Feed-forward back propagation network

Activation/Transfer function:
Though many activation functions exist, the most common is the sigmoid activation function,
which outputs a number between 0 (for low input values) and 1 (for high input values). The
resultant of this function is then passed as the input to other neurons through more connections,
each of which are weighted [24]. Sigmoid transfer function is expressed as:
(3.23)

52

Fig. 3.4: Sigmoid transfer function.

2.8.2 Adaptive Neuro-Fuzzy Inference Method:

Neuro-fuzzy refers to the combination of artificial neural network and fuzzy logic. Neuro-fuzzy
which was first proposed by Jang and Sun (1995) in fuzzy modeling environment is divided into
two areas: linguistic fuzzy modeling which is focused on interpretability is mainly the Mamdani
model; and precise fuzzy modeling that is focused on accuracy is mainly the Takagi-SugenoKang (TSK) model (Kalkan et al. 2009).
Adaptive Neuro-Fuzzy Inference System (ANFIS) is one of the most successful schemes which
combine the benefits of ANN and FIS into a single capsule (Jang 1993). The attractive features
of an ANFIS include: easy to implement, fast and accurate learning, strong generalization
abilities, excellent explanation facilities through fuzzy rules, and easy to incorporate both
linguistic and numeric knowledge for problem solving (Jang and Sun 1995; Jang, Sun, and
Mizutani 1997). According to the neuro-fuzzy approach, a neural network is proposed to
implement the fuzzy system. A typical architecture of an ANFIS, in which a circle indicates a
fixed node, whereas a square indicates an adaptive node, is shown in Figure 3.5. In this structure,
there are input and output nodes, and in the hidden layers, there are nodes functioning as
membership functions (MFs) and rules. For simplicity, we assume that the examined FIS has two

53

inputs and one output. For a first-order Sugeno fuzzy model, a classic rule set with two fuzzy "if
then" rules is as following [Terzaghi K., Peck R. B., and Mesri G., 1996]:

Fig. 3.5: First order Sugeno ANFIS architecture.

Rule 1: if x is A1 and y is B1, then f1 = p1x+q1y+r1,

(3.24a)

Rule 2: if x is A2 and y is B2, then f1 = p2x+q2y+r2.

(3.24b)

Where, x and y are the two crisp inputs, and Ai and Bi are the linguistic labels associated with
the node function.
As indicated in Fig. 3.5, the system has a total of five layers. The functioning of each
layer is described as follows [Jang, J. S. R. 1993 & Varghese, P.C. 2007].

Input node (Layer 1): Nodes in this layer contains membership functions. Parameters in this
layer are referred to as premise parameters. Every node i in this layer is a square and adaptive
node with a node function:
For i = 1, 2.

(3.25)

54

Where x is the input to node i, and Ai is the linguistic label (small, large, etc.) associated with
this node function. In other words,

is the membership function of Ai and it specifies the

degree to which the given x satisfies the quantifier Ai.

Rule nodes (Layer 2): Every node in this layer is fixed node labeled , whose output is product
of all incoming signals.
For i = 1, 2

(3.26)

Average nodes (Layer 3): Every node in this layer is fixed node labeled N. The ith node
calculates the ratio between the ith rule's firing strength to the sum of all rules' firing strengths.
Every node of these layers calculates the weight, which is normalized. For convenience, outputs
of this layer are called normalized firing strengths.

For i = 1, 2

(3.27)

Consequent nodes (Layer 4): Every node i in this layer is an adaptive node with a node function

(3.28)

Where is a normalized firing strength from layer 3 and

is the parameters set of

this node. Parameters in this layer are referred to as consequent parameters.

Output node (Layer 5): The single node in this layer is a fixed node labeled , which computes
the overall output as the summation of all incoming signals:
55

Overall output =

(3.29)

This study will be the effort to assess liquefaction potential of Allahabad city, which will
aid in determining liquefaction prone area with the help of conventional and computational
methods. The assessed liquefaction potential of site will be useful for mapping and mitigation.
2.8.3 Multiple Linear Regression (MLR) Analysis
Regression analysis is the study of establishing the functional relation between independent and
dependent variables. The independent variables may vary from one or greater than one
depending on the requirement of the dependent models. However, the number of dependent
variables is strictly restricted to one. The general formula of regression establishing relationship
between different independent variables and a dependent variable is shown below [Orlov, M.
1996]:
y = b0 + b1x1 + b2x2 + bpxp

(3.30)

or, y = b0 + bixi

(3.31)

Where, i =1, 2, p
y is the dependent variable (liquefaction potential in this case)
x1, x2 ---- xp are independent variables,
And b0, b1, b2---bp are the coefficients that has to be determined using regression analysis.
3.7 Technical Note:
56

Over the past few decades many urban areas have experienced severe damage due to
liquefaction-induced soil movements.
In this study urban area of Allahabad city is going to be divided into small zones and in each
zones some suitable points will be located to conduct geotechnical investigations in two stages
i.e. field and laboratory investigations. In the first stage, boreholes in different locations will be
drilled up to 10 m and Standard Penetration Test (SPT) will be performed at the interval of 1.5 m
or with the change in soil strata. In the second stage, experimental studies will be performed on
disturbed soil to determine the Atterbergs limits, cohesion, angle of internal friction, particle
size distribution whereas undisturbed samples will be used to determine natural water content,
bulk unit weight. All experiments will be conducted as per the guide lines of bureau of Indian
standard codes.
Soil properties obtained from field and laboratory investigations will be used to prepare bore-log
chart for datasets. These datasets will be used to determine liquefaction potential of soils by
conventional methods. As per literature review there are many conventional methods to
determine liquefaction potential of soils. Therefore, in the present study selected conventional
methods will be used to determine liquefaction potential on the basis of their merits, demerits and
applicability in the proposed research work. Some selected conventional methods based on
standard penetration test outputs, proposed by the researchers are modified Seeds method,
Tokimatsu and Yoshimi method, Idriss and Boulanger method will be used to determine
liquefaction potential. The comparisons of these methods will be carried out and they will be
critically evaluated on the basis of their limitations.

57

Liquefaction potential assessed by conventional methods will be used for the development of
soft computational models using neuro-fuzzy technique. Experimental geotechnical data will be
collected for different zones. Additional bore log data will also be collected from various sources
like consultancy projects, field investigations carried out by different labs and organization to
ensure sufficient number of datasets for training, testing and validation of models. In soft
computing method pertinent input vectors from datasets will be selected and liquefaction
potential as output vector will be obtained from the conventional methods. Some datasets will be
reserved for testing and validation of developed models. The developed model will ease the
tedious manual calculations and may be used as estimation tool for new constructed area. The
obtained values will also being useful for liquefaction potential contouring of small zone in
Allahabad city at different depth to identify the liquefaction prone areas.

58

Chapter 3

DATA COLLECTION AND GEOTECHNICAL INVESTIGATION

3.1 General
To develop the neural network (NN) model, neuro-fuzzy (NF) model and multi linear regression
(MLR) model data play an important role. Therefore, in this chapter data collection and
geotechnical investigation of the study area have been extensively discussed and presented.

Our study is based on Allahabad city which is situated in the Southern Eastern part of the Uttar
Pradesh State. It lies between the parallels of 24 28 north latitude and 81 54 east longitudes
and stands at the confluence of the Ganga and Yamuna rivers (mapsofindia.com). The location of
the study area is shown in fig. 3.1. Its metropolitan area is 63.07 km2 (24.35 sq mile). As per
2011 census, Allahabad is the seventh most-populous city in the state of Uttar Pradesh and the
thirty-sixth most-populous city in India, with an estimated population of 1.11 million in the city
and 1.21 million in its metropolitan region. In 2011, it was ranked the world's 130th fastestgrowing city. Provisional data suggest a density of 1,087 people per km2 in 2011, compared to
901 in 2001. Native people from Uttar Pradesh form the majority of Allahabad's population.

To its south west Bundelkhand region is situated, to its east and south east is the Baghelkhand
region, to its north and north east is the Awadh region and to its west is the (lower) doab of
which it itself is a part (Wikipedia.com). The city is divided by the railway line running through
it. South of the railway line is the Old Chowk area, Civil lines is situated in north. Allahabad
stands at a strategic point both geographically and culturally. A part of the Ganga-Yamuna Doab,
it is the last point of the Yamuna River, and culturally, the last point of the Indian west. As with
the rest of Doab, the soil and water are predominantly alluvial in origin.
59

Generally from February there is rapid increase in temperature, May is the hottest month with the
mean daily maximum temperature is 41.50C and mean daily minimum temperature 260C. After
the onset of the monsoon there is appreciable drop in temperature, January is the coldest month
with mean daily maximum temperature is 26.200C and mean daily minimum temperature is
9.30C. The mean monthly maximum temperature is 19.540C and mean monthly minimum
temperature is 6.230C.

3.2 Data required for assessment of liquefaction potential

Generally two types of parameters are required for the assessment of liquefaction potential.
i. Geotechnical Data: Conventional method govern the data requirements.
Therefore, geotechnical data required for liquefaction potential assessment are:
Depth; SPT-N value; Soil type; Grain size; Density; Depth of water table etc.
ii. Seismic Data: Conventional method has the relationship between geotechnical
data and seismic data therefore seismic data required for liquefaction potential
assessment are Earthquake Magnitude; maximum horizontal acceleration.
3.3 Geotechnical Data Collections
Geotechnical data required for the assessment of liquefaction potential were collected from
various sources and agencies like MNNIT Allahabad, Public Works Division (PWD) Allahabad,
Uttar Pradesh Jal Nigam Limited (UPJNL) etc. The maximum data were collected from MNNIT
Allahabad. These data were obtained from the geotechnical investigations report carried out for
various purpose. Datasets were photocopied. Most of these were obtained from the bank of river
Ganga, Yamuna and areas where urbanization grows rapidly. Numbers of data collected from
these were 350 number of borehole. Locations of these datasets are provided in table 3.1. These
60

data consist of cohesive and cohesionless soils. Out of 350 borehole datasets only 158 borehole
datasets were chosen for further study on the basis of fines content. Maximum borelog chart
contains soil profile upto a depth of 12.0 m. These borelog charts gives the information about
Grain size distribution analysis, Atterberg limits (if required), shear strength parameters (c and
), natural moisture content bulk unit weight, types of soil, SPT-N value, depth of water table
etc. After collecting borehole datasets of various locations it was thoroughly studied found that
data of some more location is required to fill the gap between geotechnical data. To bridge the
gap geotechnical investigations were performed to obtain the sufficient datasets of some new
locations.
3.4 Geotechnical Investigation of New Location
After the collection of subsurface borehole datasets by different agencies, geotechnical field and
laboratory investigation has also been carried out for those areas where borehole datasets are
absent or less in number. Standard penetration test were performed for the collection of disturbed
and undisturbed soil samples and these soil samples were brought in the laboratory. Disturbed
and undisturbed soil samples were collected from these boreholes up to the depth of 12.0 meters.
Disturbed soil samples were used to determine liquid limit; plastic limit; particle size finer than 2
mm, 0.075 mm and 0.002mm and undisturbed samples were used to determine natural water
content [25-27], bulk unit weight and strength parameters. All experiments were conducted
according to bureau of Indian standards guidelines for soil testing in the geotechnical
engineering laboratory of civil engineering department in MNNIT Allahabad. When all
properties like index properties, physical properties and shear strength properties were
determined then borelog chart for that borehole was prepared as shown in table 3.2. Similar step
were adopted to prepare borelog chart for all the boreholes as shown in annexure-A.
61

In this research we performed Standard Penetration test (SPT) at 31 locations. The SPT N-value
performed at a regular interval of 1.5 m depth [24]. Details of one borehole details out of 31
borehole data collection is shown in table 3.1 below. Investigated data were used for the
liquefaction assessment and these were also used for the validation of developed model.

62

63

Table 3.1: Datasets shown in the table for one borelog chart and calculation of FOS by Modified Seeds
method under the conditions given below.

3.5 Groundwater level data of Allahabad city

Water table is one of the important parameter used in the assessment of liquefaction potential.
Therefore, variation of ground water table level has been discussed here. About 90% of rainfall
takes place from June to September. During monsoon surplus water is available for deep
percolation to ground water. The average annual rainfall in Allahabad region is noted as 934
mm. Climate is sub humid and is characterized by hot summer, pleasant monsoon and cold
season. Ground water is mainly controlled by drainage, topography and lithological behavior. It
occurs underground water condition at shallow depths and under confined condition at deeper
depths. Depth to water in pre monsoon ranges between 3.0 to 15.0 meter below ground level and
average water level is 6.0 to 7.0 meter below ground level in Trans Ganga area. Post monsoon
water level varies between 1.45 to 13.00 meter below ground level in Trans Ganga area. Water
level varies 5.00 to 6.00 meter below ground level in Trans Yamuna area.[ ]

The minimum level of underground water is noted as eight meters, but it is found that there exist
three localities in the city where the level is above eight meters (Saumya et al., 2014). However,
in the post-monsoon period now around a dozen localities are touched this level. According to
the data provided by the state ground water department, the rain in the months of August and
September substantially recharged the underground water in the district (2009), especially in the
areas which supports majority of population.

In the pre-monsoon period, only Transport Nagar, South Malaka and Rajapur region had water
level above the eight meters. On the other hand situation of ground water level in various
localities including Civil Lines (27.30 metres below the surface level), Jhalwa (24.30 mt) Katra
(24.6 mt), Prayag (22.5 mt) Bamraulli (21.7 mt) Kucthery (20.65 mt) Beli Hospital (19.35 mt)
64

etc had reached an alarming point due to exploitation. However, due to heavy rainfall recorded in
the monsoon of 2013 water level in Civil Lines now has been improved and stands at 24.6 metres
(still far away from minimum required level). Similarly, water table in Jhalwa, Sadar Tehsil
(Katra), Bamraulli and Beli Hospital has now improved and stands at 21.30, 23.35, 19.35 and
19.10 meters respectively. As per the data recorded on August 10 (2009), the water level in
almost all localities of the city water level has improved to around two meters. Ground water
level at Rajapur now stands at 2.40 meters, at Transport Nagar it is 3.40 meters and South
Malaka it is at 4.35 meters from ground level. The water level in Bairhana area recorded as 9.10
meters in June (2009) has now risen to 6 meters (Speaking on the improved water table,
executive engineer of state water ground department, Samvedi said the, give reference instead
full quote).

At Allahapur it has now improved to 6.90 meters (august) from10.20 meters in June (2008). At
Georgetown, it has improved to three meters and now stand at 6.60 meters. The water level at
Tagore Town has now improved by 7 meters as the water level in June was recorded at 10.15
meters and now it is 3.25 meters (August, 2007). At Katghar it is now 8.20 m, Ashok Nagar 12
m, Dhoomanganj 12.40 m, Newa 10.65 m, Kasari Masari 13.80 m, Daraganj 10.90 m,
Khuldabad 8.70 m and Rasoolabad 11.45 m etc. Similarly, there are many other localities where
underground water level has remarkably improved.

65

Fig. 3.2: Post-Monsoon groundwater level map of Allahabad city.

Fig. 3.3: Pre-Monsoon groundwater level map of Allahabad city.

66

Fig. 3.4: Seasonal Fluctuation map (2011)

Now compare the location of geotechnical data collection with water table. Location during pre
monsoon, monsoon and post monsoon in tabular form in table 3.2.
Depth of Water Table from Ground Level (m)
S. No.

Location/ Area
Pre Monsoon

3.6 Zonation of Study Area

67

Monsoon

Post Monsoon

After completing data collection from existing source and geotechnical investigation the study
area have been divided into five suitable zones on the basis of geotechnical properties viz. ZoneI, II, III, IV and V which is shown in table 3.3.
Table 3.3: Zone wise Details of Boreholes.
S. No. Zone no. Allotted Total Boreholes Total Datasets
1

Zone I

40

137

Zone II

38

110

Zone III

42

167

Zone IV

32

104

Zone V

36

126

Combined

189

644

Zone-I covered 5 km radial distance from Ganga river on Allahabad-Kanpur road which include
following locations Subedarganj, Transport Nagar, Neem Sarai, Jaintipur, Jhalwa, Bamrauli etc.
On the left side of Allahabad Kanpur Road when moving towards Kanpur clayey soil with low
compressibility upto a depth of 3.0 m are present in the top strata. After 3 m soil stratum was
found to be non-plastic silty soil, whereas right side of this road covered by non-plastic silty soil
(ML) and sandy soil (SM) which is near to Ganga river.
Zone-II is covered by river ganga on it northern side and it consists of important residential
areas which includes army cantonment , Munforgunj, Katra, Rajapur etc.
Zone III is covered by Yamuna river on its southern side and it is highly populated region of
Allahabad City. One of the prime location of this zone is Allahabad Railway station and Civil
Lines with residential areas such as Refugee colony, Tulsipur, Lukerganj etc.

68

Zone IV is another important area of Allahabad city with important educational institutions and
is highly populated. Important location include Teliarganj, Prayag Railway Station, Govindpur,
Tagore Town, Allahabad University etc. This Zone consists of railway line which divides this
zone into two parts.
Zone V forms the outskirts of Allahabad city with the confluence of Ganga Yamuna. Important
areas in this zone include Rambagh Railway station , Bariana, Madhwapur etc.

3.7 Data for Parametric Studies

Two seismic parameters i.e. maximum ground acceleration (amax) and earthquake magnitude (M)
are most pertinent to determine liquefaction potential of soil by conventional approach. Present
study seismic hazard assessment in terms of ground response analysis and liquefaction potential
has been carried out for the Allahabad city, India. PGA obtained from the analysis ranges from
0.063 to 1.5g (Naik et al., 2012) and Bhuj earthquake (2001) felt 0.35g PGA. To carry out thr
parametric study data of maximum ground acceleration varied from 0.15g, 0.25g and 0.35g
(where, g is acceleration due to gravity). As earlier explained that earthquake magnitude greater
than 5.0 may cause damage and there is no history about Allahabad. Hence in this work, it has
been earthquake magnitude parameter have been varied for 6.0, 7.0 and 8.0 for the assessment of
liquefaction potential are shown in table 3.4.
Table 3.4: Parametric Data for Water Table, PGA and Earthquake Magnitude.

Depth of water table from ground level (W)

(m)
Maximum Ground Acceleration (a max)
Earthquake magnitude on Richter scale (M)

69

0.15g

0.25g

0.35g

6.0

7.0

8.0

45 combinations were formed for parametric study to calculate CSR value by semi
empirical approaches for the specific depth of water table, maximum ground acceleration and
earthquake magnitude. Similarly, CSR values were obtained for different combination of depth
and earthquake magnitude.
On the basis of these datasets liquefaction potential of soils by conventional methods have been
determined which is extensively discussed in next chapter.

70

Fig 5.1: figure shows Allahabad city which is distributed in five zones.

71

Chapter 4
ASSESSMENT OF L.P. BY CONVENTIONAL AND COMPUTATIONAL METHODS

4.1 Background

Geotechnical professionals generally carryout subsurface soil exploration to evaluate the

liquefaction potential. The most common techniques to conduct soil exploration is standard
penetration test (SPT) which is widely used. In the previous chapter details of data collected
from various agencies have been described, perform SPT test at 31 locations for soil
investigation in all five zones to collect soil samples. These collected soil samples were tested to
complete the bore log chart for all 31 locations. In this chapter use of geotechnical data have
been carried out for the evaluation of liquefaction potential by conventional and computational
methods.

4.2 Assessment of Liquefaction Potential by Conventional Methods

Conventional procedure to estimation of liquefaction potential is based on SPT-N value

associated with information about the soil strata. In this chapter, liquefaction potential have been
expressed in terms of factor of safety (FS) which is defined as the ratio of cyclic resistance ratio
(CRR) and cyclic stress ratio (CSR). Cyclic resistance ratio (CRR) is expressed as the
liquefaction resistance of the soil and cyclic stress ratio (CSR) is the earthquake induced loading.
If the value of FS is less than 1 at a particular depth indicate liquefaction susceptibility at that
depth and vice-versa.
A number of semi empirical approaches to evaluate liquefaction potential have been developed
over the years. One of the commonly used in engineering practice is modified Seeds method and

72

another techniques are Idriss & Boulangar (I&B) method and Tokimatsu & Yoshimi (T&Y)
method. These methods follow certain protocols:

Estimation of the cyclic stress ratio (CSR) induced at various depths within the soil by
the earthquake.

Estimation of the cyclic resistance ratio (CRR) of the soil, i.e. the cyclic shear stress ratio
which is required to cause initial liquefaction of the soil.

Evaluation of factor of safety against liquefaction potential of in situ soils.

To assess the liquefaction potential by these methods following steps were adopted for aforesaid
semi empirical techniques but some steps are common. These are

i.

Calculation of overburden pressure

ii.

Correction of SPT-N value

4.2.1 Calculation of Overburden Pressure

This is one of the most important parameter for the assessment of liquefaction potential of soil by
conventional methods. Bulk unit weight, natural moisture content and dry unit weight are
mentioned in bore log chart for each type of soil encountered in one borehole. On the basis of
these datasets overburden pressure and effective overburden pressure of the soil can be
determined at any depth by adopting general technique. Equations used for calculating
overburden pressure is
Effective overburden pressure = u. (4.1)
Where, = Total pressure
and u = pore water pressure
73

4.2.2 Correction for SPT-N value

To determine liquefaction potential by conventional methods, corrected SPT-N value is required.

According to Indian Standards (IS: 2131-1981) [24] one correction is due to overburden pressure
and second is due to dilatancy. Corrected SPT-N values are required for the assessment of
liquefaction potential by semi empirical approach. Hence standard procedure for correcting
SPT-N value was carried out.

Correction for overburden pressure: N- value obtained from SPT test can be corrected by the
following equation:
N1=C NN

(4.1)

Where, CN is correction factor obtained directly from the graph (figure 4.1) given in Indian
Standard Code (IS: 3121-1981).

Fig. 4.1: Correction due to overburden pressure

Correction factor can also be calculate from the given below

74

CN=0.77log10

(4.2)

Where, p is effective overburden pressure in kN/m2 [Varghese P. C].

Dilatancy Correction: The corrected N values obtained after overburden pressure, correction can
be corrected for dilatancy if the stratum consist of fine sand and silt below water table having
corrected greater N-value (N1) than 15 by following equation [O. Gijnaydin, 2009]:
Nc = 15 + 0.5(N1 -15)

(4.3)

After doing both the corrections SPT-N value have been used for assessment of liquefaction
potential of soil by semi empirical approach.
4.2.3 Estimation of Cyclic Stress Ratio (CSR)
Estimation of cyclic stress ratio is based on following assumptions
i.

Ground surface is level

ii.

Soil column is assumed to have unit weight

iii.

Length of soil column move horizontally

iv.

Soil column is assumed as a rigid body

In this work, Cyclic Stress Ratio (CSR) has been estimated by three different conventional
approaches.
i.

Modified Seeds Method [ ]

(

ii.

)(

By T&Y Method

75

CSR =
iii.

By I&B Method
(

Here
In this equation depth reduction factor rd varies with the depth from ground surface which
comes from the standard charts. A standard chart depends upon function of depth at a particular
earthquake magnitude.
4.2.4 Estimation of Cyclic Resistance Ratio (CRR)
Cyclic resistance ratio represents liquefaction resistance of in situ soil. Standard penetration test
data were used for the calculation of cyclic resistance ratio.
Cyclic Stress Ratio (CRR) have been estimated by conventional equation as given below
i.

Modified Seeds Method

ii.

By T&Y Method

[
iii.

) ]

By I&B Method

76

All the silent points and parameters are discussed in the chapter 2.

4.2.5 Estimation of Factor of Safety (FS) against Liquefaction

After estimating CSR and CRR liquefaction potential of soil as factor of safety FS i.e. can be
evaluated by conventional relation. If the cyclic stress ratio (CSR) is greater than cyclic
resistance ratio (CRR) of in situ soil then liquefaction will occur otherwise liquefaction may not
occur at that place. Higher the factor of safety soil is more resistive against liquefaction. Soil
having factor of safety slightly greater than 1 can also liquefy [ ].
Table 4.1: Value of CSR, CRR and FS by empirical methods calculated for one borehole at water table depth is at
ground level, horizontal acceleration 0.35g and earthquake magnitude 8.0.

2.
3.
4.
5.
6.
7.
8.

1.5

3.0

4.5

12

6.0

12

7.5

17

9.0

23

10.5

29

12.0

35

ML

SM

CL

0m-2.25m

SPT
Value

2.25m-9.50m

1.

Depth
(m)

9.5m-12m

Sl.
No.

Soil Profile &

Classification

Calculation of
Liquefaction Potential
by Modified Seeds
method

Calculation of
Liquefaction Potential
by I & B Method

Calculation of
Liquefaction Potential by
T & Y Method

CRR

CSR

FS

CRR

CSR

FS

CRR

CSR

FS

0.222

0.060

0.271

0.118

0.573

0.206

0.180

0.530

0.340

0.220

0.115

0.524

0.098

0.531

0.185

0.137

0.483

0.283

0.217

0.105

0.482

0.156

0.525

0.297

0.148

0.471

0.313

0.214

0.136

0.636

0.148

0.519

0.285

0.142

0.460

0.308

0.212

0.154

0.726

0.174

0.512

0.340

0.145

0.449

0.323

0.225

0.100

0.446

0.206

0.504

0.408

0.148

0.437

0.340

---

---

---

---

---

---

---

---

---

---

---

---

---

---

---

---

---

---

Hence for any there is 45 (533) sets of CRR & CSR values and for each set, liquefaction
potential have been obtained by one method. Total 644 datasets were used to analyse for chances
of liquefaction.

77

Factor of safety (FS) estimated by semi-empirical methods for all 189 borehole data i.e. 644
datasets with the variation of depth of water table; earthquake magnitude and horizontal
acceleration are presented zone wise in annexure B. The estimated FS were used for
development of model by computational models.
4.3 Development of Liquefaction Potential models by Computational Methods
Parametric study for the parameters like depth of water table, earthquake magnitude and
horizontal acceleration have been varied and liquefaction potential for the same have been
estimated by conventional method as shown in table..
To develop the ANN, ANFIS and MLR models total 644 datasets were collected in terms of
input and output values of the models for Allahabad city. Out of these 644 datasets here we
reserve 115 datasets for validation of network. Hence to develop the model 529 datasets were
used for all 5 zones. Zone wise details of datasets used for the development models are shown in
Table 4.2.
Table 4.2: Details of datasets used for development and validation of models.

Sl.
No.

Zone
Used

Boreholes no.
for validation

Number of Boreholes
Data used for Testing

Number of Datasets
reserved for

Training

Validation

Training

Validation

Zone I

4, 8, 12, 19, 21,

24, 34 and 38

32

08

110

27

Zone II

2, 5, 9, 19, 25 and
37

32

06

93

17

Zone III

2, 4, 7, 9, 18, 25,
27 and 39

34

08

135

32

Zone IV

3, 13, 18, 23, 27

and 32

26

06

85

19

78

Zone V

Combined

4, 14, 27 and 36
---------

32
32/189

04

106

20

115

529

To develop the models value of liquefaction potential (FS) was used as target value. Data
through liquefaction potential (FS) is the ratio of CRR/CSR. But in some cases this ratio found to
be too high like >50, number of these FS is less than 2% of total datasets. Those cases in which
this value (CRR/CSR ratio) outcomes too high, we restrict the FS value to 1.50.

4.3.1 Selection of Input Variables:

Selection of input parameters is one of the most important steps for developing any model. To
develop the model input datasets were selected from the collected data as development of the
model is based on available datasets which can be used directly for the assessment of
liquefaction potential. To select the input parameters 75 numbers of datasets from collected
datasets have been selected randomly. Initially two pertinent input variables i.e. SPT-N value (N)
and depth (z) were selected for the development of the computational models to predict
liquefaction potential. Initially using two inputs, no. of input variable were increased one by one
to develop computational model. Increase in input parameters minimizes the mean square error
of the models. Subsequently, increase in input variables leads to development of optimum
liquefaction potential model. Input variables were increased up to eight comprising of SPT-N
value (N), depth (d), bulk unit weight (t), particle size finer than 0.075 mm (D0.075), natural/field
moisture content (wf), particle size finer than 2.00 mm (D2.0), particle size finer than 0.002 mm
(D0.002) and angle of internal friction (). The range of values for input parameters for developing
the models are shown in table 4.3.

79

Table 4.3: Ranges of Input Parameters.

Input Parameters

Ranges

SPT-N value (N)

0-50

depth (m)

0-12

Bulk unit weight (t)

1.31-2.39

Particle finer than 0.075 mm (%) (D0.075)

18.34-99.69

Natural water content (wn)

1.16-43.9

Particle finer than 2.0 mm (%) (D2.0)

69.92-100

Particle finer than 0.002 mm (%) (D0.002)

Angle of internal friction ()

0-15
14-42.6

4.3.2 Development of ANN model:

Considering optimized input variables, artificial neural network (ANN) models were developed
and validated. Out of 189 number of boreholes, 157 number of borehole datasets were used for
training and testing whereas 32 number of borehole datasets were used for validating the ANN
models of different Zones. To develop the ANN model transfer function was used for all the
cases is given in equation 4.4. Range of tan-sigmoid transfer function is -1 to 1. The input &
output datasets used for ANN models development were normalized range between -1 to 1 using
the following equation (4.4).
(4.4)
Where,

= Normalized value.
= data which has to be normalized.
= minimum value of data.

and,

= maximum value of data.

80

To identify different combinations with its fundamental attributes a coding method was used for
different networks architecture, as MZWXAY where, MZ denotes earthquake magnitude, WX
denotes depth of water table and AY denotes maximum horizontal acceleration. The predicted
values of liquefaction potential as FS by developed models are discussed in subsequent heading.

ANN tool in MATLAB software was used for all operations in which networks were trained
with single or double hidden layers with varying numbers of neurons from 2 to 20. Liquefaction
potential evaluated from parametric study with the variation of depth of water table, maximum
horizontal acceleration and earthquake magnitude was used as output variable whereas input
variable were borehole data. Considering every condition and case of parameters and network
architecture ANN models were developed for all the zones. In case of hidden layers neurons
were varied up to 20. ANN model is simply denoted as NX, which was varied from N1 to N90
which are shown in Table 4.4. This table shows that for the development of one optimized
model. 90 models with different hidden neuron and layers were developed. Network architecture
in Table 4.4 resembles hidden layers, output layers and number of neurons.
Earlier it was mentioned that, the datasets were divided in five suitable zones which were based
on similar type of soil present in that zone. ANN models were developed for all the zones
initially. Then ANN models were developed for combined zones. In one zone, for 45
combinations of parameters and 90 cases of architecture 4050 models (i.e. 9045) models were
developed. Out of 4050 developed models, ten models based on minimum average absolute error
(AAE), minimum root mean square error (RMSE), maximum R-squared value (R2) value were
selected for further study. In second step, we select only three models out of ten models were
selected on the basis of better predictive estimate of LP which prediction was better than others
from reserved datasets. Similar steps were adopted for all individual zones and combined zones
81

for empirical method. Total ANN models developed for one zone using the datasets of three
empirical method are 12150 (i.e. 4050 3) and ANN models developed for all zones and
combined zones are 72900 (i.e. 405036).

Table4.4: Details of ANN Models and Network Architecture

ANN
Model

Network
Architecture

ANN
Model

Network
Architecture

ANN
Model

Network
Architecture

N1

4-1

N31

8-10-1

N61

14-16-1

N2

6-1

N32

8-12-1

N62

14-18-1

N3

8-1

N33

8-14-1

N63

14-20-1

N4

10-1

N34

8-16-1

N64

16-04-1

N5

12-1

N35

8-18-1

N65

16-06-1

N6

14-1

N36

8-20-1

N66

16-08-1

N7

16-1

N37

10-04-1

N67

16-10-1

N8

18-1

N38

10-06-1

N68

16-12-1

N9

20-1

N39

10-08-1

N69

16-14-1

N10

4-04-1

N40

10-10-1

N70

16-16-1

N11

4-06-1

N41

10-12-1

N71

16-18-1

N12

4-08-1

N42

10-14-1

N72

16-20-1

N13

4-10-1

N43

10-16-1

N73

18-04-1

N14

4-12-1

N44

10-18-1

N74

18-06-1

N15

4-14-1

N45

10-20-1

N75

18-08-1

N16

4-16-1

N46

12-04-1

N76

18-10-1

82

N17

4-18-1

N47

12-06-1

N77

18-12-1

N18

4-20-1

N48

12-08-1

N78

18-14-1

N19

6-04-1

N49

12-10-1

N79

18-16-1

N20

6-06-1

N50

12-12-1

N80

18-18-1

N21

6-08-1

N51

12-14-1

N81

18-20-1

N22

6-10-1

N52

12-16-1

N82

20-04-1

N23

6-12-1

N53

12-18-1

N83

20-06-1

N24

6-14-1

N54

12-20-1

N84

20-08-1

N25

6-16-1

N55

14-04-1

N85

20-10-1

N26

6-18-1

N56

14-06-1

N86

20-12-1

N27

6-20-1

N57

14-08-1

N87

20-14-1

N28

8-04-1

N58

14-10-1

N88

20-16-1

N29

8-06-1

N59

14-12-1

N89

20-18-1

N30

8-08-1

N60

14-14-1

N90

20-20-1

Hence, we developed total 72,900 models (24,3003) and selected only 54 models which were
having better predictive capability. Results obtained from these models were discussed in next
chapter.

4.3.3 Development of ANFIS model:

To develop adaptive neuro fuzzy inference system (ANFIS) models, ANFIS tool in MATLAB
software was used for all operations in which networks were trained for up to 75 numbers of
epochs and three membership functions were allotted for each of the eight input parameters. A
grid partitioning method and triangular membership function for input variables were used to

83

generate the FIS, whereas linear membership function was used for the target variable. Hybrid
optimization technique was used for training the FIS.
In the line of ANN models, ANFIS models were also developed for the input and output
variables. In this case, also various combinations were tried and tested in each zone for
Simplified procedure. Selection of developed ANFIS models were based upon the minimum
RMSE and maximum COD value obtained by the similar models for comparative study. Hence
in this thesis we developed total 180 ANFIS models. Out of these 180 developed models we
select only 54 better predictive models which means the developed models shows minimum
RMSE and maximum COD when we provided the reserved datasets.

4.3.4 Development of MLR models:

Regression analysis was also carried out to establish the functional relation between
independent and dependent variables to develop MLR models. These equations were developed
for all forty-five combinations of parameters, i.e. depth of water table, earthquake magnitude and
horizontal acceleration. The generalized form of MLR equations to predict liquefaction potential
(LP) is as follows
LP = A1+ A2N + A3d + A4 t + A5 D0.075 + A6wn+ A7 D2.0 + A8 D0.002 + A9

(4.5)

Where, A1 to A9 represents the constant of the equation with the adopted variables i.e. SPT-N
value (N), depth (d), bulk unit weight (t), percentage finer then 0.075 (D0.075), moisture content
(wn), percentage finer then 2.0 (D2.0), percentage finer then 0.002 (D0.002) and angle of internal
friction ().

84

In this study, we developed MLR equation for only 54 better predictive models which comes
from ANN model as well as ANFIS model. The all developed models will be detailed discussed
in the next chapter.

85

Chapter 5

RESULTS AND DISCUSSION

5.1 General
The assessment of liquefaction potential (LP) by semi empirical methods have been discussed
in detailed with results. Critical study on the basis of parametric variations gives a very
significant outcome. The collected details with obtained LP by these methods were adopted
for the development of computational methods. The developed models using ANN, ANFIS
and MLR have been tested and validated with the reserve datasets, which assess the
predictive capability of different models. In this chapter results of these models have been
discussed in details.

5.2 Semi Empirical Approach

5.2.1 Liquefaction Potential by Empirical Methods for Zone-I
Liquefaction potential for zone-I calculated from modified Seeds method for all 45
combinations of MxAyWz are summarized in annexure-B. A graph between percentages (%)
of chances of liquefaction vs different 45 combinations by modified Seeds method is
illustrated in Fig. 5.1. It can be seen from the figure that the maximum liquefaction occurs for
the combination of models M8A0.35W2, M8A0.35W4 & M8A0.35W6 i.e. 87.591%, 83.742% &
79.562% respectively. Model M6A0.15W0, M6A0.15W6 & M6A0.15W8 do not show any chances
of liquefaction occurrences. It is customary to mention the least prone and most susceptible
models. Since most of the further discussion is based on most probable liquefaction model
combination. Liquefaction occurrence for most susceptible models as mentioned is illustrated
in Fig. 5.1.

86

M6A0.15W0
M6A0.15W2
M6A0.15W4
M6A0.15W6
M6A0.15W8
M6A0.25W0
M6A0.25W2
M6A0.25W4
M6A0.25W6
M6A0.25W8
M6A0.35W0
M6A0.35W2
M6A0.35W4
M6A0.35W6
M6A0.35W8
M7A0.15W0
M7A0.15W2
M7A0.15W4
M7A0.15W6
M7A0.15W8
M7A0.25W0
M7A0.25W2
M7A0.25W4
M7A0.25W6
M7A0.25W8
M7A0.35W0
M7A0.35W2
M7A0.35W4
M7A0.35W6
M7A0.35W8
M8A0.15W0
M8A0.15W2
M8A0.15W4
M8A0.15W6
M8A0.15W8
M8A0.25W0
M8A0.25W2
M8A0.25W4
M8A0.25W6
M8A0.25W8
M8A0.35W0
M8A0.35W2
M8A0.35W4
M8A0.35W6
M8A0.35W8

Chances of Soil Liquefy (%)

100.000

90.000

Modified Seed's Method for Zone - I

Av. Chances of Soil Liquefaction 34.874%

80.000

70.000

60.000

50.000

40.000

30.000

20.000

10.000

0.000

Types of Combinations

Fig. 5.1: Chances of Soil Liquefaction under Different Conditions by Modified Seeds Method for zone I in Percentage.

87

90.000

M6A0.15W0
M6A0.15W2
M6A0.15W4
M6A0.15W6
M6A0.15W8
M6A0.25W0
M6A0.25W2
M6A0.25W4
M6A0.25W6
M6A0.25W8
M6A0.35W0
M6A0.35W2
M6A0.35W4
M6A0.35W6
M6A0.35W8
M7A0.15W0
M7A0.15W2
M7A0.15W4
M7A0.15W6
M7A0.15W8
M7A0.25W0
M7A0.25W2
M7A0.25W4
M7A0.25W6
M7A0.25W8
M7A0.35W0
M7A0.35W2
M7A0.35W4
M7A0.35W6
M7A0.35W8
M8A0.15W0
M8A0.15W2
M8A0.15W4
M8A0.15W6
M8A0.15W8
M8A0.25W0
M8A0.25W2
M8A0.25W4
M8A0.25W6
M8A0.25W8
M8A0.35W0
M8A0.35W2
M8A0.35W4
M8A0.35W6
M8A0.35W8

Chances of Soil Liquefy (%)

100.000

I & B Method for Zone - I

Av. Chances of Soil Liquefaction 42.320%

80.000

70.000

60.000

50.000

40.000

30.000

20.000

10.000

0.000

Types of Combinations

Fig. 5.2: Chances of Soil Liquefaction under Different Conditions by I&B Method for zone I in Percentage.

88

100.000

M6A0.15W0
M6A0.15W2
M6A0.15W4
M6A0.15W6
M6A0.15W8
M6A0.25W0
M6A0.25W2
M6A0.25W4
M6A0.25W6
M6A0.25W8
M6A0.35W0
M6A0.35W2
M6A0.35W4
M6A0.35W6
M6A0.35W8
M7A0.15W0
M7A0.15W2
M7A0.15W4
M7A0.15W6
M7A0.15W8
M7A0.25W0
M7A0.25W2
M7A0.25W4
M7A0.25W6
M7A0.25W8
M7A0.35W0
M7A0.35W2
M7A0.35W4
M7A0.35W6
M7A0.35W8
M8A0.15W0
M8A0.15W2
M8A0.15W4
M8A0.15W6
M8A0.15W8
M8A0.25W0
M8A0.25W2
M8A0.25W4
M8A0.25W6
M8A0.25W8
M8A0.35W0
M8A0.35W2
M8A0.35W4
M8A0.35W6
M8A0.35W8

Chances of Soil Liquefy (%)

T&Y Method for Zone - I

Av. Chances of Soil Liquefaction 41.590%

90.000

80.000

70.000

60.000

50.000

40.000

30.000

20.000

10.000

0.000

Types of Combinations

Fig. 5.3: Chances of Soil Liquefaction under Different Conditions by T&Y Method for zone I in Percentage.

89

From the observation of results, it is found that the model M7A0.35W0, M8A0.35W0 &
M8A0.35W2 gave highest number of liquefaction occurrences in zone-I. Total 27 datasets were
used to validate empirical models. 23 cases were found prone for liquefaction in the above
three mentioned combinations. Liquefaction did not occur for remaining 4 cases of validate
datasets (Fig. 5.2)
Model M6A0.15W4, M6A0.15W6 & M6A0.15W8 did not show any case of liquefaction
occurrences. It is customary to mention here that the above discussion and subsequent
discussion is based on the analysis of validated datasets as it is not possible to compare the
results based on training datasets. Since they are great in numbers & validation datasets itself
give an impression of overall performance of models.
The third semi empirical method is T&Y method by which liquefaction potential also
calculated. Using the same combination of MxAyWz the liquefaction occurrences from 137
data sets was calculated and illustrated in Appendix (B). It can be seen from the bar graph
that most combination of M-A-W are susceptible for liquefaction. However, most susceptible
combinations giving more than 50% chances of liquefaction were identified as M8A0.35W0,
M8A0.35W2 & M7A0.35W0.
Influence of Earthquake Magnitude:To determine the liquefaction potential initially three conventional methods, i.e. modified
Seeds method, Idriss and Boulangar (I&B) method & Tokimatsu and Yoshimi (T&Y)
method were adopted considering parametric variation like magnitude, ground acceleration
and depth of water table from ground level. In this section, influence of earthquake
magnitude on chances of liquefaction has been investigated for zone I by keeping other
parameters constant. These variations can be seen in Fig. 5.1, 5.2 and 5.3. It can be seen from
the table with the variation of magnitude itself, that earthquake magnitude play important and
significant role in determination of liquefaction potential.

90

It can be seen from Fig. 5.4 that influences of magnitude significantly affect the percentage
rise in liquefaction as magnitude increases.

The chances of soil liquefaction between

magnitude 6 and 7 have been increases by 6.569% whereas between magnitude 7 and 8
chances of liquefaction increases by 9.489%. This difference depicts the nonlinear behavior
but graphical representation between liquefaction potential and magnitude can be linear with
straight line fit curve (Fig. 5.4).
Fig. 5.4 also shows that there is no chance of liquefaction at earthquake magnitude 6.0 when
water table is at ground level and horizontal acceleration is 0.15g but increasing in
earthquake magnitude keeping other parameter constant chances of soil liquefaction

Chances of Soil Liquify (%)

increases.
18.0

Modified Seed's method

16.0

W=0.0 m & A=0.15g

14.0
12.0
10.0
8.0
6.0
4.0
2.0
0.0
5

Variation of Earthquake Magnitude

Chances of Soil Liquify (%)

Fig. 5.4: Influence of Magnitude for W.T. = 0.0m and A=0.15g by modified Seeds method for zone I.
80.0

I&B method

70.0

Depth of Water Table is at ground level and Ground Acceleration is

A=0.15g

60.0
50.0
40.0
30.0
20.0
10.0
0.0
5

7
8
Variation of Earthquake Magnitude

Fig. 5.5: Influence of Magnitude with respect to soil liquefy (%) for W.T. = 0.0m and A=0.15g by I&B method
for zone I.

91

The influence of magnitude by I&B method (Fig. 5.5) despite of the dominating tendency of
magnitude is clearly visible though the intensity of liquefaction have been reduced but
increasing trend shows the chances of liquefaction w.r.t increase in magnitude for zone-1.
Further, the graphical representation based on T&Y method varying water table and varying
ground acceleration represent that increase in magnitude always lead to increase the chances
of liquefaction. A substantial finding have been observed that, the variation in liquefaction
potential was too less i.e. 2.189% for depth of water table is at ground level and earthquake
magnitude is 0.35g with the varying of earthquake magnitude from 6.0 to 7.0. It has also been
observed that, there is no chances of liquefaction for variation of magnitude with increase in

98.0

Chances of Soil Liquify (%)

Chances of Soil Liquify (%)

water table beyond 2.0 m at constant ground acceleration 0.15g (Fig.5.6-a).

T&Y method

97.5

at W.T.= 0.0 m and A=0.35g

97.0
96.5
96.0
95.5
95.0
94.5
5

6
7
8
Variation of Earthquake Magnitude

1.0

T&Y method
at W.T.=4.0 m and A=0.15g

0.8
0.6
0.4
0.2
0.0
5

Variation of Earthquake Magnitude

(a)

(b)

Fig. 5.6: Influence of earthquake magnitude for (a) for A=0.35g and W.T.= 0.0 m (b) for A=0.15g and W.T.=
4.0 m, by T&Y method for zone I.

One is the variation of chances of liquefaction increases with the increase in earthquake
magnitude and decrease in depth of water table and another nature of results show that there
are no variation shows due to increase of earthquake magnitude with the increase in
earthquake magnitude and decrease in depth of water table (Fig.5.6-b).
It can also be shown that for varying water table and constant ground acceleration represent
that increase in magnitude always lead to increase the chances of liquefaction. A significant
finding have been observed that is the decrease in water table and increase in magnitude
92

resemble a unique relationship as increase in magnitude is dominating a unique relationship

as increase magnitude is dominating the tendency of liquefaction in case of modified seed`s
method.

Influence of Ground Acceleration:In line of the above discussion the increase in acceleration also shows the dominating
tendency of the acceleration as it directly influences the chances of liquefaction. It could be
seen from Fig. 5.7 (a) that for magnitude 6 and water table at ground level the chances of
liquefaction between two points i.e. point 0.15g and 0.25g is 7.299% and the chances of
liquefaction further increase between 0.25g and 0.35g i.e. 16.789%. The variation with the
chances of liquefaction highlights that increase in acceleration will increase in case of
increase in the chances of liquefaction where water table also plays one important role. The

30.0

Modified Seed's method

Chances of Soil Liquify (%)

Chances of Soil Liquify (%)

regular/linear behaviour can`t be depicted by modified seeds approach.

at M=6.0 and W.T.=0.0m

25.0
20.0
15.0
10.0
5.0
0.0
0.1

0.2

0.3

0.4

90.0
80.0
70.0
60.0
50.0
40.0
30.0
20.0
10.0
0.0

Modified Seed's method

at M=8.0 and W.T.=0.0 m

0.1
0.2
0.3
0.4
Variation of Ground Acceleration (g)

Variation of Ground Acceleration (g)

(a)

(b)

Fig. 5.7: Influence of Ground Acceleration for W.T. = 0.0m and (a) for M=6.0, (b) for M=8.0 by modified
Seeds method for zone I.

I & B approach for chances of liquefaction considering the variation of acceleration shows
the different trend between 0.15g and 0.25g for magnitude 8 and water table at 2.0 m the
chances of liquefaction is 40.875% between two acceleration points where as the difference
b/w the consecutive acceleration is i.e. 0.25g and 0.35g is 8.03% (Fig. 5.8). In this case,
93

chances of liquefaction increasing with the increase in acceleration but the increase in
percentage of liquefaction between last two points is decreasing drastically compared to
initial acceleration points. It shows the typical behaviour of this approach.
90.0

I&B method
at W.T.= 2.0 m at M=8.0

Chances of Soil Liquify (%)

80.0
70.0
60.0
50.0
40.0
30.0
20.0
0.1

0.2
0.3
0.4
Variation of Ground Acceleration (g)
Fig. 5.8: Influence of Ground Acceleration for W.T. = 2.0m and M=8.0 by I & B method for zone I.

Further, it may be observed from the Fig. 5.2 the chances of liquefaction in case of magnitude
with increase in ground acceleration is consistent compare to magnitude 7 & 8 though

Chances of Soil Liquify (%)

intensity of liquefaction increases with increasing liquefaction.

90.0
80.0
70.0
60.0
50.0
40.0
30.0
20.0
10.0
0.0

T&Y method
at W.T.= 4.0 m and M=8.0

0.1

0.2

0.3

0.4

Variation of Ground Acceleration (g)

Fig. 5.9: Influence of Ground Acceleration for W.T. = 4.0m and M=8.0 by T&Y method for zone I.

T&Y approach for chances of liquefaction considering the variation of acceleration shows
(Fig. 5.9) the similar trend of I&B method. In this case, chances of liquefaction increasing
with the increase in acceleration but the increase in percentage of liquefaction b/w two points

94

is reducing. Chances of liquefaction between consecutive points are 57.664% and 26.278%
for earthquake magnitude 8.0 and depth of water table 4.0 m.
Results obtained from T&Y method shows that the chances of liquefaction increases with the
increase in earthquake magnitude and decrease in depth of water table.

Influence of Water Table:Presence of ground water table plays an important role in estimation of liquefaction potential.
The influence of ground water table in determination of liquefaction potential by modified
Seeds method considering constant horizontal ground acceleration and earthquake
magnitude is shown in Fig. 5.10. If water table is at ground level, chances of liquefaction is
78% but in case of 2 m depth chances of liquefaction increased by 10% i.e. 88%. This
variation could be attributed to density of the soil at different level. This should be
specifically mentioned that the chances of liquefaction are very high when water table exists
2.0 m below the ground level. Therefore, shallow foundations at the depth of 2.0 m will be
susceptible to liquefaction leading to differential settlement. To avoid such condition antiliquefaction measures should be adopted.
As water table reducing further, there is a decreasing trend with respect to chances of
liquefaction. Therefore, depth of water table from ground level plays a significant role in
liquefaction.
92.0
90.0

Chances of Soil Liquify (%)

Chances of Soil Liquify (%)

Modified Seed's method

at M=8.0 A = 0.35g

88.0
86.0
84.0
82.0
80.0
78.0
0

2
4
6
Variation of Water Table (m)

85.0

Modified Seed's method

80.0

at M= 8.0 and A = 0.25g

75.0
70.0
65.0
60.0
55.0
50.0

95

Variation of Water Table (m)

(a)

(b)

Fig. 5.10: Influence of Depth of Water Table for (a) A=0.35g and M=8.0 (b) A=0.25g and M=8.0 by modified
Seeds method for zone I.

It is apparent from Fig. 5.11 that the chances of liquefaction in soil decrease with the decrease
in water table. The falling trends were found for all types of combinations except for
magnitude 8 and acceleration 0.35g by I&B method. In this case chances of liquefaction
potential was constant in between depth of water table 0.0 m to 2.0 m after 2.0m depth
chances of liquefaction decreasing for zone I. Though chances of liquefaction is 83.212%
when water table is at ground level for similar constant.
I&B method

70.0

Chances of Soil Liquify (%)

Chances of Soil Liquify (%)

80.0

at M=8.0 and A = 0.15g

60.0
50.0
40.0
30.0
20.0
10.0
0.0
0

83.5
83.0
82.5
82.0
81.5
81.0
80.5
80.0
79.5
79.0
78.5

2
4
6
8
Variation of Depth of Water Table (m)

I&B method
at M= 8.0 and A = 0.35g

(a)

2
4
6
8
Variation of Depth of Water Table (m)

(b)

98.0
96.0
94.0
92.0
90.0
88.0
86.0
84.0
82.0
80.0

T&Y method

Chances of Soil Liquify (%)

Chances of Soil Liquify (%)

Fig. 5.11: Influence of Depth of Water Table for (a) A=0.15g and M=8.0 (b) A=0.35g and M=8.0 by I&B
method for zone I.

at M=8.0 and A = 0.35g

2
4
6
8
Variation of Depth of Water Table (m)

70.0

T&Y method

60.0

at M = 8.0 and A = 0.15g

50.0
40.0
30.0
20.0
10.0
0.0
0

(a)

2
4
6
8
Variation of Depth of Water Table (m)

(b)

96

Fig. 5.12: Influence of Depth of Water Table for (a) A=0.35g and M=8.0 (b) A=0.15g and M=8.0 by T&Y
method for zone I.

The trend of variation of chances of liquefaction potential of soil by T&Y method is similar
to I&B method (Fig. 5.12 a & b). Maximum chances of liquefaction potential occur when
water table varies up to ground level for the combination of ground acceleration A=0.35g
and earthquake magnitude M= 8.0 i.e. 97.080%. Sudden change in the chances of
liquefaction can be seen for M=8.0 and A=0.15g at water table 4.0 m which can happen due
to decreases in acceleration.
5.2.2 Liquefaction Potential by Empirical Methods for Zone-II
Liquefaction potential for zone II calculated by different empirical formula like modified
Seeds method, I&B method and T&Y method. Zone II comprises total 38 borehole data
i.e. 110 datasets. We calculated liquefaction potential for all 45 combinations of MxWyAz are
summarized in Annexure B. A graph has been plotted in which X-axis showed that the
different 1-45 combinations and Y-axis showed that the percentages of chances of soil
liquefy.
Fig. 15.3 showed that the graph plotted for zone II by modified Seeds method. On the basis
of graph it can be easily seen that the combination M6 A0.15 W0 ; M6 A0.15 W6 & M6 A0.15 W8
did not show any liquefaction whereas at earthquake magnitude 7.0 only one condition show
that the liquefaction potential > 70% i.e. at M7 A0.35 W4. Whereas at earthquake magnitude
8.0 and horizontal acceleration 0.35g showed chances of soil liquefy > 80% which are more
prone with respect to others.
Similarly fig. 5.14 showed that the % of chances of soil liquefy by different combinations for
I & B method. In this method total 5 combinations i.e. M6 A0.15 W4 ; M6 A0.15 W6 ; M6 A0.15
W8; M7 A0.15 W6 and M7 A0.15 W8 did not show chances of soil liquefy. In modified Seeds
method only 5 combinations showed the soil liquefy>80% but in I & B method there are nine

97

combinations showed that the chances of soil liquefy by greater than 80% i.e. M6 A0.35 W0 ;
M7 A0.35 W0 ; M7 A0.35 W2 ; M8 A0.25 W0 ; M8 A0.35 W0 ; M8 A0.35 W2 ; M8 A0.35 W4 ; M8 A0.35
W6 and M8 A0.35 W8. Maximum value of chance of soil liquefy is 88.182% for M8 A0.35 W0.

98

M6A0.15W0
M6A0.15W2
M6A0.15W4
M6A0.15W6
M6A0.15W8
M6A0.25W0
M6A0.25W2
M6A0.25W4
M6A0.25W6
M6A0.25W8
M6A0.35W0
M6A0.35W2
M6A0.35W4
M6A0.35W6
M6A0.35W8
M7A0.15W0
M7A0.15W2
M7A0.15W4
M7A0.15W6
M7A0.15W8
M7A0.25W0
M7A0.25W2
M7A0.25W4
M7A0.25W6
M7A0.25W8
M7A0.35W0
M7A0.35W2
M7A0.35W4
M7A0.35W6
M7A0.35W8
M8A0.15W0
M8A0.15W2
M8A0.15W4
M8A0.15W6
M8A0.15W8
M8A0.25W0
M8A0.25W2
M8A0.25W4
M8A0.25W6
M8A0.25W8
M8A0.35W0
M8A0.35W2
M8A0.35W4
M8A0.35W6
M8A0.35W8

Chances of Soil Liquefy(%)

100.00

Modified Seed's Method for Zone II

90.00

Av. Chances of Soil Liquefaction 25.293%

80.00

70.00

60.00

50.00

40.00

30.00

20.00

10.00

0.00

Types of Combinations

Fig. 5.13: Chances of Soil liquefaction under Different Conditions by Modified Seeds Method for zone II in Percentage.

99

M6A0.15W0
M6A0.15W2
M6A0.15W4
M6A0.15W6
M6A0.15W8
M6A0.25W0
M6A0.25W2
M6A0.25W4
M6A0.25W6
M6A0.25W8
M6A0.35W0
M6A0.35W2
M6A0.35W4
M6A0.35W6
M6A0.35W8
M7A0.15W0
M7A0.15W2
M7A0.15W4
M7A0.15W6
M7A0.15W8
M7A0.25W0
M7A0.25W2
M7A0.25W4
M7A0.25W6
M7A0.25W8
M7A0.35W0
M7A0.35W2
M7A0.35W4
M7A0.35W6
M7A0.35W8
M8A0.15W0
M8A0.15W2
M8A0.15W4
M8A0.15W6
M8A0.15W8
M8A0.25W0
M8A0.25W2
M8A0.25W4
M8A0.25W6
M8A0.25W8
M8A0.35W0
M8A0.35W2
M8A0.35W4
M8A0.35W6
M8A0.35W8

Chances of Soil Liquefy(%)

100.000

I & B Method for Zone II

90.000

Av. Chances of Soil Liquefy: 35.475%

80.000

70.000

60.000

50.000

40.000

30.000

20.000

10.000

0.000

Types of Combinations

Fig. 5.14: Chances of Soil liquefaction under Different Conditions by I&B Method for zone II in Percentage.

100

90

M6A0.15W0
M6A0.15W2
M6A0.15W4
M6A0.15W6
M6A0.15W8
M6A0.25W0
M6A0.25W2
M6A0.25W4
M6A0.25W6
M6A0.25W8
M6A0.35W0
M6A0.35W2
M6A0.35W4
M6A0.35W6
M6A0.35W8
M7A0.15W0
M7A0.15W2
M7A0.15W4
M7A0.15W6
M7A0.15W8
M7A0.25W0
M7A0.25W2
M7A0.25W4
M7A0.25W6
M7A0.25W8
M7A0.35W0
M7A0.35W2
M7A0.35W4
M7A0.35W6
M7A0.35W8
M8A0.15W0
M8A0.15W2
M8A0.15W4
M8A0.15W6
M8A0.15W8
M8A0.25W0
M8A0.25W2
M8A0.25W4
M8A0.25W6
M8A0.25W8
M8A0.35W0
M8A0.35W2
M8A0.35W4
M8A0.35W6
M8A0.35W8

Chances of Soil Liquefy (%)

100

T&Y Method for Zone II

Av. Chances of Soil Liquefy: 43.333%

80

70

60

50

40

30

20

10

Types of Model

Fig. 5.15: Chances of Percentage of Soil liquefaction under Different Conditions by T&Y Method for zone II

101

Fig. 5.15 showed that the chances of soil liquefy by T & Y method. In which two cases did
not showed any chances of soil liquefy which is M6 A0.15 W6 & M6 A0.15 W8. In ten cases
chances of soil liquefy showed less than 1%. There are only six conditions which showed that
the chances of soil liquefy > 80%. Maximum value is 84.545% for M8 A0.35 W0.
Influence of Earthquake Magnitude:In zone II most of the behavior is similar to zone I. Therefore it could be seen from Fig. 5.16
a & b that at ground acceleration 0.15g liquefaction chances are maximum at the depth of
water table 2.0 m i.e. 16.363%. On the other hand liquefaction chances are maximum at
ground acceleration 0.35g when water table is at 2.0 m below ground level i.e. 74.545 % and
84.545% respectively for magnitude 7.0 & 8.0. These characteristics show that liquefaction
chances increases due to increase in earthquake magnitude. But liquefaction chances were
maximum at the depth of 2.0 m water table for any ground acceleration which shows that the

Modified Seed's method

16.0

Modified Seed's method

Percentage of Soil Liquify (%)

Percentage of Soil Liquify (%)

soil strength properties on the top layer of soil is better than that of soil are at 2.0 m depth.

at W= 2.0 m & A=0.15g

14.0
12.0
10.0
8.0
6.0
4.0
2.0
0.0
5

at W= 2.0 m and A=0.35g

90.0
80.0
70.0
60.0
50.0
40.0
30.0
20.0
10.0
0.0
5

6
7
8
9
Variation of Earthquake Magnitude

Variation of Earthquake Magnitude

(a)

(b)

Fig. 5.16: Influence of Magnitude for (a) Depth of W.T. = 2.0 m and A=0.15g & (b) Depth of W.T. = 2.0 m and
A=0.35g by modified Seeds Method for zone II.

The chance of liquefaction in the soil for zone II by I&B method is also increases due to rise
in water table and increase in ground acceleration (Fig. 5.17 (a) & (b)). But at constant
ground acceleration leads to increase in chance of liquefaction up to 58.182% at depth of
102

water table is at ground surface whereas in another cases i.e. due to decrease in water table

I&B method

70.0

at A=0.15g and W. T.=0.0 m

60.0

I&B method

50.0
Percentage of Soil Liquify (%)

Percentage of Soil Liquify (%)

the variation shows up to below than 12%.

at A=0.25g and W. T.= 6.0 m

40.0

50.0

30.0

40.0
30.0

20.0

20.0

10.0

10.0
0.0
5

0.0
5

6
7
8
9
Variation of Earthquake Magnitude

6
7
8
9
Variation of Earthquake Magnitude

(a)

(b)

Fig. 5.17: Influence of Magnitude with respect to soil liquefy (%) for (a) W.T. = 0.0m and A=0.15g and (b)
W.T. = 6.0m and A=0.25g by I&B method for zone II.

Chances of Soil Liquify (%)

T&Y method
at W.T.= 0.0 m and A = 0.15g

T&Y method

18.0
16.0
14.0
12.0
10.0
8.0
6.0
4.0
2.0
0.0

at W.T.= 2.0 m and A = 0.15g

Chances of Soil Liquify (%)

10.0
9.0
8.0
7.0
6.0
5.0
4.0
3.0
2.0
1.0
0.0

6
7
8
9
Variation of Earthquake Magnitude

(a)

6
7
8
9
Variation of Earthquake Magnitude

(b)

Fig. 5.18: Influence of Magnitude for (a) Depth of W.T. = 0.0 m (b) Depth of W.T. = 2.0 m at A=0.15g by
modified T&Y for zone II.

On the basis of this difference in influence of liquefaction chances can also be investigated,
after comparing the results obtained from T&Y approach for zone-II with respect to
earthquake magnitude. It showed that for ground acceleration 0.15g and at ground level depth
of water table gave marginal variation of liquefaction potential i.e. liquefaction chances
increases to 3.636% when magnitude of earthquake varies from 6.0 to 7.0 and liquefaction

103

chances increases to 6.364 % for magnitude of earthquake changes from 7.0 to 8.0 (Fig. 5.18
(a)). Whereas, at 2.0 m depth of water table with constant ground acceleration 0.15g showed
from Fig. 5.18(b) that liquefaction potential changes from 2.727% to 5.455% when
earthquake magnitude changes from 6 to 7 although liquefaction potential changes from
5.455% to 13.636% when earthquake magnitude changes from 7 to 8.
But greater than 0.15g ground acceleration, chances of soil liquefies drastically increases.
This shows that the soil is more resistive up to 0.15g ground acceleration and at 2.0 m depth
of water table.

Influence of Ground Acceleration:Influence due to ground accelerations are shown in Fig.5.19 a & b for zone II by modified
Seeds method. It can be seen from the Fig. 5.19(a) that the liquefaction chances are low
between 0.15g to 0.25g whereas chances of liquefaction has increased drastically from 0.25g
to 0.35g for water table 2.0 m and earthquake magnitude 6.0. Similarly Fig. 5.19(b) shows in

60.0

Modified Seed's method

Percentage of Soil Liquify (%)

Percentage of Soil Liquify (%)

liquefaction potential gradual increase for higher magnitude at the same water table

at W.T.: 6.0 m and M=6.0

50.0
40.0
30.0
20.0
10.0
0.0
0.1

0.2

0.3

0.4

90.0
80.0
70.0
60.0
50.0
40.0
30.0
20.0
10.0
0.0

Modified Seed's method

at W.T.: 6.0 m and M=8.0

0.1
0.2
0.3
0.4
Variation of Horizontal Acceleration (g)

Variation of Horizontal Acceleration (g)

(a)

(b)

Fig. 5.19: Influence of Ground Acceleration at W.T. = 6.0m for (a) M=6.0 and (b) M=8.0 by modified Seeds
method for zone II.

I & B approach for chances of liquefaction is 40.875% between 0.15g and 0.25g (Fig. 5.14)
for magnitude 8 and water table at 2m. The chance of liquefaction between the consecutive
104

points 0.25g and 0.35g is 8.03%. Which is very low compared to initial points. In this case,
also the chances of liquefaction increasing with the increase in acceleration but increase is
non-uniform.
The chances of liquefaction by T&Y method shows the similar trend if compared with I&B
method. The difference of chances of liquefaction between two consecutive points i.e. 0.15g
and 0.25g is 57.664% and between 0.25g and 0.35g is 26.278% for earthquake magnitude 8.0

Chances of Soil Liquify (%)

and at depth of water table 4.0 m (Fig.5.21 b).

90.0

I&B method

80.0

at W.T.= 2.0 m at M=8.0

70.0
60.0
50.0
40.0
30.0
20.0

0.2
0.3
0.4
Variation of Ground Acceleration (g)
Fig. 5.20: Influence of Ground Acceleration for W.T. = 2.0m and M=8.0 by I & B method for zone II.

Chances of Soil Liquify (%)

0.1

90.0
80.0
70.0
60.0
50.0
40.0
30.0
20.0
10.0
0.0

T&Y method
at W.T.= 4.0 m and M=8.0

0.1

0.2
0.3
Variation of Ground Acceleration (g)

0.4

Fig. 5.21: Influence of Ground Acceleration for W.T. = 4.0m and M=8.0 by T&Y method for zone II.

By the analysis of the results obtained from T&Y method, it can be seen that in all cases
chances of liquefaction increases with the increase in earthquake magnitude with decrease in
depth of water table.

105

Influence of Water Table:The influence of water table on chances of liquefaction obtained by modified Seeds method,
I&B method & T&Y method could be seen in Fig. 5.22, 5.23 and 5.24 respectively. The
variation of water table with constant horizontal ground acceleration and earthquake
magnitude follows the similar pattern of zone I. This should be specifically mentioned that
the chances of liquefaction are very high when water table is at 2.0 m depth below the ground
level.

Modified Seed's method

Chances of soil liquefy (%)

Chances of soil liquefy (%)

85.0

at M=8.0 A = 0.35g

84.5
84.0
83.5
83.0
82.5
0

2
4
6
Variation of Water Table (m)

(a)

Modified Seed's method

80.0

at M=8.0A = 0.25g

70.0
60.0
50.0
40.0
30.0
0

Variation of Water Table (m)

(b)

Fig. 5.22: Influence of Depth of Water Table for (a) A=0.35g and M=8.0 (b) A=0.25g and M=8.0 by modified
Seeds method for zone II.

A regular variation can be seen from the curves. After 2.0 m depth of water table from
ground level, there is a decreasing trend of chances of liquefaction. Therefore depth of water
table from ground level plays a significant role in liquefaction. Though, chances of
liquefaction by modified Seeds method are very high in case of magnitude 8 and ground
acceleration 0.35g considering water table at 2.0 m from ground level which is 84.545% (Fig.
5.22(a)). In this case it could also be observed that liquefaction potential below 4.0 m depth
of water table from ground level the percentage of liquefaction in zone-II is constant i.e.
82.727%. Rather than this, chances of liquefaction initially increases up to 2.0 m depth of
water table again chances of liquefaction decreases with the decrease in water table which is
shown in Fig. 5.22 (b).
106

89.0

I&B method

60.0

Chances of Soil Liquify (%)

Chances of Soil Liquify (%)

70.0

at M=8.0, A = 0.15g

50.0
40.0
30.0
20.0
10.0
0.0
0

2
4
6
Variation of Water Table (m)

I&B method

88.0

at M=8.0, A = 0.35g

87.0
86.0
85.0
84.0
83.0
82.0
0

(a)

2
4
6
Variation of Water Table (m)

(b)

Fig. 5.23: Influence of Depth of Water Table for (a) A=0.15g and M=8.0 (b) A=0.35g and M=8.0 by I&B
method for zone II.

It is apparent from Fig. 5.23 (a) that the chances of liquefaction potential of soil decrease with
the increasing of water table. The falling trends were found for all types of combinations
except Fig. 5.23 (b) in which chances of liquefaction potential was constant in between depth
of water table 0.0 m to 2.0 m after 2.0 depth trend also decreases by I&B method of zone II.
The chances of liquefaction are very high in case of earthquake magnitude 8 and ground

85.0
84.0
83.0
82.0
81.0
80.0
79.0
78.0
77.0
76.0

T&Y method

Chances of Soil Liquify (%)

Chances of Soil Liquify (%)

acceleration 0.35g considering water table at ground level which is 84.545%.

at M=8.0, A = 0.35g

2
4
6
Variation of Water Table (m)

80.0
70.0
60.0
50.0
40.0
30.0
20.0
10.0
0.0

T&Y method
at M=8.0, A = 0.15g

(a)

2
4
6
Variation of Water Table (m)

(b)

Fig. 5.24: Influence of Depth of Water Table for (a) A=0.35g and M=8.0 (b) A=0.15g and M=8.0 by T&Y
method for zone II.

The trend of variation of chances of liquefaction potential of soil is similar to modified

Seeds method and I&B method. But in this method, chances of decreasing of liquefaction

107

drastically changes with rapid rate from ground level of water table to 4.0 m depth of water
table (Fig. 5.24 (b)).

5.2.3 Liquefaction Potential by Empirical Methods for Zone-III

Zone III covers total 42 borehole data i.e. 167 datasets. Chances of soil liquefaction in zone
III by different empirical methods is summarized in the Fig. 5.25. 5.26 & 5.27. Fig. 5.25
shows that the chances of soil liquefaction obtained from modified Seeds method for
parameters like M6A0.15W0, M6A0.15W6 and M6A0.15W8 are negligible where as M7A0.35W6,
M8A0.25W2, M8A0.35W0, M8A0.35W2, M8A0.35W4, M8A0.35W6 & M8A0.35W8 chances of soil
liquefaction is greater than 80%. It means properties/type of soil present in 134 data sets are
prone to liquefaction. In case of M8A0.35W2 total 156 data sets of soils having liquefaction
potential value less than 1. Therefore, liquefaction potential obtained by modified Seeds
method for earthquake magnitude 8.0, horizontal acceleration 0.35g and depth of water table
at 2.0 m is vulnerable for this zone.
Liquefaction potential determined for zone III from I & B method is shown in fig. 5.26. The
combination of M6A0.15W4, M6A0.15W6 & M6A0.15W8

do not have any chances of

liquefaction but other combination like M7A0.35W0, M8A0.25W0, M8A0.35W0, M8A0.35W2,

M8A0.35W4, M8A0.35W6, M8A0.35W8 have the chance of liquefaction which is greater than
80%.

Liquefaction potential estimated by T&Y method presented in Fig. 5.27 for zone III.
According to this figure, 10 combinations do not show any chances of soil liquefaction but 12
combinations have the chances of soil liquefaction by 90%. Average chances of soil
liquefaction in this zone is 48.902% by T&Y method which is greater than average chances
of liquefaction potential obtained by other methods.

108

M6A0.15W0
M6A0.15W2
M6A0.15W4
M6A0.15W6
M6A0.15W8
M6A0.25W0
M6A0.25W2
M6A0.25W4
M6A0.25W6
M6A0.25W8
M6A0.35W0
M6A0.35W2
M6A0.35W4
M6A0.35W6
M6A0.35W8
M7A0.15W0
M7A0.15W2
M7A0.15W4
M7A0.15W6
M7A0.15W8
M7A0.25W0
M7A0.25W2
M7A0.25W4
M7A0.25W6
M7A0.25W8
M7A0.35W0
M7A0.35W2
M7A0.35W4
M7A0.35W6
M7A0.35W8
M8A0.15W0
M8A0.15W2
M8A0.15W4
M8A0.15W6
M8A0.15W8
M8A0.25W0
M8A0.25W2
M8A0.25W4
M8A0.25W6
M8A0.25W8
M8A0.35W0
M8A0.35W2
M8A0.35W4
M8A0.35W6
M8A0.35W8

Chances of Soil Liquefy (%)

100.000
90.000

Modified Seed's Method for Zone III

Av. Chances of Soil Liquefaction 36.141%

80.000

70.000

60.000

50.000

40.000

30.000

20.000

10.000

0.000

Types of Combinations

Fig. 5.25: Chances of Soil liquefaction under Different Conditions by Modified Seeds Method for zone III in Percentage.

109

90

M6A0.15W0
M6A0.15W2
M6A0.15W4
M6A0.15W6
M6A0.15W8
M6A0.25W0
M6A0.25W2
M6A0.25W4
M6A0.25W6
M6A0.25W8
M6A0.35W0
M6A0.35W2
M6A0.35W4
M6A0.35W6
M6A0.35W8
M7A0.15W0
M7A0.15W2
M7A0.15W4
M7A0.15W6
M7A0.15W8
M7A0.25W0
M7A0.25W2
M7A0.25W4
M7A0.25W6
M7A0.25W8
M7A0.35W0
M7A0.35W2
M7A0.35W4
M7A0.35W6
M7A0.35W8
M8A0.15W0
M8A0.15W2
M8A0.15W4
M8A0.15W6
M8A0.15W8
M8A0.25W0
M8A0.25W2
M8A0.25W4
M8A0.25W6
M8A0.25W8
M8A0.35W0
M8A0.35W2
M8A0.35W4
M8A0.35W6
M8A0.35W8

Chances of Soil Liquefy (%)

100

I&B Method for Zone III

Av. Chances of Soil Liquefaction 43.247%

80

70

60

50

40

30

20

10

Types of Combinations

Fig. 5.26: Chances of Soil liquefaction under Different Conditions by I&B Method for zone III in Percentage.

110

100.000

M6A0.15W0
M6A0.15W2
M6A0.15W4
M6A0.15W6
M6A0.15W8
M6A0.25W0
M6A0.25W2
M6A0.25W4
M6A0.25W6
M6A0.25W8
M6A0.35W0
M6A0.35W2
M6A0.35W4
M6A0.35W6
M6A0.35W8
M7A0.15W0
M7A0.15W2
M7A0.15W4
M7A0.15W6
M7A0.15W8
M7A0.25W0
M7A0.25W2
M7A0.25W4
M7A0.25W6
M7A0.25W8
M7A0.35W0
M7A0.35W2
M7A0.35W4
M7A0.35W6
M7A0.35W8
M8A0.15W0
M8A0.15W2
M8A0.15W4
M8A0.15W6
M8A0.15W8
M8A0.25W0
M8A0.25W2
M8A0.25W4
M8A0.25W6
M8A0.25W8
M8A0.35W0
M8A0.35W2
M8A0.35W4
M8A0.35W6
M8A0.35W8

Chances of Soil Liquefy (%)

T&Y Method for Zone III

Av. Chances of Soil Liquefaction 48.902%

90.000

80.000

70.000

60.000

50.000

40.000

30.000

20.000

10.000

0.000

Types of Combinations

Fig. 5.27: Chances of Soil liquefaction under Different Conditions by T&Y Method for zone III in Percentage.

111

Influence of Earthquake Magnitude:It could be observed from Fig. 5.28 liquefaction percentage is increasing with the increase in
magnitude. The percentage of liquefaction between magnitude 6 and 7 is 7.186% and it is
9.580% between magnitude 7 and 8. The behaviour of liquefaction potential is not uniform

Percentage of Soil Liquify(%)

with respect to variation in magnitude.

18.0

Modified Seed's method

16.0

at A=0.15g and W.T.=0.0 m

14.0
12.0
10.0
8.0
6.0
4.0
2.0
0.0
5

7
8
9
Variation of Earthquake Magnitude
Fig. 5.28: Influence of Magnitude for W.T. = 0.0m and A=0.15g by modified Seeds method for zone III.

In this case modified seed`s method resembles that the increase in magnitude; the increase
magnitude is dominates the chances of liquefaction. The similar patterns have also been
observed by I&B method (Fig. 5.29). The difference is with respect to chances of liquefaction
in percentage by I&B method when compared to modified Seeds method for some
combinations in zone III.

Percentage of Soil Liquify (%)

70.0

I&B method

60.0

at W.T.= 0.0 m and A=0.15g

50.0
40.0
30.0
20.0
10.0
0.0
5

6
7
8
9
Variation of Earthquake Magnitude
Fig. 5.29: Influence of Earthquake Magnitude for W.T. = 0.0m and A=0.15g by I&B method for zone III.

112

T&Y method also indicates that increase in magnitude will increase the chances of
liquefaction (Fig. 5.30 (a)) when water table is at ground level and ground acceleration is
0.35g. It has been observed that the increase in liquefaction potential is very low between two
points as 2.189% for magnitude 6.0 & 7.0. When water table is at 4.0m and the ground
acceleration is kept constant at 0.15g there is no probability of liquefaction between the
earthquake magnitude 6.0 to 7.0 (Fig. 5.30 (b)).
The first nature shows the tendency of increase in chances of liquefaction with increase in
earthquake magnitude and decrease in water table. Whereas another Fig. 5.30 (b) shows that
there are no chances of liquefaction for with increase in earthquake magnitude from 6 to 7
and insignificant increment have been observed when the earthquake magnitude increases

97.8

T&Y method

97.6

at W: 0.0 m and A=0.35g

97.4
97.2
97.0
96.8
96.6
96.4

Percentage of Soil Liquify(%)

Percentage of Soil Liquify (%)

from 7 to 8 for constant depth of water table.

96.2
5

T&Y method

0.7

at W: 4.0 m and A=0.15g

0.6
0.5
0.4
0.3
0.2
0.1
0.0
5

6
7
8
9
Variation of Earthquake Magnitude

Variation of Earthquake Magnitude

(a)

(b)

Fig. 5.30(a): Influence of earthquake magnitude for (a) for A=0.35g and W.T.= 0.0 m and (b) for A=0.15g and
W.T.= 4.0 m by T&Y method for zone III.

Influence of Ground Acceleration:The present analysis represents the variation in the tendency of liquefaction with respect to
the variation in ground acceleration by different approach. The Fig 5.31(a) shows the chances
of liquefaction for the earthquake magnitude 6.0 and water table at ground level by modified
Seeds method. When the ground acceleration increases from 0.15g to 0.25g, the change in

113

the chances of liquefaction is 9.581% and for ground acceleration variation from 0.25 to 0.35
it is 10.778%. which is non-uniform in nature.
Modified Seed's method

25.0

at M= 6.0 and W= 0.0 m

Percentage of Soil Liquify (%%)

Percentage of Soil Liquify(%)

25.0
20.0
15.0
10.0
5.0
0.0

Modified Seed's method

at M= 6.0 and W= 8.0 m

20.0
15.0
10.0
5.0
0.0

0.1 Variation0.2
0.3
0.4
of Horizontal
Acceleration
(g)

0.1
0.2
0.3
0.4
Variation of Horizontal Acceleration (g)

(a)

(b)

Fig. 5.31: Influence of Ground Acceleration for (a) W.T. = 0.0m and for M=6.0 and (b) W.T. = 0.0m and for
M=8.0 by modified Seeds method for zone III.

Percentage of Soil Liquify (%)

90.0

I&B method
at Magnitude 8.0 and depth of Water Table: 2.0 m

85.0
80.0
75.0
70.0
65.0
60.0
55.0
0.1

0.2
0.3
0.4
Variation of Horizontal Acceleration (g)
Fig. 5.32: Influence of Ground Acceleration for W.T. = 2.0m and M=8.0 by I & B method for zone III.

Estimated liquefaction potential by I&B method the analysis of graph shows the linear
behaviour as in seeds method. For water table 2m from ground level and earthquake
magnitude 8.0, the chances of liquefaction for ground acceleration between 0.15g and 0.25g
increases by 15.03% but between from 0.25g to 0.35g it is 8.383%, which represents the
increment in non-uniform manner.

114

By T&Y method, also the chances of liquefaction is increasing with the increase in
acceleration for initial points of acceleration but the percentage of liquefaction decreasing for
consecutive point of acceleration (Fig. 5.33).

Percentage of Soil Liquify (%)

100.0

T&Y method
at M= 8.0 and W.T.= 4.0 m

95.0
90.0
85.0
80.0
75.0
70.0
0.1

0.2
0.3
0.4
Variation of Horizontal Acceleration (g)
Fig. 5.33: Influence of Earthquake Magnitude M=8.0 and Depth of Water Table is at 4.0 m by T&Y method for
zone III.

Influence of Water Table:Fig 5.34(a) represents the variation in liquefaction potential with respect to change in the
ground water level keeping the earthquake magnitude constant at 8.0 and horizontal ground
acceleration 0.35g for zone III by modified Seeds method. The tendency of liquefaction
increases rapidly when the water table changes from ground level to 2.0m depth. This
variation can be seen in all conditions, thus could be significantly attributed to the density of
the soil at that level. Therefore, shallow foundations at the depth of 2.0m will be susceptible
to liquefaction leading to differential settlement. To avoid such condition anti-liquefaction
measures should be adopted. Fig 5.34(b) shows, the liquefaction potential of 82.036% when
the water table is at 2.0m depth from ground level.

115

Modified Seed's method

Modified Seed's method

85.0

at M= 8.0 and A = 0.35g

Percentage of Soil Liquify (%)

Percentage of Soil Liquify (%)

94.0
92.0
90.0
88.0
86.0
84.0
82.0
0

at M= 8.0 and A = 0.25g

80.0
75.0
70.0
65.0
60.0
55.0
50.0

2
4
6
8
Variation of Depth of Water Table

2
4
6
Variation of Depth of Water Table

(a)

(b)

Fig. 5.34: Influence of Depth of Water Table for (a) A=0.35g and M=8.0 and (b) A=0.25g and M=8.0 by
modified Seeds method for zone III.

I&B method

60.0

Percentage of Soil Liquify (%)

Percentage of Soil Liquify (%)

70.0

at A = 0.15g & M=8.0

50.0
40.0
30.0
20.0
10.0
0.0
0

2
4
6
8
Variation of Depth of Water Table (m)

(a)

89.5
89.0
88.5
88.0
87.5
87.0
86.5
86.0
85.5
85.0
84.5

I&B method
at A = 0.35g & M=8.0

2
4
6
8
Variation of Depth of Water Table (m)

(b)

Fig. 5.35: Influence of Depth of Water Table for (a) A=0.15g and M=8.0 and (b) A=0.35g and M=8.0 by I&B
method for zone III.

The liquefaction potential with respect to the variation in the depth of water table for zone III
by I&B method are shown in Fig. 5.35(a) & (b). The liquefaction of soil decreases as water
table decreases from G.L. i.e. 0.0 m to 8.0 m depth. The falling trends were found for all
types of combinations except for the combination of M=8.0 and A=0.35g (Fig. 5.35(b)) in
which chances of liquefaction potential was constant in between depth of water table 2.0 m to
4.0 m, and chances of liquefaction is maximum when the water table is at ground level.

116

T&Y method
at A = 0.35g & M=8.0

Percentage of Soil Liquify (%)

Percentage of Soil Liquify u(%)

98.0
97.5
97.0
96.5
96.0
95.5
95.0
94.5
94.0
93.5

T&Y method

90.0
80.0
70.0
60.0
50.0
40.0
30.0
20.0
10.0
0.0

at A = 0.15g & M=8.0

2
4
6
8
Variation of Depth of Water Table

2
4
6
8
Variation of Depth of Water Table

(a)

(b)

Fig. 5.36: Influence of Depth of Water Table for (a) A=0.35g and M=8.0 and (b) A=0.15g and M=8.0 by T&Y
method for zone III.

The graphical representation in Fig 5.36 (a) & (b) shows the variation in liquefaction
potential with the variation in water table keeping the earthquake magnitude and ground
acceleration constant for zone III by T&Y method. The decreasing trend of liquefaction
potential have been observed between water table is 0.0 m to 4.0 m depth and consecutively it
remains very high i.e. 94.012% for 6.0m & 8.0 m. Liquefaction potential for ground
acceleration 0.15g and earthquake magnitude 8.0 is maximum at ground level i.e. 79.641%,
and it suddenly drops down to 0% at water table 4.0m depth comparative variation for 0.15g
and 0.35g is very significant.
5.2.4 Liquefaction Potential by Empirical Methods for Zone-IV
Zone IV comprises the study of liquefaction potential based on 32 borehole data i.e. 104
datasets. It has been observed from Fig. 5.37 that liquefaction potential percentage at
earthquake magnitude 6.0 did not exceed 20% for every combination. Fig. 5.37 also shows
that at every depth of water table for earthquake magnitude 7 and horizontal acceleration
0.15g has the liquefaction potential percentage below 20%. In contrast to above combination
M8 A0.35 W2 crosses the chances of liquefaction potential greater than 80%. In this zone,

117

average chances of liquefaction potential obtained by modified Seeds method is 28.034%

which is less in comparison other zones.

118

M6A0.15W0
M6A0.15W2
M6A0.15W4
M6A0.15W6
M6A0.15W8
M6A0.25W0
M6A0.25W2
M6A0.25W4
M6A0.25W6
M6A0.25W8
M6A0.35W0
M6A0.35W2
M6A0.35W4
M6A0.35W6
M6A0.35W8
M7A0.15W0
M7A0.15W2
M7A0.15W4
M7A0.15W6
M7A0.15W8
M7A0.25W0
M7A0.25W2
M7A0.25W4
M7A0.25W6
M7A0.25W8
M7A0.35W0
M7A0.35W2
M7A0.35W4
M7A0.35W6
M7A0.35W8
M8A0.15W0
M8A0.15W2
M8A0.15W4
M8A0.15W6
M8A0.15W8
M8A0.25W0
M8A0.25W2
M8A0.25W4
M8A0.25W6
M8A0.25W8
M8A0.35W0
M8A0.35W2
M8A0.35W4
M8A0.35W6
M8A0.35W8

Chances of Soil Liquefy(%)

100.000

90.000

Modified Seed's Method for Zone IV

Av. Chances of Soil Liquefaction 28.034%

80.000

70.000

60.000

50.000

40.000

30.000

20.000

10.000

0.000

Types of Combination

Fig. 5.37: Chances of Soil liquefaction under Different Conditions by Modified Seeds Method for zone IV in Percentage.

119

80.000

M6A0.15W0
M6A0.15W2
M6A0.15W4
M6A0.15W6
M6A0.15W8
M6A0.25W0
M6A0.25W2
M6A0.25W4
M6A0.25W6
M6A0.25W8
M6A0.35W0
M6A0.35W2
M6A0.35W4
M6A0.35W6
M6A0.35W8
M7A0.15W0
M7A0.15W2
M7A0.15W4
M7A0.15W6
M7A0.15W8
M7A0.25W0
M7A0.25W2
M7A0.25W4
M7A0.25W6
M7A0.25W8
M7A0.35W0
M7A0.35W2
M7A0.35W4
M7A0.35W6
M7A0.35W8
M8A0.15W0
M8A0.15W2
M8A0.15W4
M8A0.15W6
M8A0.15W8
M8A0.25W0
M8A0.25W2
M8A0.25W4
M8A0.25W6
M8A0.25W8
M8A0.35W0
M8A0.35W2
M8A0.35W4
M8A0.35W6
M8A0.35W8

Soil Liquefy under Different Condotions (%)

90.000

I&B Method for Zone IV

Av. Chances of Soil Liquefaction 35.021%

70.000

60.000

50.000

40.000

30.000

20.000

10.000

0.000

Types of Combination

Fig. 5.38: Chances of Soil liquefaction under Different Conditions by I&B Method for zone IV in Percentage.

120

90

M6A0.15W0
M6A0.15W2
M6A0.15W4
M6A0.15W6
M6A0.15W8
M6A0.25W0
M6A0.25W2
M6A0.25W4
M6A0.25W6
M6A0.25W8
M6A0.35W0
M6A0.35W2
M6A0.35W4
M6A0.35W6
M6A0.35W8
M7A0.15W0
M7A0.15W2
M7A0.15W4
M7A0.15W6
M7A0.15W8
M7A0.25W0
M7A0.25W2
M7A0.25W4
M7A0.25W6
M7A0.25W8
M7A0.35W0
M7A0.35W2
M7A0.35W4
M7A0.35W6
M7A0.35W8
M8A0.15W0
M8A0.15W2
M8A0.15W4
M8A0.15W6
M8A0.15W8
M8A0.25W0
M8A0.25W2
M8A0.25W4
M8A0.25W6
M8A0.25W8
M8A0.35W0
M8A0.35W2
M8A0.35W4
M8A0.35W6
M8A0.35W8

Chances of Soil Liquefy (%)

100

T&Y Method for Zone IV

Av. Chances of Soil Liquefaction 35.513%

80

70

60

50

40

30

20

10

Types of Combination

Fig. 5.39: Chances of Soil liquefaction under Different Conditions by T&Y Method for zone IV in Percentage.

121

The average chances of liquefaction potential by I&B method for zone IV is 35.021% as
shown in Fig. 5.38. It can be seen from Fig. 5.38 that liquefaction potential for all
combination are less than 80%. But there are seven combinations in which chances of
liquefaction potential lies between 70 80%.
According to T & Y method average chances of soil susceptibility to liquefaction is 35.513%
as per Fig. 5.39. The chances of liquefaction for the combinations of M6 A0.35 W0; M7 A0.35
W0; M8 A0.25 W0; M8 A0.35 W0 are greater than 80% whereas the combinations of M7 A0.35 W0
and M8 A0.35 W0 are 90.385% and 95.192% respectively which is vulnerable in nature.
Influence of Earthquake Magnitude:It could be seen from Fig 5.40 that chances of liquefaction rise in percentage is significantly
high with respect to increase in magnitude. The difference between magnitude 6.0 and 7.0 is
39.5% and difference between magnitude 7.0 and 8.0 is 25%. These graphs represent the
relationship between the two parameters which is linear in nature by modified Seeds
approach.

Percentage of Soil Liquify (%)

90.0

Modified Seed's method

80.0

at W.T.= 4.0 m & A=0.35g

70.0
60.0
50.0
40.0
30.0
20.0
10.0
0.0
5

Variation of Earthquake Magnitude

Fig. 5.40: Influence of Magnitude for W.T. = 4.0m and A=0.35g by modified Seeds method for zone IV.

I&B method shows the dominating tendency of magnitude despite of decrease in water table.
Fig 5.41 shows that the liquefaction potential difference between the magnitude 6 & 7 is
34.615% and 7 & 8 is 20.554%. The variation of percentage of liquefaction with magnitude is
122

non-uniform but linear in nature for horizontal acceleration 0.35g. This behaviour shows that
chances of liquefaction increase with magnitude (Fig. 5.41).
By T&Y method also increase in magnitude tends to increase the chances of liquefaction.
When the water table is at ground level and ground acceleration is 0.25g the variation in
liquefaction potential from 24.036% to 44.123% due to varies in earthquake magnitude from
6.0 to 7.0 (Fig. 5.42(a)). For another observation when the water table is 0.0m and the ground
acceleration is kept constant at 0.15g there is also nonlinear variation in chances of

Percentage of Soil Liquify (%)

liquefaction occurrence (Fig. 5.42(b)).

I&B method

90.0
80.0
70.0
60.0
50.0
40.0
30.0
20.0
10.0
0.0

at W.T.= 6.0 m and A=0.35g

7
Variation of Magnitude

Fig. 5.41: Influence of Earthquake Magnitude for W.T. = 6.0m and A=0.35g by I&B method for zone IV.

at W.T.= 2.0 m and A=0.25g

Percentage of Soil Liquify (%)

Percentage of Soil Liquify (%)

T&Y method
70.0
60.0
50.0
40.0
30.0
20.0
10.0
0.0
5

6
7
8
9
Variation of Earthquake Magnitude
(a)

T&Y method

60.0

at W.T.= 0.0 m and A=0.15g

50.0
40.0
30.0
20.0
10.0
0.0
5

6
7
8
9
Variation of Earthquake Magnitude
(b)

Fig. 5.42: Influence of earthquake magnitude for (a) A=0.25g and W.T.= 2.0 m and (b) A=0.15g and W.T.= 0.0
m by T&Y method for zone IV.

123

Influence of Ground Acceleration:This analysis represents the variation in the tendency of liquefaction with respect to the
variation in ground acceleration. The graph shows a dominating trend as the increase in
ground acceleration also increases the chances of liquefaction estimated by modified Seeds
method. Fig. 5.43 (a) shows relation between percentage of liquefaction and horizontal
acceleration at earthquake magnitude 8.0. The difference in liquefaction potential between
0.15g & 0.25g is 34.615% and 0.25g & 0.35g is 25.000%. Fig. 5.43(b) shows slightly

90.0

Modified Seed's method

80.0

at W= 4.0 m & M=8.0

Percentage of Soil Liquify (%)

Percentage of Soil Liquify (%)

different behaviour as compare to previous one.

70.0
60.0
50.0
40.0
30.0
20.0
10.0
0.0
0.1

0.2

0.3

50.0
45.0
40.0
35.0
30.0
25.0
20.0
15.0
10.0
5.0
0.0

Modified Seed's Method

at W = 6.0 m &M=6.0

0.1

0.4

Variation of Horizontal Acceleration (g)

0.2

0.3

0.4

Variation of Horizontal Acceleration (g)

(a)

(b)

Fig. 5.43: Influence of Ground Acceleration for (a) W.T. = 4.0m and M=8.0 and (b) W.T. = 6.0m and M=6.0 by
modified Seeds method for zone IV.

The analysis of I&B (Fig. 5.44) liquefaction potential shows the same behaviour as in
modified Seeds method. For the water table 8.0m and earthquake magnitude 7.0 but at
ground acceleration 0.15g liquefaction potential is zero and it increases from 0.15g to 0.25g.
The increase in liquefaction potential between 0.25g to 0.35g is 55.770% which is very high.
Compare to above cases of Seeds method and I&B method, the T&Y approach also shows
the similar trend of liquefaction percentage with variation in acceleration. In T&Y method
variation in percentage of liquefaction potential is nearly constant between 0.15g and 0.25g.
124

Then there is sudden increase in liquefaction percentage between 0.25g to 0.35g and this
percentage change is 81.731% for water table 8.0m and magnitude 7.0.
80.0

I&B method

Percentage of Soil Liquify (%)

70.0

at W= 8.0 m at M=7.0

60.0
50.0
40.0
30.0
20.0
10.0
0.0
0.1

0.2
0.3
0.4
Variation of Horizontal Acceleration (g)
Fig. 5.44: Influence of Ground Acceleration for W.T. = 8.0m and M=7.0 by I & B method for zone IV.

Percentage of Soil Liquify (%)

100.0

T&Y method
at depth of Water Table: 8.0 m at M=7.0

80.0
60.0
40.0
20.0
0.0
0.1

0.2

0.3

0.4

Variation of Horizontal Acceleration (g)

Fig. 5.45: Influence of Ground Acceleration for W.T. = 8.0m and M=7.0 by T&Y method for zone IV.

By the analysis of the results obtained from T&Y method, it can be seen that in all cases
variation of chances of liquefaction increases with the increase in earthquake magnitude and
decrease in depth of water table (Fig. 5.39).
Influence of Water Table:Liquefaction potential is highly affected by the presence and the level of ground water table.
Fig 5.46a represents the variation in liquefaction potential with respect to change in the
ground water level keeping the earthquake magnitude constant at 8.0 and horizontal ground

125

acceleration 0.35g for zone IV by modified Seeds method. The tendency of liquefaction
percentage is maximum when water table is at 2.0m depth from ground level. These trend
variations have been seen in previous zone also. Liquefaction percentage is nearly constant
between 4.0m to 6.0m of water table and it has been noted that it is move that 75% at all the
trends of water.
Fig 5.46(b) shows another variation of liquefaction potential for M=7.0 and A=0.35g. In this
case abrupt change has not been seen between ground level to 2.0m of water table and
percentage of liquefaction potential is decreasing after 2.0m depth and this tendency is
negligible between 6.0m to 8.m which is 2.884% only.
Modified Seed's method
Percentage of Soil Liquify (%)

Percentage of Soil Liquify (%)

87.0

at A = 0.35g & M=8.0

85.0
83.0
81.0
79.0
77.0
75.0

65.0

Modified Seed's method

60.0

at A = 0.35g & M=7.0

55.0
50.0
45.0
40.0
35.0
30.0

2
4
6
8
Variation of Depth of Water Table (m)

(a)

2
4
6
8
Variation of Depth of Water Table (m)

(b)

Fig. 5.46: Influence of Depth of Water Table for (a) A=0.35g and M=8.0 and (b) A=0.35g and M=7.0 by
Modified Seeds method for zone IV.

Fig 5.47 shows the similar trend of percentage liquefaction potential with respect to the
variation in the depth of water table for zone IV by I&B method. It could be clearly depicted
that the chances of liquefaction of soil decrease with the increasing of water table which is
different from Seeds method. Fig. 5.47(a) shows the decrease in percentage liquefaction
from ground level to 4.0m depth beyond which it is zero. Fig. 5.47(b) shows that liquefaction
potential is decreasing with water table depth.

126

12.0

I&B method

80.0

I&B method

at A = 0.15g at M=6

70.0

at A = 0.25g & M=7.0

Percentage of Soil Liquify(%)

Percentage of Soil Liquify (%)

14.0

10.0
8.0
6.0
4.0
2.0

60.0
50.0
40.0
30.0
20.0
10.0
0.0

0.0
0

2
4
6
8
Variation of Depth of Water Table (m)

2
4
6
8
Variation of Depth of Water Table (m)

(a)

(b)

90.0
80.0
70.0
60.0
50.0
40.0
30.0
20.0
10.0
0.0

Percentage of Soil Liquify (%)

Percentage of Soil Liquify (%)

Fig. 5.47: Influence of Depth of Water Table for (a) A=0.15g and M=6.0 and (b) A=0.25g and M=7.0 by I&B
method for zone IV.

T&Y method
at A = 0.35g & M=6.0

2
4
6
8
Variation of Depth of Water Table (m)

90.0
80.0
70.0
60.0
50.0
40.0
30.0
20.0
10.0
0.0

T&Y method
at A = 0.25g & M=7.0

Variation of Depth of Water Table (m)

(a)

(b)

Fig. 5.48: Influence of Depth of Water Table for (a) A=0.35g and M=6.0 and (b) A=0.25g and M=7.0 by T&Y
method for zone IV.

Fig 5.48 shows the variation in liquefaction potential obtained by T&Y method for zone IV.
It also shows the decreasing trend in percentage liquefaction like I&B method. The maximum
chance of liquefaction occurs when the water table is at ground level which is around
82.692% (Fig. 5.48(a)) and then it decreases to 19.231% at 8.0m depth. Fig.5.48 (b) shows
that maximum liquefaction is 78.846% at ground level when A=0.25g and M=7.0.

5.2.5 Liquefaction Potential by Empirical Methods for Zone-V

127

Zone V comprises of 36 borehole data with 126 datasets. On the basis of these datasets barchart of 45 combinations with chances of liquefy in percent obtained by three conventional
method is plotted in Fig. 5.49, 5.50 & 5.51. The maximum chance of soil liquefaction by
conventional method is around 80%. Average value of soil liquefaction by modified Seeds
method is 41.975%. At earthquake magnitude 6.0, chances of soil liquefaction is less than
50% except for the model M6 A0.35 W2.
Variation from M8A0.25W0 to M8A0.35W8 (Fig. 5.50) shows that the chances of soil
liquefaction remain constant or in another way, the variation of chances of soil liquefaction is
too less. There are three conditions i.e. M6A0.15W4, M6A0.15W6 & M6A0.15W8 which do not
show any chances of soil liquefaction.
In case of T&Y method, it has been observed that maximum chances of occurrence of soil
liquefaction potential have been obtained for combination M8A0.35W0 is 80.159% (Fig. 5.51).

128

M6A0.15W0
M6A0.15W2
M6A0.15W4
M6A0.15W6
M6A0.15W8
M6A0.25W0
M6A0.25W2
M6A0.25W4
M6A0.25W6
M6A0.25W8
M6A0.35W0
M6A0.35W2
M6A0.35W4
M6A0.35W6
M6A0.35W8
M7A0.15W0
M7A0.15W2
M7A0.15W4
M7A0.15W6
M7A0.15W8
M7A0.25W0
M7A0.25W2
M7A0.25W4
M7A0.25W6
M7A0.25W8
M7A0.35W0
M7A0.35W2
M7A0.35W4
M7A0.35W6
M7A0.35W8
M8A0.15W0
M8A0.15W2
M8A0.15W4
M8A0.15W6
M8A0.15W8
M8A0.25W0
M8A0.25W2
M8A0.25W4
M8A0.25W6
M8A0.25W8
M8A0.35W0
M8A0.35W2
M8A0.35W4
M8A0.35W6
M8A0.35W8

Chances of Soil Liquefy (%)

80

70

Modified Seed's Method for Zone V

Av. Chances of Soil Liquefaction 41.975%

60

50

40

30

20

10

Types of Combinations

Fig. 5.49: Chances of Soil Liquefaction under Different Conditions by Modified Seeds Method for zZone V in Percentage.

129

80

M6A0.15W0
M6A0.15W2
M6A0.15W4
M6A0.15W6
M6A0.15W8
M6A0.25W0
M6A0.25W2
M6A0.25W4
M6A0.25W6
M6A0.25W8
M6A0.35W0
M6A0.35W2
M6A0.35W4
M6A0.35W6
M6A0.35W8
M7A0.15W0
M7A0.15W2
M7A0.15W4
M7A0.15W6
M7A0.15W8
M7A0.25W0
M7A0.25W2
M7A0.25W4
M7A0.25W6
M7A0.25W8
M7A0.35W0
M7A0.35W2
M7A0.35W4
M7A0.35W6
M7A0.35W8
M8A0.15W0
M8A0.15W2
M8A0.15W4
M8A0.15W6
M8A0.15W8
M8A0.25W0
M8A0.25W2
M8A0.25W4
M8A0.25W6
M8A0.25W8
M8A0.35W0
M8A0.35W2
M8A0.35W4
M8A0.35W6
M8A0.35W8

Chances of Soil Liquefy (%)

90

I&B Method for Zone V

Av. Chances of Soil Liquefaction 45.644%

70

60

50

40

30

20

10

Types of Combinations

Fig. 5.50: Chances of Soil Liquefaction under Different Conditions by I&B Method for zone V in Percentage.

130

90

M6A0.15W0
M6A0.15W2
M6A0.15W4
M6A0.15W6
M6A0.15W8
M6A0.25W0
M6A0.25W2
M6A0.25W4
M6A0.25W6
M6A0.25W8
M6A0.35W0
M6A0.35W2
M6A0.35W4
M6A0.35W6
M6A0.35W8
M7A0.15W0
M7A0.15W2
M7A0.15W4
M7A0.15W6
M7A0.15W8
M7A0.25W0
M7A0.25W2
M7A0.25W4
M7A0.25W6
M7A0.25W8
M7A0.35W0
M7A0.35W2
M7A0.35W4
M7A0.35W6
M7A0.35W8
M8A0.15W0
M8A0.15W2
M8A0.15W4
M8A0.15W6
M8A0.15W8
M8A0.25W0
M8A0.25W2
M8A0.25W4
M8A0.25W6
M8A0.25W8
M8A0.35W0
M8A0.35W2
M8A0.35W4
M8A0.35W6
M8A0.35W8

Chances of Soil Liquefy(%)

100

T&Y Method for Zone V

Av. Chances of Soil Liquefaction 44.515%

80

70

60

50

40

30

20

10

Types of Combinations

Fig. 5.51: Chances of Soil Liquefaction under Different Conditions by T&Y Method for Zone V in Percentage.

131

Influence of Earthquake Magnitude:It has been already been establish by the study of previous zones that earthquake magnitude
plays an important role. Though zone V shows that value of liquefaction potential at
earthquake magnitude 6.0 to 8.0 is less than 50% for horizontal acceleration 0.15g. At
horizontal acceleration 0.25g chances of L.P. again lies below than 50 % for earthquake
magnitude 6 and 7 except at depth of water table 2.0m (Fig. 5.52(a)).
Chances of percentage in liquefaction again increase with the increase in earthquake
magnitude at horizontal acceleration 0.35g. It increases by 71% when water table is at 4.0m.
Fig. 5.52(b) also shows that the chances of liquefaction potential increases to 74.603% when
earthquake magnitude increases from 6.0 to 7.0 and chances of liquefaction potential
increases and reached to 75.397% when earthquake magnitude changes from 7.0 to 8.0,

80.0

Modified Seed's method

70.0

at A=0.25g and W=2.0m

60.0
50.0
40.0
30.0
20.0
5

6
7
8
Variation of Earthquake Magnitude

Percentage of Soil Liquify (%)

Percentage of Soil Liquify (%)

whereas in other cases its value is below 20%.

(a)

80.0

Modified Seed's method

at A=0.35g and W= 2.0m

70.0
60.0
50.0
40.0
5
6
7
Variation of Earthquake Magnitude

(b)

Fig. 5.52: Influence of Earthquake Magnitude for (a) A=0.25g and W=2.0 and (b) A=0.35g and W.T.=2.0 by
Modified Seeds method for zone V.

The influence of magnitude can also be analyzed in case of I&B method despite the decrease
in water table the dominating tendency of magnitude is clearly visible though the percentage
of soil liquefaction have been reduced.
Fig 5.50 shows that in three cases which are M6A0.25W0, M6A0.35W0 and M6A0.35W2 having
chances of liquefaction potential at earthquake magnitude 6.0 is greater than 50% which are

132

57.937%, 71.429% and 63.492% respectively for zone V by I&B method. But at earthquake
intensity 6.0,7.0 & 8.0 and horizontal acceleration at 0.15g there is only one condition
M8A0.15W0 shows chances of liquefaction potential greater than 50% i.e. 64.286% (Fig.
5.53(a)). This figure also shows that linear variation in percentage of liquefaction occurrences

70.0

I&B method

60.0

at W.T.= 0.0 m and A=0.15g

Percentage of Soil Liquify(%)

Percentage of Soil Liquify (%)

due to variation in earthquake magnitude.

50.0
40.0
30.0
20.0
10.0
0.0
5

I&B method

76.0

at W.T.= 0.0 m and A=0.35g

75.0
74.0
73.0
72.0
71.0

6
7
8
Variation of Earthquake Magnitude

(a)

6
7
8
Variation of Earthquake Magnitude

(b)

Fig. 5.53: Influence of Earthquake Magnitude for (a) A=0.15g and W=0.0 and (b) A=0.35g and W.T.=0.0 by
I&B method for zone V.

Least variation can be seen in Fig. 5.53(b) at horizontal acceleration 0.35g when water table
is at ground level the chances of liquefaction potential at 6.0, 7.0 and 8.0 are respectively

70.0

T&Y method

60.0

at W.T.= 0.0 m and A=0.15g

Percentage of Soil Liquify (%)

Percentage of Soil Liquify (%)

71.429%, 74.603% and 75.397%.

50.0
40.0
30.0
20.0
10.0
0.0
5

6
7
8
Variation of Earthquake Magnitude

T&Y method

80.5

at W.T.= 0.0 m and A=0.35g

80.0
79.5
79.0
78.5
78.0
5

(a)

6
7
8
Variation of Earthquake Magnitude

(b)

Fig. 5.54: Influence of earthquake magnitude for (a) for A=0.15g and W.T.= 0.0 m (b) for A=0.35g and W.T.=
0.0 m by T&Y method for zone V

133

Further, the graphical representation based on T&Y method shows that for varying water
table and ground acceleration increase in magnitude always lead to increase the chances of
liquefaction. A substantial finding have been observed that, the variation in liquefaction
potential was 27.778% with the increase in earthquake magnitude from 6.0 to 7.0 and
variation was 15.079% with the increase of earthquake magnitude from 7.0 to 8.0 at
horizontal acceleration 0.15g when water table is at ground level (Fig. 5.54(a)). Another
observation was that, chances of liquefaction with variation of magnitude increases when
horizontal ground acceleration is increasing. The variations was found at horizontal
acceleration 0.35g is very less i.e. 0.794% between 6.0 to 7.0 and 7.0 to 8.0 magnitude of
earthquake (Fig. 5.54(b)).
Influence of Ground Acceleration:In line of the above discussion the increase in acceleration also shows the dominating
tendency of the acceleration as it directly influences the chances of liquefaction. It could be
seen from Fig. 5.55 (a) that for magnitude 7.0 when water table at ground level the chances of
liquefaction between two points i.e. point 0.15g and 0.25g is 26.19% and the chances of
liquefaction further increase between 0.25g to 0.35g by 26.195%. The variation with the
chances of liquefaction shows linear trend which means in case of increase in acceleration the
regular/linear behavior be obtained by Modified Seeds approach. Furthermore at earthquake
magnitude 8 there is marginal increase in chances of liquefaction when ground acceleration
increases from 0.25 to 0.35 at 2.0m depth of water table.

134

70.0

Modified Seed's method

Modified Seed's method

at depth of W.T.=4.0 m & M=7.0

Percentage of Soil Liquify (%)

Percentage of Soil Liquify (%)

80.0
60.0
50.0
40.0
30.0
20.0
10.0
0.0

0.1
0.2
0.3
0.4
Variation of Horizontal Acceleration (g)

at depth of W.T.= 2.0 m & M=8.0

80.0
70.0
60.0
50.0
40.0
30.0
20.0
10.0
0.0

0.1
0.2
0.3
0.4
Variation of Horizontal Acceleration (g)

(a)

(b)

Fig. 5.55: Influence of Ground Acceleration (a) for W.T. = 4.0m and for M=7.0 (b) for W.T.=2.0m and M=8.0
by modified Seeds method for zone V.

By I&B approach chances of liquefaction considering the variation of acceleration shows

maximum variation for earthquake magnitude 6.0 and 2.0m water table depth (Fig. 5.56(a)).
It is having a linear variation of liquefaction potential between two points i.e. 0.15g and 0.25g
and 0.25g and 0.35g. In this case, also the chances of liquefaction increasing with the
increase in acceleration but the increase in percentage of liquefaction b/w two points is
constant. It shows the typical behavior of this approach.
Consider earthquake magnitude 8 for different depths variation in percentage liquefaction
potential is rapid for the horizontal acceleration for 0.15g to 0.25g but for next interval it is
constant for 0.25g to 0.35g (Fig. 5.56(b)). Liquefaction potential increases between 0.15g to
0.25g by 37.301% and then becomes constant between 0.25g to 0.35g for magnitude 8 when
water table is at 2.0m depth.

135

at depth of W.T.= 2.0 m and M=6.0

Percentage of Soil Liquify(%)

Percentage of Soil Liquify(%)

60.0

I&B method

I&B method

70.0

50.0
40.0
30.0
20.0
10.0
0.0
0.1

0.2

0.3

80.0
75.0
70.0
65.0
60.0
55.0
50.0
45.0
40.0
35.0
30.0

at depth of W.T.=2.0 m at M=8.0

0.1

0.4

0.2

0.3

0.4

Variation of Horizontal Acceleration (g)

Variation of Horizontal Acceleration (g)

(a)

(b)

Fig. 5.56: Influence of Ground Acceleration for (a) W.T. = 2.0m and M=6.0 (b) W.T.=2.0m and M=8.0 by I&B
method for zone V.

Considering the variation of acceleration by T&Y approach chances of liquefaction are

increasing with the increase in acceleration but the increase in percentage of liquefaction b/w
two points is reducing. The difference between 0.15g to 0.25g the chances of liquefaction is
50.794% and between 0.25g and 0.35g marginal increase of 4.761% has been observed for
water table ground level and earthquake magnitude 6 (Fig. 5.57 (a)). Fig 5.57 (b) shows
liquefaction potential follows a linear trend with very less variation. In this the liquefaction is
analyzed at ground level and for earthquake magnitude 8 which shows increase of 1.01%
between points 0.15g and 0.25g and increase of 1.00% between 0.25g and 0.35g which is

90.0
80.0
70.0
60.0
50.0
40.0
30.0
20.0
10.0
0.0

T&Y method

Percentage of Soil Liquify (%)

Percentage of Soil Liquify(%)

very less.

at M= 6.0 and W.T.: 0.0 m

0.1
0.2
0.3
0.4
Variation of Horizontal Acceleration (g)

T&Y method

80.4
80.2
80.0
79.8
79.6
79.4
79.2
79.0
78.8
78.6
78.4

at M=8.0 and W.T.: 0.0 m

0.1

0.2

0.3

0.4

Variation of Horizontal Acceleration (g)

(a)

(b)
136

Fig. 5.57: Influence of Ground Acceleration for (a) W.T. = 0.0m and M=8.0 (b) W.T. = 0.0m and M=8.0 by
T&Y method for zone V.

Influence of Water Table:The influence of ground water table in determination of liquefaction potential by modified
Seeds method can be seen for constant horizontal ground acceleration and earthquake
magnitude in Fig 5.58. It could be seen for zone-V hat chances of liquefaction are increasing
up to depth of 2m and then reducing for the depth of water table 4m, 6m & 8m. Similar to
earlier cases maximum value is observed at 2m depth for ground accelerations 0.15g, 0.25g
and 0.35g and earthquake magnitudes viz. 6.0, 7.0, and 8.0. Fig 5.58(a) shows the variation of
soil liquefaction in percent for horizontal ground acceleration 0.35g and earthquake
magnitude 6.0. It shows the increase in liquefaction potential from 0 m water table to 2.0 m
water table i.e. 39.683% to 52.381% and then decreases for water table 4m, 6m and 8m with

55.0

Modified Seed's method

Percentage of Soil Liquify(%)

Percentage of Soil Liquefy(%)

minor variation.

at M= 6.0 A = 0.35g

50.0
45.0
40.0
35.0
30.0
0

2
4
6
Depth of Water Table (m)

75.6

Modified Seed's method

75.4

at M= 8.0 and A = 0.35g

75.2
75.0
74.8
74.6
74.4
0

(a)

2
4
6
Depth of Water Table (m)

(b)

Fig. 5.58: Influence of Depth of Water Table for (a) A=0.35g and M=6.0 (b) A=0.35g and M=8.0 by modified
Seeds method for zone V.

137

I&B method

70.0

at M=7.0 and A = 0.25g

Percentage of Soil Liquify (%)

Percentage of Soil Liquify (%)

80.0
60.0
50.0
40.0
30.0
20.0
0

2
4
6
Variation of Water Table (m)

75.6

I&B method

75.4

at M= 8.0 and A = 0.35g

75.2
75.0
74.8
74.6
74.4

(a)

2
4
6
Variation of Water Table (m)

(b)

Fig. 5.59: Influence of Depth of Water Table for (a) A=0.25g and M=7.0 (b) A=0.35g and M=8.0 by I&B
method for zone V.

Fig 5.58(b) depicts the curve of variation in liquefaction potential at horizontal ground
acceleration 0.35g and earthquake magnitude 8.0. A constant value of 75.4% has been
observed between 2.0 m to 4.0 m of water table and attains a minimum value of 74.6% at all
other depth.
I&B method shows in Fig. 5.59(a) a decreasing trend of liquefaction potential with the depth
of water table. The maximum value of liquefaction potential has been observed at ground
level i.e. 71.429%. However, in Fig 5.59(b) liquefaction potential decreases marginally from

81.0
80.0
79.0
78.0
77.0
76.0
75.0
74.0

Percentage of Soil Liquify(%)

Percentage of Soil Liquify(%)

75.4% to 74.6% up to 2m depth and then becomes constant.

T&Y method
at M= 8.0 and A = 0.35g

2
4
6
Variation of Water Table

70.0

T&Y method

60.0

at M= 8.0 and A = 0.15g

50.0
40.0
30.0
20.0
10.0
0.0
0

(a)

2
4
6
Variation of Water Table

(b)

Fig. 5.60: Influence of Depth of Water Table for (a) A=0.35g and M=8.0 (b) A=0.15g and M=8.0 by T&Y
method for zone V.

In case of T&Y method, liquefaction potential decreases with the decrease in the depth of
water table. The maximum chances of liquefaction potential occur at the ground level when
138

ground acceleration is 0.35g and earthquake magnitude is 8.0 i.e. 80.159%. Then it shows a
decreasing trend with the water table and remains constant for water table 4.0-6.0 m at
75.379% and further decreases to 74.603% at water table 8m (Fig. 5.60 (a)). It can be seen in
the Fig. 5.60 (b) that the liquefaction potential is 65.873% at ground surface and it decreases
to 10% when water table is at 4.0 m. It remains constant for water table 6.0 m and 8.0 m.

5.2.6 Liquefaction Potential by Empirical Methods for Combined Zone

Combined zone study comprises 175 borehole data or 644 datasets. Basically this study is the
combination of all zones.
Fig. 5.61 for liquefaction occurrence based on modified Seeds method also observed that
initially by decreasing of water table chances of liquefy increases up to a depth of 2m after
that decreases & this happen only when soil having S.P.T.-N value less at depth 2.0m. S.P.T.N value depends on various parameters of soil like density of soil, water content of soil,
particle size at that level etc. Maximum soil liquefy at M8 A0.35 W2 i.e. 88% (approx.).
Fig. 5.62 based on I & B method. In this figure we observed that M6 A0.15 W4; M6 A0.15 W6;
M6 A0.15 W8 i.e. those three cases showed did not any chances of soil liquefy. But at
earthquake magnitude 8.0 and horizontal acceleration 0.35g, we observe that the chances of
soil liquefy decreases with the decreasing depth of water table but decreasing rate would be
very slow with respect to others. But average chances of soil liquefy is 42.219%.
Fig. 5.63 observes that the average chances of soil liquefy is 45.1946% by T & Y method for
combined zone. In this graph it is clearly showed that the model M6 A0.35 W0; M7 A0.35 W0; M8
A0.25 W0; M8 A0.35 W0 crosses the chances of liquefaction potential by 70% that means approx.
580 datasets goes to the direction of liquefaction on above combination.

139

M6A0.15W0
M6A0.15W2
M6A0.15W4
M6A0.15W6
M6A0.15W8
M6A0.25W0
M6A0.25W2
M6A0.25W4
M6A0.25W6
M6A0.25W8
M6A0.35W0
M6A0.35W2
M6A0.35W4
M6A0.35W6
M6A0.35W8
M7A0.15W0
M7A0.15W2
M7A0.15W4
M7A0.15W6
M7A0.15W8
M7A0.25W0
M7A0.25W2
M7A0.25W4
M7A0.25W6
M7A0.25W8
M7A0.35W0
M7A0.35W2
M7A0.35W4
M7A0.35W6
M7A0.35W8
M8A0.15W0
M8A0.15W2
M8A0.15W4
M8A0.15W6
M8A0.15W8
M8A0.25W0
M8A0.25W2
M8A0.25W4
M8A0.25W6
M8A0.25W8
M8A0.35W0
M8A0.35W2
M8A0.35W4
M8A0.35W6
M8A0.35W8

Chances of Soil Liquefy (%)

100

Modified Seed's Method for Combined Zone

90

Av. Chances of Soil Liquefy: 35.172%

80

70

60

50

40

30

20

10

Types of Model

Fig. 5.61: Chances of Percentage of Soil Liquefy under Different Conditions by Modified Seeds Method for combined zone.

140

M6A0.15W0
M6A0.15W2
M6A0.15W4
M6A0.15W6
M6A0.15W8
M6A0.25W0
M6A0.25W2
M6A0.25W4
M6A0.25W6
M6A0.25W8
M6A0.35W0
M6A0.35W2
M6A0.35W4
M6A0.35W6
M6A0.35W8
M7A0.15W0
M7A0.15W2
M7A0.15W4
M7A0.15W6
M7A0.15W8
M7A0.25W0
M7A0.25W2
M7A0.25W4
M7A0.25W6
M7A0.25W8
M7A0.35W0
M7A0.35W2
M7A0.35W4
M7A0.35W6
M7A0.35W8
M8A0.15W0
M8A0.15W2
M8A0.15W4
M8A0.15W6
M8A0.15W8
M8A0.25W0
M8A0.25W2
M8A0.25W4
M8A0.25W6
M8A0.25W8
M8A0.35W0
M8A0.35W2
M8A0.35W4
M8A0.35W6
M8A0.35W8

Chances of Soil Liquefy

100

I&B Method for Combined Zone

90

Av. Chances of Soil Liquefy: 42.219%

80

70

60

50

40

30

20

10

Types of Model

Fig. 5.62: Chances of Percentage of Soil Liquefy under Different Conditions by I&B Method for combined zone

141

M6A0.15W0
M6A0.15W2
M6A0.15W4
M6A0.15W6
M6A0.15W8
M6A0.25W0
M6A0.25W2
M6A0.25W4
M6A0.25W6
M6A0.25W8
M6A0.35W0
M6A0.35W2
M6A0.35W4
M6A0.35W6
M6A0.35W8
M7A0.15W0
M7A0.15W2
M7A0.15W4
M7A0.15W6
M7A0.15W8
M7A0.25W0
M7A0.25W2
M7A0.25W4
M7A0.25W6
M7A0.25W8
M7A0.35W0
M7A0.35W2
M7A0.35W4
M7A0.35W6
M7A0.35W8
M8A0.15W0
M8A0.15W2
M8A0.15W4
M8A0.15W6
M8A0.15W8
M8A0.25W0
M8A0.25W2
M8A0.25W4
M8A0.25W6
M8A0.25W8
M8A0.35W0
M8A0.35W2
M8A0.35W4
M8A0.35W6
M8A0.35W8

Chances of Soil Liquefy (%)

100

T&Y Method for Combined Zone

90

Av. Chances of Soil Liquefy: 45.195%

80

70

60

50

40

30

20

10

Types of Model

Fig. 5.63: Chances of Percentage of Soil Liquefy under Different Conditions by T&Y Method for combined zone

142

Influence of Earthquake Magnitude:In this section, we discuss the effect due to earthquake magnitude, depth of water table and
horizontal acceleration for all datasets i.e. zone-I to zone-V. There are 644 datasets. At
earthquake magnitude 6.0 chances of liquefaction occurrence does not greater than 50% in any
considered conditions for wholly datasets. The effect of chances of liquefaction prone areas due
to earthquake magnitude could be easily depicted from the fig. 5.64 (a, b) which shows that the
variation in liquefaction potential significantly depends on magnitude. On the basis of the results
obtained we found that the chances of liquefaction potential is almost for same at 0.0 m & 2.0 m
depth of water table from ground surface at horizontal acceleration 0.15g at earthquake
magnitude 6.0, 7.0 and 8.0. Generally the series found is of convex nature on the influence of
earthquake magnitude. But in two cases i.e. at depth of water table at 2.0m & 4.0m and
horizontal acceleration 0.35g nature of curve reverses and shows maximum deviation i.e.
chances of liquefaction potential increases by 37.733% due to increases earthquake magnitude
by 6.0 to 7.0 and liquefaction potential increases by 19.099% due to increases earthquake
magnitude by 7.0 to 8.0 (fig. 5.64 (b)).

90.0
80.0
70.0
60.0
50.0
40.0
30.0
20.0

Modified Seed's method

at W.T.=8.0 m and A=0.15g

20.0
15.0
10.0
5.0
0.0
5.5

6.5

7.5

8.5

Percentage of Soil Liquify (%)

Percentage of Soil Liquify (%)

25.0

Variation of Earthquake Magnitude

(a)

Modified Seed's method

at W.T.= 4.0 m A=0.35g

5.5

6.5
7.5
8.5
Variation of Earthquake Magnitude

(b)

Fig. 5.64: Influence of Magnitude for (a)W.T. = 8.0m and A=0.15g and (b) W.T. = 4.0m and A=0.35g by modified
Seeds method for combined site.

143

The influence of earthquake magnitude follow previous pattern by I&B method. From the graph
rise in liquefaction potential between two consecutive points 6 and 7 is 1.250% and 10.938%
between points 7 and 8 at horizontal acceleration 0.15g (Fig. 5.65 a) which is least variation due
to earthquake magnitude. Fig. 5.65 (b) shows that the chances of liquefaction are greater than
75% but variation due to earthquake magnitude is also less. Out of 15 plotted graphs 7 graphs
follow convex pattern and 8 graphs shows concave pattern. At horizontal acceleration 0.15g and

14.0

88.0

I&B method

12.0

Percentage of Soil Liquify (%)

Percentage of Soil Liquify (%)

0.25g all five graph shows convex type and concave type graph respectively.

at W.T.= 8.0 m and A=0.15g

10.0
8.0
6.0
4.0
2.0
0.0
5.5

6.5

7.5

8.5

86.0

I&B method
at W.T.= 0.0 m and A=0.35g

84.0
82.0
80.0
78.0
76.0
74.0
5.5
6.5
7.5
8.5
Variation of Earthquake Magnitude

Variation of Earthquake Magnitude

(a)

(b)

Fig. 5.65: Influence of Magnitude for (a)W.T. = 8.0m and A=0.15g and (b) W.T. = 0.0m and A=0.35g by I&B
method for combined site.

Further graphical representation based on T&Y method reveals the same trend of variation as
shown by preceding methods. At horizontal acceleration 0.15g Chances of liquefaction are
increases by 32.187% due to increasing in earthquake magnitude by 6.0 to 7.0 and 26.407% due
to increasing in earthquake magnitude by 7.0 to 8.0. This increase in two steps keeps minor
difference with respect to increment hence the graph (Fig. 5.66 (a)) shows linear type graph.
Influence due to earthquake magnitude at horizontal acceleration 0.35g at ground level however
percentage rise in liquefaction is too less between the magnitudes which is 1.875% between 6.0
and 7.0 and about 1.250% between 7.0 and 8.0. There is high probability of chances of

144

liquefaction at moment magnitude 6 i.e. 90.469% after that the graph (Fig. 5.66 (b)) shows minor

80.0
70.0
60.0
50.0
40.0
30.0
20.0
10.0
0.0

T&Y method

Percentage of Soil Liquify (%)

Percentage of Soil Liquify (%)

increments due to increase in earthquake magnitude.

at W.T.= 0.0 m and A=0.15g

5.5

94.0
93.5
93.0
92.5
92.0
91.5
91.0
90.5
90.0

T&Y method
at W.T.=0.0 m and A=0.35g

5.5
6.5
7.5
8.5
Variation of Earthquake Magnitude

6.5
7.5
8.5
Variation of Earthquake Magnitude

(a)

(b)

Fig. 5.66: Influence of Magnitude for (a)W.T. = 0.0m and A=0.15g and (b) W.T. = 0.0m and A=0.35g by I&B
method for combined site.

Influence of Ground acceleration:In the line of above discussion analysis of variation of liquefaction potential with respect to
ground acceleration is done by all the three methods. It could be seen from fig. 5.67(a) that
maximum value of liquefaction potential is 23.758% at ground level for earthquake magnitude 6
however in fig. 5.67(b) value of liquefaction is comparatively high which increases by
45.031%between 0.15g and 0.25g and about11.335%between points 0.25g and 0.35g attaining
maximum value of 88.043%.It is clearly visible from the graph the increase in horizontal ground
acceleration also leads to increase in chances of liquefaction.

145

Modified Seed's method

25.0
20.0
15.0
10.0
5.0
0.0
0.1

0.2

0.3

Modified Seed's method

100.0

at W.T,= 0.0 m & M=6.0

Percentage of Soil Liquify %

Percentage of Soil Liquify %

30.0

at W.T.= 2.0 m & M=8.0

90.0
80.0
70.0
60.0
50.0
40.0
30.0
20.0

0.4

0.1

Variation of Horizontal Acceleration (g)

0.2
0.3
0.4
Variation of Horizontal Acceleration (g)

(a)

(b)

90.0
80.0
70.0
60.0
50.0
40.0
30.0
20.0
10.0
0.0

I&B method

I&B method

at W.T.= 8.0 m & M=7.0

Percentage of Soil Liquify %

Percentage of Soil Liquify %

Fig. 5.67: Influence of Ground Acceleration for (a) W.T. = 0.0m and M=6.0 & (b) for W.T. = 2.0m M=8.0 by
modified Seeds method for combine site.

0.1
0.2
0.3
0.4
Variation of Horizontal Acceleration (g)

90.0
80.0
70.0
60.0
50.0
40.0
30.0
20.0
10.0
0.0

at W.T.= 6.0 m & M=8.0

0.1
0.2
0.3
0.4
Variation of Horizontal Acceleration (g)

(a)

(b)

Fig. 5.68: Influence of Ground Acceleration for (a) W.T. = 8.0m and M=7.0 and (b) W.T=6.0and M=8.0 by I&B
method for comb site.

Evaluation of liquefaction by I&B method shows that increasing variation of liquefaction which
increases by 26.719% between points 0.15g and 0.25g and similarly by 54.531%between 0.25g
and 0.35g. Maximum value of liquefaction observed is 82.5%when depth of water table is 8m
and earthquake magnitude 7. Likewise in fig. 5.68(b) there is 69.218% rise between points 0.15g
and 0.35g for magnitude 8 and water table at 6.0 m from ground level.

146

Percentage of Soil Liquify (%)

T&Y method

100.0
90.0
80.0
70.0
60.0
50.0
40.0
30.0
20.0
10.0
0.0

Percentage of Soil Liquify (%)

at M= 6.0 and W: 0.0 m

T&Ymethod

100.0
95.0
90.0
85.0
80.0
75.0
70.0
65.0
60.0

at W: 0.0 m at M=8.0

0.1
0.1

0.2
0.3
0.4
Variation of Horizontal Acceleration (g)

0.2

0.3

0.4

Variation of Horizontal Acceleration (g)

(a)

(b)

Fig. 5.69: Influence of Ground Acceleration for (a) W.T. = 0.0m and M=6.0, (b) for W.T=0.0and M=8.0 by T&Y
method for comb site.

Furthermore analysis is done by T&Y method which shows variation of 71.406% between two
consecutive point 0.15g and 0.25g and 7.344% between points 0.25g and 0.35g at ground level
and magnitude 6.0. For fig. 5.69(b) liquefaction is quite high at 0.15g ground acceleration which
increases from 70.313% to 93.594%between 0.15g and 0.35g for moment magnitude 8.
Influence of Water Table:Depth of ground water table plays an important role in estimation of liquefaction potential. The
influence of ground water table in determination of liquefaction potential by modified Seeds
method can be seen for variation of water table with respect to constant horizontal ground
acceleration and earthquake magnitude in Fig. 5.70 (a & b). From the graph it can be seen that
maximum value of liquefaction is observed at 2.0 m depth of water table thereby identifying a
highly susceptible liquefaction zone. In fig. 5.70(a) 42.857% liquefaction is identified which
goes on decreasing with depth of water table for ground acceleration 0.35g and magnitude 6
similarly 88.043% liquefaction percentage is seen from fig. 5.70(b) for ground acceleration
0.35g and magnitude 8. From the graphs it can clearly visualized with decrease in depth of water

147

table chances of liquefaction are also minimized however maximum value is attained at depth

50.0

Modified Seed's method

89.0

Modified Seed's method

45.0

at A = 0.35g and M=6.0

88.0

at A = 0.35g & M=8.0

Percentage of Soil Liquify (%)

Percentage of Soil Liquify (%)

2.0 m.

40.0
35.0
30.0
25.0
20.0
15.0

87.0
86.0
85.0
84.0
83.0
82.0
81.0

2
4
6
Variation of Water Table

2
4
6
Variation of Water Table

(a)

(b)

Fig. 5.70 : Influence of Depth of Water Table for (a) A=0.35g and M=6.0 (b) A=0.35g and M=8.0 by modified
Seeds method for comb site.

It is apparent from Fig. 5.71 (a & b) that the chances of liquefaction potential of soil decreases
with the increasing of water table and no peak is observed as in previous one. The falling trends
were found up to depth of 4.0 m and attains minimum value of 0 up to depth 8m.Maximum value
is observed at ground level 13.906% in fig 5.71(a) when ground acceleration is 0.15g and
magnitude 6 similarly falling trend is observed in fig 5.71(b) from maximum value of 82.344%
at ground level and 60.156% at 8.0m depth for ground acceleration 0.25g and magnitude 8.
Percentage of Soil Liquify (%)

Percentage of Soil Liquify (%)

I&B method

16.0
14.0
12.0
10.0
8.0
6.0
4.0
2.0
0.0

at M= 6.0 & A = 0.15g

2
4
6
Variation of Water Table

85.0

I&B method

80.0

at M= 8.0 and A = 0.25g

75.0
70.0
65.0
60.0
55.0
50.0

148

2
4
6
Variation of Water Table

(a)

(b)

Fig. 5.71 : Influence of Depth of Water Table for (a) A=0.35g and M=6.0 and (b) A=0.35g and M=8.0 by I&B
method for comb site.

T&Y method

94.0

at A = 0.35g and M=6.0

Percentage of Soil Liquify (%)

Percentage of Soil Liquify (%)

96.0

92.0
90.0
88.0
86.0
84.0
82.0
0

2
4
6
Variation of Water Table

50.0
45.0
40.0
35.0
30.0
25.0
20.0
15.0
10.0
5.0
0.0

T&Y method
at M=7.0 and A = 0.15g

(a)

2
4
6
Variation of Water Table

(b)

Fig. 5.72: Influence of Depth of Water Table for (a) A=0.35g and M=6.0 (b) A=0.15g and M=7.0 by T&Y method
for comb site.

The trend of variation of chances of liquefaction potential of soil is similar to modified Seeds
method and I&B method. Maximum chances of liquefaction potential occur at the depth of water
table is at ground surface with the combination of ground acceleration A=0.35g and earthquake
magnitude M= 8.0 i.e. 93.594% Fig. 5.72 (a). It is also shown in the Fig. 5.72 (b) that the
liquefaction chance is 43.906% at depth of water table is at ground surface and it is suddenly
decreases continuously up to depth 8.0 m.

149

5.3 Liquefaction Potential by Computational Methods

5.3.1 L.P. by ANN, ANFIS & MLR Models of Zone-I
To estimate the liquefaction potential of soil for Zone-I which contains 137 datasets (from 40
boreholes) were used to develop the computational model. Out of 137 datasets, 27 datasets (from
8 boreholes) were reserved for validating the different developed model using ANN, ANFIS and
MLR techniques. The datasets of eight boreholes i.e. BH- 4, 8, 12, 19, 21, 24, 34 and 38 were
reserved as mentioned in chapter 4.
Table5.1: Zone wise List of Models in which percentage of chances of soil liquefies is more susceptible
through various approaches.

Zone

Zone I

Zone II

Zone III

Zone IV

Zone V

Combined
Zone

Semi-Empirical
Approach
Modified Seeds

M8A0.35W2

M8A0.35W4

M8A0.35W6

I&B

M8A0.35W0

M8A0.35W2

M7A0.35W0

T&Y

M8A0.35W0

M7A0.35W0

M6A0.35W0

Modified Seeds

M8A0.35W2

M8A0.35W0

M8A0.35W4

I&B

M8A0.35W0

M7A0.35W0

M8A0.25W0

T&Y

M8A0.35W0

M7A0.35W0

M6A0.35W0

Modified Seeds

M8A0.35W2

M8A0.35W4

M8A0.35W6

I&B

M8A0.35W0

M8A0.35W2

M8A0.35W4

T&Y

M7A0.35W0

M8A0.35W0

M8A0.25W0

Modified Seeds

M8A0.35W2

M8A0.35W4

M8A0.35W6

I&B

M7A0.35W0

M8A0.35W2

M8A0.35W4

T&Y

M8A0.35W0

M8A0.25W0

M7A0.35W0

Modified Seeds

M8A0.35W2

M8A0.35W4

M7A0.35W0

I&B

M8A0.35W0

M8A0.25W0

M7A0.35W0

T&Y

M8A0.35W0

M7A0.35W0

M8A0.25W0

Modified Seeds

M8A0.35W2

M8A0.35W4

M8A0.35W6

I&B

M8A0.35W0

M8A0.35W2

M8A0.35W4

T&Y

M8A0.35W0

M7A0.35W0

M6A0.35W0

Model Description

150

5.3.1.1. Results of Developed Models Based on Modified Seeds Method

It has been already mentioned that 110 data sets were used to train the ANN & ANFIS models
and remaining 27 datasets were used to validate the developed model. The analyses of prediction
capability of computational models are based on average absolute error (AAE) & root mean
square error (RMSE) values (Annexure-C). AAE and RMSE were calculated for liquefaction
potential obtained from computational methods in comparison to actual values of modified
seeds method for all 45 combination of MxAyWz. In this section few results of some best models
and models with maximum chances of liquefaction have been discussed in Table 5.1.
It can be observed from Table C-1 that only four incorrect predictions were made by ANN,
remaining results shows that liquefaction potential from ANN was in accordance with modified
Seeds method. Four incorrect predictions are at serial no. 9, 26 in model M8A0.35W2 and at serial
no. 13 &16 in model M8A0.35W4. ANN model M8A0.35W6 showed best predictive capability in
which all 27 results were predicted in accordance with actual liquefaction potential obtained
from modified seeds method.
Results of the other developed ANN models have the good prediction ability. The remaining 42
combinations of ANN models have predicted more/less in accordance with modified seeds
method. This indicates that ANN models are useful and can be successfully applied for
determination of liquefaction potential.
The minimum AAE & RMSE are 0.047 & 0.214 obtained by ANN model M6A0.15W8. In contrast
to minimum AAE & RMSE, ANN model M8A0.35W6 have AAE=18.744% and RMSE=45.236%.
Above discussion indicates that AAE & RMSE are high for some ANN models. Similarly,
ANFIS models to estimate L.P. were developed for all 45 combinations. The minimum and
maximum AAE & RMSE by ANFIS models are same as obtained by ANN models i.e. minimum

151

AAE & RMSE is obtained by ANFIS model M6A0.15W8 and maximum AAE & RMSE by
ANFIS model M8A0.35W6. Four incorrect predictions were also done which can be seen in table
C-4 at serial no. 9, 16 and 26 in model M8A0.35W4 and serial no. 16 in model M8A0.35W6.
Therefore it may be said that the ANN & ANFIS models have the prediction capability more
than 95% (Table C-4).
As discussed previously that in addition to computation models multiple linear regression models
were also developed for evaluation of L.P. Hence MLR model for selected combinations of
MxAyWz were developed as it was unreasonable to develop MLR models for entire combinations
of MxAyWz. The developed MLR models were M8A0.35W2, M8A0.35W4 & M7A0.35W6 to estimate
L.P. by these three models following equations can be used.
For M8A0.35W2:
L.P.= 0.7888N-0.3417m-0.1701t+0.1660D0.075-0.0814wn-0.2458D2.0-0.2572D0.002-0.2631+0.3657..(1)

For M8A0.35W4:
L.P.= 0.9039N-0.4013m-0.2050t+0.0836D0.075-0.0143wn+0.0376D2.0-0.0145D0.002+0.0322+0.0720.(2)

For M8A0.35W6:
L.P.= 0.9355N-0.2229m-0.3518t+0.0815D0.075-0.0675wn-0.1366D2.0+0.0136D0.002-0.0319+0.2773.(3)

The calculated AAE & RMSE are shown in Table C-7 has been observed that AAE & RMSE
are higher in MLR models in comparison to soft computing models.

152

ANN

L.P. by Modified Seed's Method

L.P. by Modified Seed's Method for M8A0.35W2for Zone-I

1.6
From ANN Methods From ANFIS Method
1.4
y = 0.758x + 0.0381
y = 1.0604x +0.0371
R = 0.7861
1.2
R = 0.6909
1
0.8
0.6
0.4
0.2
0
0.2
0.4
0.6
0.8
1
-0.2 0
-0.4
L.P. by ANN, ANFIS and MLR Technique

ANFIS
MLR

From MLR Technique

y = 0.9655x + 0.0042
R = 0.7980
1.2

1.4

1.6

Fig.5.73: Comparative Study of L.P. between Modified Seeds Method and Various Computational Methods i.e.
ANN, ANFIS and MLR for Zone I for Model M8A0.35W2.

The comparative study between ANN, ANFIS and MLR models were carried out for some
selected combination of models M8A0.35W2 (Fig. 5.73), M8A0.35W4

(Fig. 5.74) & M7A0.35W6

(Fig. 5.75). it can be seen from figures that coefficient of determination i.e. R2 is higher for two
MLR models i.e. for M8A0.35W2 and M8A0.35W6 by MLR technique but in Fig. 5.74 showed that
the ANN model value for model M8A0.35W4 R2 values is high.
L.P. by Modified Seed's Method for M8A0.35W4for Zone-I

L.P. by Modified Seed's Method

1.6

From ANN Methods From ANFIS Method

y = 1.0065x -0.0028 y = 0.8141x + 0.0459
R = 0.8827
R = 0.9585

1.4
1.2
1
0.8
0.6
0.4

From MLR Technique

y = 0.8132x + 0.0627
R = 0.9431

0.2

ANN
ANFIS
MLR

0
-0.2

0.2

0.4
0.6
0.8
1
L.P. by ANN, ANFIS and MLR Technique

1.2

1.4

1.6

Fig. 5.74: Comparative Study of L.P. between Modified Seeds Method and Various Computational Methods i.e.
ANN, ANFIS and MLR for Zone I for Model M8A0.35W4.

153

L.P. by Modified Seed's Method

L.P. by Modified Seed's Method for M8A0.35W6 for Zone-I

1.6

From ANN Methods

y = 0.8005x +0.128
R = 0.8079

1.4
1.2

From ANFIS Method

y = 0.7926x + 0.1161
R = 0.8713

ANN
ANFIS
MLR

1
0.8
0.6
0.4

From MLR Technique

y = 0.8985x + 0.0364
R = 0.9591

0.2
0
0

0.2

0.4
0.6
0.8
1
1.2
1.4
1.6
L.P. by ANN, ANFIS and MLR Technique
Fig. 5.75: Comparative Study of L.P. between Modified Seeds Method and Various Computational Methods i.e.
ANN, ANFIS and MLR for Zone I for Model M8A0.35W6.

5.3.1.2 Results of Developed Models based on I&B Method

The developed model M8A0.35W0, M8A0.35W2 & M7A0.35W0 results have been discussed in this
section (Table 5.1). Total 27 datasets were used to validate computational models. As per Table
C-2 liquefaction did not occur in four, six and five datasets for the model M8A0.35W0, M8A0.35W2
& M7A0.35W0.
Models like M6A0.15W4, M6A0.15W6 & M6A0.15W8 (Table C-2) did not show any chance of
liquefaction. It is to mention here that the above discussion and subsequent discussion is based
on the analysis of validated datasets as it is not possible to compare the results based on training
datasets. Since they are great in numbers & validation datasets itself give an impression of
overall performance of models.
Predicted values of liquefaction potential by ANN model for all 45 combinations are shown in
Appendix -C. Using actual and predicted values, statistical parameters such as AAE and RMSE
have been calculated for each combination and these results are summarized in Appendix-D. The
lowest AAE & RMSE have been yielded by model M6A0.15W8 and highest AAE & RMSE by

154

model M8A0.15W0. Similarly lowest RMSE = 0.1375% was yielded by model M6A0.15W8 and
highest RMSE = 32.850% was yielded by model M8A0.15W0.
Table 5.2: AAE and RMSE values of ANN models based on I&B method

Model

M8A0.35W0

M8A0.35W2

M7A0.35W0

AAE

12.486%

16.632%

20.479%

RMSE

19.728%

28.487%

30.563%

Some of the predictions are incorrect as given in table C-2 at serial no. 16 for model M8A0.35W0,
at serial no. 13, 16 and 17 in model M8A0.35W2 and at serial no. 27 for model M7A0.35W0. 5 cases
have been found where predictions by developed ANN models are incorrect. The mean
prediction accuracy of ANN models are greater than 90%.
Liquefaction potential estimated by ANFIS models with AAE and RMSE are shown below in
Table 5.3.
Table5.3: AAE and RMSE value for most vulnerable combinations by I&B method for site-I.

Model

M8A0.35W0

M8A0.35W2

M7A0.35W0

AAE

15.410%

17.380%

14.700%

RMSE

25.970%

28.890%

20.040%

From Table C-5 it can be seen that the prediction by M8A0.35W0 model at serial no. 16 is not
correct. Similarly prediction by M7A0.35W0 model is incorrect at serial no. 8, 16 and 25 whereas
for model M8A0.35W2 all 27 predictions are correct. Out of 81 predictions, 4 predictions of L.P.
by ANFIS models are incorrect. Therefore it can be said that ANFIS predictions are above 95%.
MLR models for selected combinations of MxAyWz were developed. The developed models are
M8A0.35W0, M8A0.35W2 & M7A0.35W0. AAE & RMSE value of MLR models were compared with
ANN and ANFIS models which is higher in comparison to ANN and ANFIS values of AAE and
RMSE but coefficient of determination is reached up to 0.9526.
155

The developed relationship by MLR models are as follows.

For M8A0.35W0:
L.P. =0.0377+1.2485N-0.2133m-0.0856t-0.0728D0.075+0.1278wn-0.1922 D2.0+0.1391 D0.002+0.0665 ...(4)

For M8A0.35W2:
L.P.= 0.2504+1.1652N-0.4764m-0.1019 t -0.0214D+0.0838 wn -0.1747 D2.0+0.1237 D0.002-0.0067 ....(5)

For M7A0.35W0:

L.P. by I&B Method

L.P.= 0.0297+1.2064N-0.2062m+0.2028t-0.0444D0.075+0.0406wn-0.1210 D2.0+0.1525 D0.002 + 0.0432 (6)

1.8
1.6
1.4
1.2
1
0.8
0.6
0.4
0.2
0
-0.2 0
-0.4

L.P. by I&B Method for M8A0.35W0 for Zone I

From ANN Methods
y = 1.0259x - 0.0009
R = 0.9612

From ANFIS Method

y = 0.9707x + 0.0201
R = 0.9243

From MLR Technique

y = 0.8789x + 0.2138
R = 0.9292

ANN
ANFIS
MLR
0.2

0.4
0.6
0.8
1
L.P. by ANN, ANFIS and MLR Technique

1.2

1.4

1.6

Fig. 5.76: Comparative Study of L.P. between I&B Method and Various Computational Methods i.e. ANN, ANFIS
and MLR for Zone I for Model M8A0.35W0.

L.P. by I&B Method for M8A0.35W2 for Zone I

1.6

From ANN Methods

y = 0.8261x - 0.0623
R = 0.8774

L.P. by I&B Method

1.4
1.2

From ANFIS Method

y = 01.021x - 0.0085
R = 0.9651

From MLR Technique

y = 0.8003x + 0.105
R = 0.9526

1
0.8
0.6
0.4
0.2
0
0

0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
L.P. by ANN, ANFIS and MLR Technique
Fig. 5.77: Comparative Study of L.P. between I&B Method and Various Computational Methods i.e. ANN, ANFIS
and MLR for Zone I for Model M8A0.35W2.

156

ANN
ANFIS
MLR

L.P. by I&B Method

L.P. by I&B Method for M7A0.35W0 for Zone I

1.8
1.6
1.4
1.2
1
0.8
0.6
0.4
0.2
0
-0.2 0
-0.4

From ANN Methods

y = 1.0259x - 0.0009
R = 0.9612

0.2

From ANFIS Method

y = 1.0498x - 0.0201
R = 0.9612

0.4

0.6

From MLR Technique

y = 0.6244x + 0.0004
R = 0.8362

0.8

1.2

1.4

1.6

L.P. by ANN, ANFIS and MLR Technique

Fig. 5.78: Comparative Study of L.P. between I&B Method and Various Computational Methods i.e. ANN, ANFIS
and MLR for Zone I for Model M6A0.35W0.

It can be clearly seen from Fig. 5.76, 5.77 and 5.88 that for model M8A0.35W0 & M7A0.35W0
ANN have better coefficient of determination (COD) value whereas for model M8A0.35W2
coefficient of determination value 0.8774. Model M8A0.35W2 showed COD value 0.9526 which
is better than ANN model (Fig. 5.77).
5.3.1.3 Results of Developed Model based on T&Y Method
Using the same combination of MxAyWz the liquefaction occurrences from 137 data sets was
calculated and illustrated in Appendix (B) estimated by T&Y method. The most susceptible
combinations of liquefaction potential were identified as M8A0.35W0, M7A0.35W0 & M6A0.35W0
and the average chances of soil liquefaction by T&Y method for zone-I is 41.590% (Fig. 5.3).
Liquefaction potential predicted by ANN models is shown in Table C-3 for above three
combinations. It can be seen from Table C-3 that chances of liquefaction were very high for all
27 data sets. It can also be seen that liquefaction potential value predicted by ANN was in
accordance with T&Y method for all which indicates strong prediction capability of ANN
models.

157

The predicted liquefaction potential by ANFIS models were also in accordance with T & Y
method. Prediction capability by ANFIS models are also good compared to ANN.
This indicates that both ANN & ANFIS methods have strong training and predicting capability.
The AAE of ANN & ANFIS model M8A0.35W0 is 13.844% and 12.510% respectively. RMSE of
ANN & ANFIS model M7A0.35W0 is 17.020% and 16.580% respectively. However AAE of
ANN and ANFIS models M7A0.35W0 are 9.948% and 15.650%. It indicates that performance of
ANN and ANFIS in terms of statistical parameters was more or less same.
MLR models of M8A0.35W0, M7A0.35W0 & M6A0.35W0 were developed. Multi linear equations for
the developed model are as:
For model M8A0.35W0:
L.P.= -0.0833+0.4508N-0.4512m-0.2187 t +0.3116 D0.075+0.1088 wn +0.0380 D2.0+0.0607 D0.002-0.0480 ...(7)

For model M7A0.35W0:

L.P.= 0.0838+0.4661N-0.3944m-0.4005 t +0.3519 D0.075+0.0951 wn -0.0820 D2.0+0.0189 D0.002-0.0738 .(8)

For model M6A0.35W0:

L.P. = =0.1575+0.4697N-0.7508m-0.5244t+0.4041D0.075+0.1740wn -0.0291 D2.0+0.2737 D0.002+ 0.0322 (9)

L.P. by T&Y Method

1.6
1.4
1.2

ANN
ANFIS
MLR

L.P. by T&Y Method for M8A0.35W0 for Zone-I

From ANN Methods From ANFIS Method
y = 1.1988x +0.0387 y = 0.9702x + 0.0039
R = 0.9439
R = 0.9334

From MLR Technique

y = 1.3699x - 0.0346
R = 0.6067

1
0.8
0.6
0.4
0.2
0
-0.2 0
-0.4

0.2

0.4
0.6
0.8
1
L.P. by ANN, ANFIS and MLR Technique

1.2

1.4

1.6

Fig. 5.79: Comparative Study of L.P. between T&Y Method and Various Computational Methods i.e. ANN, ANFIS
and MLR for Zone I for Model M8A0.35W0.

158

ANN
ANFIS
MLR

L.P. by T&Y Method for M7A0.35W0 for Zone-I

L.P. by T&Y Method

1.6

From ANN Methods

y = 1.0258x -0.012
R = 0.9438

1.4

From ANFIS Method

y = 1.0297x + 0.0176
R = 0.8394

From MLR Technique

y = 1.2894x - 0.0607
R = 0.6425

1.2
1
0.8
0.6
0.4
0.2
0
-0.2

0.2

0.4
0.6
0.8
1
L.P. by ANN, ANFIS and MLR Technique

1.2

1.4

1.6

Fig. 5.80: Comparative Study of L.P. between T&Y Method and Various Computational Methods i.e. ANN, ANFIS
and MLR for Zone I for Model M7A0.35W0.

L.P. by T&Y Method for M6A0.35W0 for Zone-I

L.P. by T&Y Method

1.6

From ANN Methods

y = 1.0463x+0.0161
R = 0.9654

1.4
1.2

From ANFIS Method

y = 0.77214x+0.0476
R = 0.8362

From MLR Technique

y = 1.3932x - 0.1353
R = 0.8083

1
0.8
0.6
0.4
0.2
0
-0.2

0.2

0.4

0.6

0.8

1.2

1.4

1.6

L.P. by ANN, ANFIS and MLR Technique

Fig.5.81: Comparative Study of L.P. between T&Y Method and Various Computational Methods i.e. ANN, ANFIS
and MLR for Zone I for Model M6A0.35W0.

Fig. 5.79, 5.80 and 5.81 showed comparative analysis between results obtained by T&Y method
and results predicted by different developed models. Best linear fit equation and coefficient of
determination are also showed in these figures. From these figures it could be seen that COD

159

value is better in ANN than ANFIS and MLR. MLR showed poor COD value for these three
developed models.

These results depicts that ANN & ANFIS models have better predictive capability in comparison
to MLR models. However regression models have their own advantages such as fast
convergence and generalization etc.

5.3.2 L.P. by ANN, ANFIS & MLR models for Zone-II

To evaluate the liquefaction potential of soil for Zone-II 38 borehole i.e. 110 datasets were used.
Out of 110 datasets 17 datasets (6 bore hole data) were reserved for validating the models
developed by ANN, ANFIS and MLR techniques. The datasets reserved from six boreholes are
BH- 2, 5, 9, 19, 25 and 37.
5.3.2.1 Results of Developed Models based on Modified Seeds Method
Liquefaction potential for zone II calculated from modified Seeds method for all 45
combinations of MxAyWz are summarized in Annexure B. Chances of liquefaction in percentage
by modified Seeds method is illustrated in fig. 5.13. It could be seen from the figure that the
combination of M8A0.35W2, M8A0.35W0 & M8A0.35W4 are having maximum chances of
liquefaction i.e. 84.545%, 82.727% & again 82.727% respectively. Since further discussion is
based on most probable liquefaction model combination as shown in Table 5.1 for zone-II. The
L.P. value obtained by modified Seeds method is shown in Table C-11, C-14 and C-17. There
are two L.P. data in the combination of M8A0.35W2, M8A0.35W0 & M8A0.35W4 which shows no
chances of liquefaction.
93 datasets were used to train the ANN & ANFIS models and remaining 17 datasets were used to
validate the models. It can be seen from Table C-10 that prediction done by ANN models are

160

excellent. Results shows that the liquefaction potential predicted by ANN models were in
accordance with Modified Seeds method. Network architecture N39, N39 and N45 respectively
gave better results for model M8A0.35W2, M8A0.35W0 & M8A0.35W4.
Similar to other results the prediction accuracy of ANN models for remaining 42 combinations
has more/less in accordance with modified seeds method. This indicates that ANN is the useful
method and can be successfully applied for the determination of liquefaction potential.
The minimum AAE & RMSE were 0.047% & 0.214% for ANN models M6A0.15W8. Similarly
maximum AAE & RMSE were obtained for ANN model M8A0.35W6 i.e. 18.744% and 45.236%
respectively. Above discussion indicates that AAE & RMSE are higher for few ANN models
only. These results may be attributed to the less variation in the liquefaction potential values for
such models.
The minimum and maximum AAE & RMSE given by ANFIS model are similar to ANN models
i.e. minimum AAE & RMSE has been given by model M6A0.15W8 and maximum AAE & RMSE
have been given by model M8A0.35W6. It is quiet obvious that the higher statistical measures is
done due to same reason as discussed for ANN models.
MLR models for M8A0.35W2, M8A0.35W0 & M8A0.35W4 combination were developed and AAE &
RMSE is shown in Table C-16. Multi linear equations for the developed models are as:
For model M8A0.35W2:
L.P.= 0.0579+0.9000N-0.6070m-0.1060 t +0.0683 D0.075-0.0127 wn +0.0922 D2.0-0.2351 D0.002-0.0188 ....(10)

For model M7A0.35W0:

L.P. = 0.1003+0.8811N-0.3034m-0.1055 t +0.0067 D0.075-0.0841 wn +0.1261 D2.0-0.1501 D0.002-0.0350 ...(11)

For model M8A0.35W4:

L.P. = 0.1150+0.8572N-0.5277m-0.1185 t +0.0136 D0.075-0.0477 wn +0.1143 D2.0-0.1095 D0.002-0.0015 ...(12)

161

Fig. 5.82, 5.83 and 5.84 showed comparative results of liquefaction potential between validation
results obtained by modified Seeds method and the predicted liquefaction potential by ANN,
ANFIS and MLR models M8A0.35W2, M8A0.35W0 & M8A0.35W4 respectively. COD value of ANN
and MLR models are better than ANFIS models. ANN models M8A0.35W2 is having COD value
of 0.9975 which is best among these models.

L.P. by Modified Seed's Method

ANN
ANFIS
MLR

L.P. by Modified Seed's Method for M8A0.35W2 of Zone-II

1.8

From ANN Methods

y = 1.0041x +0.004
R = 0.9975

1.6
1.4

From ANFIS Method

y = 0.8588x + 0.2326
R = 0.9515

From MLR Technique

y = 0.9852x + 0.0302
R = 0.9849

1.2
1
0.8
0.6
0.4
0.2
0
0

0.2

0.4
0.6
0.8
1
L.P. by ANN, ANFIS and MLR Technique

1.2

1.4

1.6

L.P. by Modified Seed's Method

Fig.5.82: Comparative Study of L.P. between Modified Seeds Method and Various Computational Methods i.e.
ANN, ANFIS and MLR for Zone II for Model M8A0.35W2.
ANN
L.P. by Modified Seed's Method for M8A0.35W0 of Zone-II
1.8
ANFIS
From ANN Methods From ANFIS Method From MLR Technique
MLR
1.6
y = 0.994x +0..0187 y = 0.9247x + 0.0688 y = 0.9487x + 0.0605
R = 0.9876
R = 0.9688
R = 0.9822
1.4
1.2
1
0.8
0.6
0.4
0.2
0
0

0.2

0.4
0.6
0.8
1
L.P. by ANN, ANFIS and MLR Technique

1.2

1.4

1.6

Fig.5.83: Comparative Study of L.P. between Modified Seeds Method and Various Computational Methods i.e.
ANN, ANFIS and MLR for Zone II for Model M8A0.35W0.

162

1.8

L.P. by Modified Seed's Method

ANN

L.P. by Modified Seed's Method for M8A0.35W4 of Zone-II

From ANN Methods
y = 0.9632x +0.0261
R = 0.9778

1.6
1.4

From ANFIS Method

y = 0.9945x -0.089
R = 0.9204

ANFIS

From MLR Technique

y = 0.9442x + 0.0607
R = 0.9811

MLR

1.2
1
0.8
0.6
0.4
0.2
0
0

0.2

0.4
0.6
0.8
1
L.P. by ANN, ANFIS and MLR Technique

1.2

1.4

1.6

Fig.5.84: Comparative Study of L.P. between Modified Seeds Method and Various Computational Methods i.e.
ANN, ANFIS and MLR for Zone II for Model M8A0.35W4.

5.3.2.2 Results of Developed Model based on I&B method

The models M8A0.35W0, M7A0.35W0 & M8A0.25W0 gave highest percentage of liquefaction
occurrences in zone-II i.e. 88.182%, 87.273% and 86.364% (Fig. 5.14). Total 17 datasets were
used to validate these models. 15 cases were found prone to liquefaction in the above three
mentioned combinations. Liquefaction did not occur for remaining 2 cases of validate datasets
(Table C-11).
Five models M6A0.15W4, M6A0.15W6, M6A0.15W8, M7A0.15W6 & M7A0.15W8 did not show any
chances of liquefaction (Fig 5.14). ANN & ANFIS models were developed to predict
liquefaction potential. Estimated values of liquefaction potential from optimal ANN model is
shown in Appendix -II for all 45 combinations. Using actual and predicted values, statistical
parameters such as average absolute error (AAE) and root mean square error (RMSE) were also
calculated for each combinations of MxAyWz and the results are summarized in Appendix-III.
The lowest AAE = 0.339 yielded by ANN model M6A0.15W8 and highest AAE = 15.077 was
163

yielded by ANN model M8A0.15W0. Similarly lowest RMSE = 0.1375 was yielded by ANN
model M6A0.15W8 and highest RMSE = 32.850 was yielded by ANN model M8A0.15W0.
Liquefaction potential predicted by ANFIS models for all 45 combinations is summarized in
Appendix-C. AAE & RMSE values from actual and predicted datasets for each combinations of
MxAyWz are shown in Appendix-D. Some of the ANFIS model showed considerably higher
value of AAE & RMSE due to which variation in the target parameters for models M 6A0.15W0,
M6A0.15W6 & M6A0.15W8 were showing very less number of liquefaction occurrence in target
parameters. Hence L.P.> 1.0 in most of the target parameters, which yielded less variation in
target values i.e. in between 1-1.50. Computation methods produce low error when data contains
high variation. Therefore ANN & RMSE was higher in such combinations of models. AAE &
RMSE for remaining ANFIS model are low in comparison to ANN models.
Models M8A0.35W0, M7A0.35W0 & M8A0.25W0 were developed. AAE & RMSE of the MLR model
M8A0.35W0 is >100%, for the model M7A0.35W0 it is 60.15% and 65.27% whereas for the model
M8A0.25W0, it is 19.49% and 22.92%. It can also be seen that AAE & RMSE was higher in MLR
models in comparison to soft computing models. Equations of the developed MLR models are
as:
For the model M8A0.35W0:
L.P.= -0.2326+1.1292N-0.2572m-0.1154 t +0.1001 D0.075+0.1185 wn -0.0091 D2.0-0.0933 D0.002 0.0765 .(13)

For the model M7A0.35W0:

L.P.= -0.1827+1.0471N-0.2549m+0.0492 t +0.0913 D0.075-0.0045 wn +0.0982 D2.0-0.2229 D0.002+0.0384 (14)

For the model M8A0.25W0:

L.P.= -0.1650+1.0209N-0.2940m+0.0736 t +0.0920 D0.075-0.0278 wn +0.1245D2.0-0.2543D0.002+0.0346 (15)

164

Fig. 5.85, Fig. 5.86 and Fig. 5.87 showed that COD by ANN models are 0.9884 0.9938 and
0.9989 which are nearer to 1 in comparison to other models. COD are greater than 0.9 by ANN,
ANFIS and MLR techniques for three combinations.

ANFIS

From ANN Methods From ANFIS Method From MLR Technique

y = 1.0002x +0.014
y = 0.9766x -0.0319
y = 1.0029x + 0.0062
R = 0.9894
R = 0.9041
R = 0.9783

1.6
L.P. by I&B Method

ANN

L.P. by I&B Method for M7A0.35W0 for Zone-II

1.8
1.4

MLR

1.2
1
0.8
0.6
0.4
0.2
0
0

0.2

0.4

0.6

0.8

1.2

1.4

1.6

L.P. by ANN, ANFIS and MLR Technique

Fig. 5.85: Comparative Study of L.P. between I&B Method and Various Computational Methods i.e. ANN, ANFIS
and MLR for Zone II for Model M7A0.35W0.

L.P. by I&B Method for M8A0.25W0 for Zone-II

From ANN Methods
y = 0.9439x -0.0224
R = 0.9938

L.P. by I&B Method

1.8
1.6

From ANFIS Method

y = 0.9224x -0.0389
R = 0.9638

From MLR Technique

y = 1.0857x - 0.0518
R = 0.922

1.4
1.2
1
0.8
0.6
ANN

0.4

ANFIS

0.2

MLR

0
0

0.2

0.4
0.6
0.8
1
1.2
L.P. by ANN, ANFIS and MLR Technique

1.4

1.6

Fig.5.86: Comparative Study of L.P. between I&B Method and Various Computational Methods i.e. ANN, ANFIS
and MLR for Zone II for Model M8A0.25W0.

165

L.P. by I&B Method for M8A0.35W0 for Zone-II

1.6
L.P. by I&B Method

1.4
1.2

From ANN Methods

y = 0.9974x +0.0025
R = 0.9989

From ANFIS Method

y = 0.9787x + 0.0117
R = 0.9669

From MLR Technique

y = 0.9979x - 0.0046
R = 0.9403

1
0.8
0.6

ANN
ANFIS
MLR

0.4
0.2
0
-0.2 0
-0.4

0.2

0.4

0.6

0.8

1.2

1.4

1.6

L.P. by ANN, ANFIS and MLR Technique

Fig.5.87: Comparative Study of L.P. between I&B Method and Various Computational Methods i.e. ANN, ANFIS
and MLR for Zone II for Model M8A0.35W0.

5.3.2.3 Results of Developed Model based on T&Y method

The third semi empirical method from which liquefaction potential was evaluated is T&Y
method. Using the same combination of MXAYWZ the liquefaction occurrences from 137
datasets were calculated and illustrated in Appendix-B. It can be seen from the Fig. 5.15 that
most combination are susceptible to liquefaction. However, M8A0.35W0, M7A0.35W0 &
M6A0.35W0 combination were choosing for comparative study (Table 5.1) whose chances of
liquefaction potential are 84.545%, 83.636% and again 83.636%.
Liquefaction potential predicted by ANN models are mentioned in Table C.12 for above three
combinations. It can be seen from this table that chances of liquefaction are very high in all 17
data for all three combinations. It can also be seen that liquefaction potential predicted by ANN
models are in accordance with T&Y method for all 17 data.
Liquefaction potential predicted by ANFIS models are in accordance with T&Y method (Table
C-15). AAE and RMSE are also shown in Table C-15. This indicates that ANN & ANFIS

166

methods can be adopted for the development of L.P. models. The AAE and RMSE of ANN &
ANFIS models are shown in Annexure-D.
Models M8A0.35W0, M7A0.35W0 & M6A0.35W0 were developed for T&Y method. Equations of the
developed MLR models are as:
For the model M8A0.35W0:
L.P.=-0.3709+0.9038N-0.6665m-0.5021 t +0.3777 D0.075+0.0627 wn +0.3996 D2.0-0.7131 D0.002+0.0114 .....(16)

For the model M7A0.35W0:

L.P.=-0.2491+0.8376N-0.6615m-0.3627 t +0.3910 D0.075+0.0140 wn +0.3347 D2.0-0.7108 D0.002-0.0283 ..(17)

For the model M6A0.35W0:

L.P.=-0.1305+0.7446N-0.6413m-0.0919 t +0.4124 D0.075-0.0476 wn +0.2573 D2.0-0.6889 D0.002-0.0596 ..(18)

ANN
ANFIS
MLR

L.P. by T&Y Method for M8A0.35W0 for Zone-II

L.P. by T&Y Method

From ANN Methods

y = 0.9955x+0.0052
R = 0.9974

1.5

From ANFIS Method

y = 0.9958x+0.0032
R = 0.9946

From MLR Technique

y = 0.9954x - 0.0759
R = 0.8892

1
0.5
0
0
-0.5

0.2

0.4

0.6

0.8

1.2

1.4

1.6

L.P. by ANN, ANFIS and MLR Technique

Fig.5.88: Comparative Study of L.P. between T&Y Method by ANN, ANFIS and MLR for Zone II for Model
M8A0.35W0.

167

ANN
ANFIS
MLR

L.P. by T&Y Method for M7A0.35W0 for Zone-II

L.P. by T&Y Method

1.8
1.6
1.4

From ANN Methods

y = 1.0178x+0.0095
R = 0.9604

From ANFIS Method

y = 1.0022x+0.0045
R = 0.9946

From MLR Technique

y = 1.021x - 0.0653
R = 0.9010

1.2
1
0.8
0.6
0.4
0.2
0
-0.2 0

0.2

0.4
0.6
0.8
1
L.P. by ANN, ANFIS and MLR Technique

1.2

1.4

1.6

Fig.5.89: Comparative Study of L.P. between T&Y Method by ANN, ANFIS and MLR for Zone II for Model
M7A0.35W0.

L.P. by T&Y Method for M6A0.35W0 for Zone-II

1.8

L.P. by T&Y Method

1.6
1.4

From ANN Methods

y = 1.0041x+0.0035
R = 0.9979

From ANFIS Method

y = 1.0124x+0.0068
R = 0.9901

From MLR Technique

y = 1.0429x - 0.0427
R = 0.9259

1.2
1
0.8
0.6

ANN

0.4

ANFIS

0.2

MLR

0
-0.2 0

0.2

0.4

0.6
0.8
1
1.2
L.P. by ANN, ANFIS and MLR Technique

1.4

1.6

Fig.5.90: Comparative Study of L.P. between T&Y Method by ANN, ANFIS and MLR for Zone II for Model
M6A0.35W0.

It can be referred from these results that ANN and ANFIS models showed better performance in
terms of statistical parameters compared with MLR.

168

5.3.3 L.P. by ANN, ANFIS & MLR Models of zone-III

Zone-III is having 167 datasets of 42 boreholes. Among the 42 borehole data, 34 borehole data
were reserved to train and development ANN, ANFIS and MLR models. Eight borehole data
(i.e. 32 datasets) were reserved for validating the different developed models by ANN, ANFIS
and MLR techniques. These eight reserved boreholes are BH no. 2, 4, 7, 9, 18, 25, 27 and 39.
5.3.3.1. Results of Developed models Based on Modified Seeds Method
Liquefaction potential for zone III calculated from Modified Seeds method for all 45
combinations are summarized in Annexure B. It could be seen from the Fig. 5.25 average
chances of liquefaction of soil is 36.141% and the maximum liquefaction occurs for a
combination M8A0.25W2. Models M6A0.15W0, M6A0.15W6 and M6A0.15W8 do not show any
chances of liquefaction (Fig. 5.25). Three combinations i.e. M8A0.35W2, M8A0.35W4 and
M8A0.35W6 were chosen for discussion (Table 5.1).
It can be observed from Table C-19 that prediction of L.P. by ANN models are in accordance
with actual liquefaction potential value. 100% chances of liquefaction do not occur for
combinations M8A0.35W2, M8A0.35W4 and M8A0.35W6.
ANN models for these three combinations are developed through all 91 network architecture.
N48, N57 and N22 gave minimum error and maximum COD value for models M8A0.35W2,
M8A0.35W4 and M8A0.35W6 respectively. AAE and RMSE value for model M8A0.35W2,
M8A0.35W4 and M8A0.35W6 are 21.250%, 15.020%, 13.150% and 28.230%, 24.660%, 22.970%
respectively. ANN models showed good prediction accuracy of L.P. when compared with actual
values of L.P.
Results obtained by ANFIS models for these combinations are shown in Table C-22 The
minimum AAE of ANFIS model M6A0.15W8 is 2.470 and RMSE is 1.345 for the ANFIS model

169

M6A0.15W0. The maximum value of AAE and RMSE computed is 26.347% and 54.102%
respectively, for the ANFIS model M7A0.35W4. Two incorrect predictions were observed at serial
no. 30 in ANFIS models M8A0.35W2 and M8A0.35W6 (Table C-22). AAE and RMSE value for
model M8A0.35W2 is 12.870% and 19.510%, for model M8A0.35W4 is 11.150% and 22.290% and
for model M8A0.35W6 is 10.500% and 18.700%.
Multiple linear regressions models M8A0.35W2, M8A0.35W4 & M7A0.35W6 were also developed for
evaluating L.P. of soil. The equations developed by these combinations are as follows:
For model M8A0.35W2:
L.P.= -0.1153+0.8254N-0.3075m+0.0884 t +0.1187 D0.075-0.0554 wn +0.0707 D2.0-0.0646 D0.002+ 0.0753 (19)

For M8A0.35W4:
L.P.= 0.0560+0.9493N-0.3590m+0.0763 t +0.0781 D0.075-0.0228 wn -0.0215 D2.0-0.1014 D0.002+

0.0227 (20)

For M8A0.35W6:
L.P.= 0.0722+0.8691N-0.2681m-0.0581 t +0.1347 D0.075-0.0422 wn -0.0490 D2.0-0.0672 D0.002+0.1274 ..(21)

From MLR techniques, results were calculated by using input datasets which were reserved for
validation for zone-III which are summarized in Table C-25. The prediction performances of
MLR models are also acceptable i.e. only two predictions are incorrect which are s. no. 30 for
model M8A0.35W2 and M8A0.35W4. AAE & RMSE were also calculated which are 61.66% and
>100% for model M8A0.35W2, AAE & RMSE both are greater than 100% for model M8A0.35W4
and 67.21% and >100% for the model M8A0.35W6 but the COD are 0.9452, 0.857 and 0.967
respectively.
ANN, ANFIS and MLR models were compared for most susceptible combination of M8A0.35W2
(Fig. 5.91), M8A0.35W4 (Fig. 5.92) & M7A0.35W6 (Fig. 5.93). The coefficient of determination
value i.e. R2 of MLR model M8A0.35W2 is good compared to other models. COD of ANN
models M8A0.35W4 and M8A0.35W6 are 0.9689 and 0.9673 respectively.

170

L.P. by Modified Seed's Method for M8A0.35W2 for Zone-III

L.P. by Modified Seed's Method

1.8

From ANFIS Method

y = 1.1006x -0.0416
R = 0.9573

From ANN Method

y = 0.9171x + 0.0415
R = 0.8812

1.6
1.4

From MLR Technique

y = 0.878x -0.0057
R = 0.9452

ANN
ANFIS
MLR

1.2
1
0.8
0.6
0.4
0.2
0
-0.2 0

0.2

0.4
0.6
0.8
1
L.P. by ANN, ANFIS and MLR Technique

1.2

1.4

1.6

L.P. by Modified Seed's Method

Fig.5.91: Comparative Study of L.P. between Modified Seeds Method and Various Computational Methods i.e.
ANN, ANFIS and MLR for Zone III for Model M8A0.35W2.

L.P. by Modified Seed's Method for M8A0.35W4 for Zone-III

From ANN Method

y = 0.9838x - 0.0107
R = 0.9689

1.5

Fronm ANFIS Method

y = 0.9315x - 0.0536
R = 0.8507

From MLR Technique

y = 1.0817x -0.0673
R = 0.8569

ANN
ANFIS
MLR

0.5

0
0

0.2

0.4

0.6
0.8
1
1.2
L.P. by ANN, ANFIS and MLR Technique

1.4

1.6

-0.5
Fig. 5.92: Comparative Study of L.P. between Modified Seeds Method and Various Computational Methods i.e.
ANN, ANFIS and MLR for Zone III for Model M8A0.35W4.

171

L.P. by Modified Seed's Method for M8A0.35W6 for Zone-III

L.P. by Modified Seed's Method

1.6

From ANN Method

y = 1.0054x - 0.0042
R = 0.9874

1.4
1.2

Fronm ANFIS Method

y = 0.949x +0.0075
R = 0.9667

ANN
ANFIS
MLR

1
0.8
0.6
From MLR Technique
y = 0.8914x +0.0542
R = 0.9673

0.4
0.2
0
-0.2

0.2

0.4

0.6

0.8

1.2

1.4

1.6

L.P. by ANN, ANFIS and MLR Technique

Fig. 5.93: Comparative Study of L.P. between Modified Seeds Method and Various Computational Methods i.e.
ANN, ANFIS and MLR for Zone III for Model M8A0.35W6.

5.3.3.2 Results of Developed Models based on I&B Method

In zone-III an average chance of liquefaction potential by I&B approach is 43.247% (Fig. 5.26).
Combinations of M8A0.35W0, M8A0.35W2 & M8A0.35W4 calculated maximum chances of
liquefaction where as M6A0.15W4, M6A0.15W6 & M6A0.15W8 combination showed no chances of
liquefaction as per Figure 5.26.
All results predicted by ANN models were in accordance with the results evaluated by I&B
method. AAE and RMSE value for ANN model M8A0.35W0 is 30.036% and 77.853%, for model
M8A0.35W2 is 17.673% and 38.743% and for model M8A0.35W4 is 12.174% and 20.766% (Table
C-20). M8A0.35W0 gave minimum RMSE and maximum COD under the network architecture N4
where as models M8A0.35W2 and M8A0.35W0 gave minimum RMSE and maximum COD under
the network N28 and N4 respectively. Here N4 is single hidden layer network but N28 is double
hidden layer network.
Table C-23 shows incorrect predictions at serial no. 30 for model M8A0.35W0, M8A0.35W2 &
M8A0.35W4. Thus out of 96, 3 predictions by ANFIS models are not matching. AAE and RMSE

172

for ANFIS model M8A0.35W0 is 11.320% and 17.390%, for ANFIS model M8A0.35W2 is 12.560%
and 17.930% and for ANFIS model M8A0.35W4 is 14.693% and 22.707%.
The MLR models M8A0.35W0, M8A0.35W2 & M8A0.35W4 were developed. The multi linear
equations of MLR models are:
For model M8A0.35W0:
L.P.= -0.1039+1.0884N-0.3708m+0.1042 t +0.0694 D0.075-0.0222 wn + 0.0340 D2.0+0.0647 D0.002-0.0085.(22)

For model M8A0.35W2:

L.P.= 0.0198+1.0150N-0.3549m-0.0169 t +0.0528D0.075-0.0259 wn +0.0415 D2.0+0.0290 D0.002-0.0036.(23)

For model M8A0.35W4:

L.P.= 0.0884+0.9790N-0.3094m-0.0920 t +0.0377 D0.075-0.0198 wn +0.0358 D2.0-0.0088 D0.002-0.0127.(24)

Results of MLR models were found by using input datasets which were reserved for validation
for zone-III which are summarized in Table C-26. There are five predictions by MLR models
which are incorrect and can be seen at serial no. 3, 4, 7, 8 and 26 for model M8A0.35W4. AAE &
RMSE were also calculated which are >100% for all three models, but the COD are 0.876, 0.945
and 0.971 respectively for M8A0.35W0, M8A0.35W2 & M8A0.35W4.

L.P. by I&B Method for M8A0.35W0 for Zone-III

ANN
ANFIS
MLR

L.P. by I&B Method

1.8
From ANN Methods From ANFIS Method From MLR Technique
1.6
y = 0.9487 - 0.0035
y = 1.0085x -0.0084
y = 0.9912 x + 0.0775
1.4
R = 0.9722
R = 0.9919
R = 0.8763
1.2
1
0.8
0.6
0.4
0.2
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
-0.2 0
L.P.
by
ANN,
ANFIS
and
MLR
Technique
-0.4
Fig. 5.94: Comparative Study of L.P. between I&B Method and Various Computational Methods i.e. ANN, ANFIS
and MLR for Zone III for Model M8A0.35W0.

173

ANN
ANFIS
MLR

L.P. by I&B Method for M8A0.35W2 for Zone-III

L.P. by I&B Method

1.8
From ANFIS Method From MLR Technique
From ANN Methods
y = 0.9513 - 0.0074
1.6
y = 1.0239x + 0.0551
y = 1.038x -0.0057
R = 0.9802
R = 0.9450
1.4
R = 0.9771
1.2
1
0.8
0.6
0.4
0.2
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
-0.2 0
L.P. by ANN, ANFIS and MLR Technique
Fig. 5.95: Comparative Study of L.P. between I&B Method and Various Computational Methods i.e. ANN, ANFIS
and MLR for Zone III for Model M8A0.35W2.

1.8
L.P. by I&B Method

1.6
1.4

ANN

L.P. by I&B Method for M8A0.35W4 for Zone-III

From ANN Methods
y = 0.9811x +0.0043
R = 0.9881

ANFIS

MLR

From ANFIS Method

y = 0.9386x + 0.0228
R = 0.9772

1.2
1
0.8
0.6
From MLR Technique
y = 1.0185x + 0.027
R = 0.9713

0.4
0.2
0
-0.2 0

0.2

0.4
0.6
0.8
1
L.P. by ANN, ANFIS and MLR Technique

1.2

1.4

1.6

Fig. 5.96: Comparative Study of L.P. between I&B Method and Various Computational Methods i.e. ANN, ANFIS
and MLR for Zone III for Model M8A0.35W4.

The results of ANN, ANFIS and MLR models were compared as shown in Fig. 5.94, 5.95 and
5.96. It has been observed that COD of ANN models are better than others models except for
model M8A0.35W2. In this case COD of ANFIS model is better but overall performances of ANN
models are more accurate than other models.
5.3.3.3 Results of Developed Model based on T&Y Method
An average chance of soil liquefaction in this zone result by the T&Y method is 48.902% (Fig.
5.27). However, combinations most susceptible to liquefaction are M8A0.35W0, M7A0.35W0 &
174

Liquefaction potential predicted by ANN models are shown in Table C-21. Only one prediction
by ANN models is incorrect out of 96 predictions i.e. at serial no. 8 for model M8A0.35W0. AAE
and RMSE value are also summarized in Table C-21.
Table C-24 showed the results obtained from ANFIS models. The AAE and RMSE for models
M8A0.35W0, M7A0.35W0 & M8A0.25W0 are 11.110%, 9.090%, 8.350%, 21.900%, 15.430% and
14.280% respectively.
MLR models of M8A0.35W0, M7A0.35W0 & M8A0.25W0 combinations were developed. The
equations are follows:
For model M8A0.35W0:
L.P.= -0.2769+0.4274N-0.2024m+0.1671 t +0.3170 D0.075-0.0948 wn +0.1057 D2.0+0.0511 D0.002+ 0.0046.(25)

For model M7A0.35W0:

L.P. = -0.2651+0.4681N-0.2190m+ 0.1926t+0.3555D0.075-0.1089 wn +0.1112 D2.0+0.0484 D0.002+ 0.0067.(26)

For model M8A0.25W0:

L.P.= -0.2126+0.4948N-0.2309m+0.2381 t +0.3893 D0.075-0.1240 wn + 0.1011D2.0+0.0229 D0.002+ 0.0030.(27)

Out of 96 predictions by MLR models only 4 predictions are not correct compared to T&Y
method (Table C-27). The AAE for MLR models M8A0.35W0, M7A0.35W0 & M8A0.25W0 are
40.970%, 29.800% and 20.870% where as RMSE for model M8A0.35W0, M7A0.35W0 &
M8A0.25W0 are 50.090%, 35.960% and 25.090% respectively.

175

ANN
ANFIS
MLR

L.P. by T&Y Method for M8A0.35W0 for Zone-III

1.6

From ANN Methods

y = 0.9727x+0.0065
R = 0.9941

1.4

From ANFIS Method

y = 0.9244x+0.098
R = 0.9726

From MLR Technique

y = 0.8265x - 0.04
R = 0.6626

L.P. by T&Y Method

1.2
1
0.8
0.6
0.4
0.2
0
-0.2

0.2

0.4
0.6
0.8
1
L.P. by ANN, ANFIS and MLR Technique

1.2

1.4

1.6

Fig. 5.97: Comparative Study of L.P. between T&Y Method and Various Computational Methods i.e. ANN, ANFIS
and MLR for Zone I for Model M8A0.35W0.

L.P. by T&Y Method for M7A0.35W0 for Zone-III

From ANN Methods
y = 1.0683x - 0.0094
R = 0.9062

1.6

L.P. by T&Y Method

1.4

From ANFIS Method

y = 0.8684x + 0.0353
R = 0.9436

From MLR Technique

y = 0.7802x + 0.0655
R = 0.6660

1.2
1
0.8
0.6

ANN

0.4

ANFIS

0.2

MLR

0
0

0.2

0.4
0.6
0.8
1
L.P. by ANN, ANFIS and MLR Technique

1.2

1.4

1.6

Fig. 5.98: Comparative Study of L.P. between T&Y Method and Various Computational Methods i.e. ANN, ANFIS
and MLR for Zone I for Model M7A0.35W0.

176

1.8

From ANN Methods From ANFIS Method

y = 1.1717x+0.0561 y = 1.0411x+0.0167
R = 0.8578
R = 0.9519

1.6
1.4
L.P. by T&Y Method

ANN
ANFIS
MLR

L.P. by T&Y Method for M8A0.25W0 for Zone-III

From MLR Technique
y = 0.7922x - 0.0833
R = 0.7063

1.2
1
0.8
0.6
0.4
0.2
0
0

0.2

0.4
0.6
0.8
1
L.P. by ANN, ANFIS and MLR Technique

1.2

1.4

1.6

Fig.5.99: Comparative Study of L.P. between T&Y Method and Various Computational Methods i.e. ANN, ANFIS
and MLR for Zone I for Model M6A0.35W0.

Fig. 5.97, 5.98 and 5.99 showed comparative of ANN, ANFIS and MLR models. COD value of
ANN models is better than ANFIS and MLR models. MLR showed poor COD value for these
three developed models.
5.3.4 L.P. by ANN, ANFIS and MLR Models for Zone IV
Zone-IV contains 32 borehole or 104 datasets for which Liquefaction Potential has been
evaluated. Out of 32 borehole data (i.e. 137 datasets) six borehole data were reserved for
validating the different developed ANN, ANFIS and MLR models. The data of boreholes BH- 3,
13, 18, 23, 27 and 32 were used for validation of models

5.3.4.1. Results of Developed Models Based on Modified Seeds Method

Liquefaction potential for zone IV calculated by Modified Seeds method for all 45 combinations
are summarized in Annexure B. The average chances of liquefaction by Modified Seeds method
are 28.034%. Maximum chances of liquefaction occurs for three combination are M8A0.35W2,
M8A0.35W4 and M8A0.35W6 Out of 19 datasets two datasets of combination M8A0.35W2 and three
177

datasets of combination l M8A0.35W4 and M8A0.35W6 shows no liquefaction condition.(Table

5.37).
ANN model were developed for all network architecture. Out of

91 developed network

architecture N76, N52 and N39 predictions for the ANN models M8A0.35W2,
M8A0.35W6

M8A0.35W4 and

are better than other .AAE for model M8A0.35W2, M8A0.35W4 and M8A0.35W6 are

15.836%, 16.747% and 13.555% whereas RMSE is 20.530%, 23.543% and 16.761%
respectively.
Liquefaction potential of soils calculated for all 45 ANFIS models are shown in annexure-B. The
minimum AAE given by ANFIS model is 2.636 for the model M6A0.15W8 and for RMSE is
6.320. The maximum value of AAE and RMSE computed by ANFIS is 48.221% and 120.062%
respectively, for the model M7A0.25W0. Only one incorrect prediction is found by ANFIS models
i.e. in serial no. 9 for model M8A0.35W2.
Multiple linear regressions models developed for these combinations are as follows.
For model M8A0.35W2:
L.P.= 0.0933+1.0074N-0.6003m+0.1398 t +0.0062 D0.075+0.0379 wn +0.1476 D2.0-0.0111 D0.002+ 0.0207 (28)

For model M8A0.35W4

L.P.= =0.0006+1.0218N-0.4790m-0.0821 t +0.0264 D0.075+0.0344 wn +0.1461 D2.0-0.1378 D0.002-0.0239..(29)

For model M8A0.35W6:

L.P =0.1245+1.0404N-0.3626m-0.2585 t +0.0117 D0.075+0.0556 wn +0.0650 D2.0-0.1903 D0.002-0.0595.(30)

178

L.P. by Modified Seed's Method

L.P. by Modified Seed's Method

ANN
L.P. by Modified Seed's Method for M8A0.35W2 for Zone-IV
1.8
ANFIS
From ANN Method Fronm ANFIS Method
1.6
y = 0.92x - 0.0032
y = 0.8783x +0.2025
MLR
1.4
R = 0.9338
R = 0.7611
1.2
1
0.8
0.6
From MLR Technique
0.4
y = 0.8789x +0.0493
0.2
R = 0.8121
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
-0.2 0
L.P.
by
ANN,
ANFIS
and
MLR
Technique
-0.4
Fig.5.100: Comparative Study of L.P. between Modified Seeds Method and Various Computational Methods i.e.
ANN, ANFIS and MLR for Zone IV for Model M8A0.35W2.
ANN
L.P. by Modified Seed's Method for M8A0.35W4 for Zone-IV
1.6
From ANN Method Fronm ANFIS Method From MLR Technique
ANFIS
1.4 y = 0.949x - 0.0307
y = 0.8689x +0.0486
y = 0.8773x +0.145
R = 0.9567
R = 0.9718
R = 0.8206
1.2
1
0.8
0.6
0.4
0.2
0
0.4
0.6
0.8
1
1.2
1.4
1.6
L.P. by ANN, ANFIS and MLR Technique
Fig. 5.101: Comparative Study of L.P. between Modified Seeds Method and Various Computational Methods i.e.
ANN, ANFIS and MLR for Zone IV for Model M8A0.35W4.

L.P. by Modified Seed's Method

-0.2

0.2

ANN

L.P. by Modified Seed's Method for M8A0.35W6 for Zone-IV

1.6

From ANN Method

y = 0.8524x +0.0566
R = 0.9584

1.4
1.2

Fronm ANFIS Method

y = 0.7937x +0.1638
R = 0.885

From MLR Technique

y = 0.8918x +0.0033
R = 0.9689

ANFIS

1
0.8
0.6
0.4
0.2
0
0

0.2

0.4
0.6
0.8
1
L.P. by ANN, ANFIS and MLR Technique

179

1.2

1.4

1.6

Fig. 5.102: Comparative Study of L.P. between Modified Seeds Method and Various Computational Methods i.e.
ANN, ANFIS and MLR for Zone IV for Model M8A0.35W6.
Table 5.3: Statistical Parameters Calculated between Modified Seeds method and Developed Models.
Models

M8A0.35W2

M8A0.35W4

ANN

ANFIS

MLR

ANN

ANFIS

AAE (%)

16..216

39.421

57.872

16.533

RMSE (%)

24.468

91.996

>100

COD

0.9338

0.7611

0.8121

M8A0.35W6
MLR

ANN

ANFIS

MLR

41.385

19.968

13.797

42.248

39.545

23.154

48.003

30.252

17.784

89.019

80.987

0.9718

0.8206

0.9567

0.9584

0.8850

0.9689

In addition to individual ANN, ANFIS and MLR models the comparative performance of most
susceptible combination of Liquefaction M8A0.35W2 (Fig. 5.100), M8A0.35W4

(Fig. 5.101) &

M8A0.35W6 (Fig. 5.102) was also studied. Coefficient of correlation value i.e. R2 is higher for two
models i.e. for M8A0.35W2 and M8A0.35W4 by ANN technique but in Fig. 5.102 showed that the
MLR technique having higher R2 value for model M8A0.35W6 (Table 5.3).
5.3.4.2 Results of Developed Models based on I&B Techniques
Average chances of liquefaction potential evaluated by I&B method for zone IV is 35.021%
(Figure 5.38). The following combinations M8A0.35W0, M8A0.35W2 & M8A0.35W4 were selected
for comparative study.
ANN & ANFIS models were developed to predict liquefaction potential. The results of above
models has been compared and discussed in following literature. Prediction through ANN is
accordance with the results obtained through I&B formula. AAE for the model M8A0.35W0,
M8A0.35W2 and M8A0.35W4 are 14.830%, 14.169% and 14.092% and RMSE are 20.205%,
20.108% and 17.596% whereas network architecture are N76, N64 and N34 respectively (Table
5.29). Whereas, ANN techniques given the lowest AAE = 0.932% and RMSE =3.804% value
yielded for model M6A0.15W6 having network architecture N7 and highest AAE = 23.406% and
RMSE =39.354% was yielded for model M8A0.35W2 having network architecture N28.
180

Table 5.4: Statistical Parameters Calculated between I&B method and Developed Models.
M8A0.35W0

M8A0.35W2

M8A0.35W4

Models
ANN

ANFIS

MLR

ANN

ANFIS

MLR

ANN

AAE (%)

14.125

16.333

39.633

14.133

RMSE (%)

19.411

23.846

71.167

COD

0.9841

0.9467

0.9104

ANFIS

MLR

11.015

32.933

14.946

13.060

>100

17.649

13.757

36.537

20.030

16.578

>100

0.9874

0.9897

0.9321

0.9831

0.9086

0.8697

ANFIS techniques given the lowest AAE = 2.228% yielded for model M6A0.15W8 and highest
AAE = 28.074% was yielded for model M7A0.25W0. Similarly lowest RMSE = 4.276335% was
yielded for model M6A0.15W4 and highest RMSE = 60.64037% was yielded for model
M7A0.25W0. Table 5.32 showed that the value of AAE for the models M8A0.35W0, M8A0.35W2 and
M8A0.35W4 are 17.260%, 13.090% and 13.220% respectively whereas the values of RMSE for
the same models are 24.020%, 15.280% and 16.440% respectively.
The MLR models M8A0.35W0, M8A0.35W2 & M7A0.35W4 were developed and liquefaction for
calculating the L.P. by these three models three multi linear equations had been generated which
are:
For model M8A0.35W0:
L.P.= -0.0320+1.1947N-0.2381m+0.0479 t -0.0190 D0.075+0.0917 wn -0.0015 D2.0-0.3511 D0.002-0.0734...(31)

For model M8A0.35W2:

L.P.= 0.0274+1.1771N-0.2601m+0.0220 t -0.0202 D0.075+0.0853 wn -0.0195 D2.0-0.3513 D0.002-0.0858..(32)

For model M8A0.35W4:

L.P. =-0.3473+1.2829N-0.3791m+ 0.2543 t -0.0375 D0.075+0.1712 wn +0.1590 D2.0-0.3503 D0.002-0.0328.(33)

From the above three equation we obtain the L.P. for 17 reserved datasets. Out of these 51
results only one prediction is adverse which serial no. 9 for model M8A0.35W2. AAE & RMSE
were also calculated which are shown in Table C-35. It can also be seen that AAE & RMSE was
higher in MLR models in comparison to computational models.
181

From Fig. 5.103, 5.104 and 5.105, it can be clearly seen that COD determined by results obtained
through ANN model having 0.9841, 9874 and 0.9831 for the model M8A0.35W0, M8A0.35W2 &
M7A0.35W4 respectively. But COD value from result obtained by ANFIS model is 0.9897 for
model M8A0.35W2 (Table 5.4).

1.7

L.P. by I&B Method

1.5
1.3

ANN

L.P. by I&B Method for M8A0.35W0 for Zone-IV

From ANN Methods
y = 0.9502x +0.007
R = 0.9841

From ANFIS Method

y = 0.9378x + 0.0495
R = 0.9467

ANFIS

From MLR Technique

y = 0.9145x - 0.0367
R = 0.9104

MLR

1.1
0.9
0.7
0.5
0.3
0.1

L.P. by I&B Method

-0.1 0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
L.P. by ANN, ANFIS and MLR Technique
-0.3
Fig. 5.103: Comparative Study of L.P. between I&B Method and Various Computational Methods i.e. ANN, ANFIS
and MLR for Zone IV for Model M8A0.35W0.
ANN
L.P. by I&B Method for M8A0.35W2 for Zone-IV
ANFIS
1.6
MLR
From ANN Methods From ANFIS Method From MLR Technique
1.4
y = 0.9149x - 0.0254
y = 0.9296x + 0.0023
y = 0.9929x +0.022
R = 0.9321
R = 0.9897
R = 0.9874
1.2
1
0.8
0.6
0.4
0.2
0
-0.2 0
-0.4

0.2

0.4
0.6
0.8
1
L.P. by ANN, ANFIS and MLR Technique

1.2

1.4

1.6

Fig. 5.104: Comparative Study of L.P. between I&B Method and Various Computational Methods i.e. ANN, ANFIS
and MLR for Zone IV for Model M8A0.35W2.

182

L.P. by I&B Method

L.P. by I&B Method for M8A0.35W4 for Zone-IV

1.6
From
ANN
Methods From ANFIS Method
1.4
y = 0.9903x - 0.0094
y = 0.9311x + 0.0885
1.2
R = 0.9831
R = 0.9086
1
0.8
0.6
From MLR Technique
0.4
y = 0.9125x - 0.0498
0.2
R = 0.8697
0
0.2
0.4
0.6
0.8
1
1.2
1.4
-0.2 0
L.P. by ANN, ANFIS and MLR Technique
-0.4

ANN
ANFIS
1.6

Fig. 5.105: Comparative Study of L.P. between I&B Method and Various Computational Methods i.e. ANN, ANFIS
and MLR for Zone IV for Model M8A0.35W4.

5.3.4.3 Results of Developed Models based on T&Y Method

An average chance of liquefaction evaluated by T&Y Method is 35.513%. The combinations
having maximum chances of liquefaction are M8A0.35W0, M7A0.35W0 & M8A0.25W0. In all the
combinations more or less chances of liquefaction potential happen.
Liquefaction potential predicted by ANN model is shown in Table C-30 are M8A0.35W0,
M7A0.35W0 & M8A0.25W0. The chances of liquefaction are very high for all 19 data sets reserved
for validation. It can be also seen from Table C-30 that liquefaction predicted by ANN model is
not matching with T&Y method for one combinations at serial no. 7 for M7A0.35W0.
The predicted liquefaction potential from ANFIS model are in accordance with T&Y method
except for one combination i.e. at serial no. 7 again for ANFIS model M7A0.35W0 (Table C-33).
The following models were developed using MLR Technique:
MLR model M8A0.35W0:
L.P.= -0.2500+0.5428N-0.5558m+0.2190 t +0.2988 D0.075+0.1208 wn +0.2377 D2.0-0.7065 D0.002-0.0714(34)

MLR model M7A0.35W0:

L.P.= 0.2666+0.5897N-0.5924m+0.2591 t +0.3482 D0.075+0.0905 wn +0.2703 D2.0-0.6694 D0.002-0.0700..(35)

MLR model M8A0.25W0:

183

L.P. = -0.2814+0.6501N-0.6225m+0.3315 t +0.4139 D0.075+0.0295 wn +0.2926 D2.0-0.5841 D0.002-0.0463....(36)

From the above three equation L.P. for 19 reserved datasets were obtained. Out of 57 results
three predictions were adverse which are at serial no. 7, 8 and 9 for MLR model M8A0.35W0.
Table 5.5: Statistical Parameters Calculated between T&Y method and Developed Models.
M8A0.35W0

M7A0.35W0

M8A0.25W0

Models
ANN

ANFIS

MLR

ANN

ANFIS

MLR

ANN

ANFIS

MLR

AAE (%)

10.175

14.390

>100

10.624

13.170

30.410

10681

16.161

33.983

RMSE (%)

12.677

24.309

>100

13.716

17.523

47.141

14693

23.295

48.957

COD

0.9984

0.5272

0.8338

0.9614

0.8886

0.6634

0.9576

0.8760

0.6632

Fig. 5.106, 5.107 and 5.108 showed comparative results of ANN ,ANFIS, MLR analysis and
from these figures it could be seen that COD value is better in ANN than ANFIS and then MLR.
MLR showed poor COD value for these three developed models.
L.P. by T&Y Method for M8A0.35W0 for Zone-IV

L.P. by T&Y Method

1.6

From ANN Methods

y = 1.0138x-0.0055
R = 0.9984

1.4
1.2

From ANFIS Method

y = 0.2922x+0.1585
R = 0.5272

1
0.8
0.6
From MLR Technique
y = 0.8415x + 0.2528
R = 0.8338

0.4
0.2
0
0

0.2

0.4

0.6

0.8

1.2

1.4

1.6

L.P. by ANN, ANFIS and MLR Technique

Fig.5.106: Comparative Study of L.P. between T&Y Method and Various Computational Methods i.e. ANN,
ANFIS and MLR for Zone IV for Model M8A0.35W0.

184

L.P. by T&Y Method

1.6

ANN
ANFIS
MLR

L.P. by T&Y Method for M7A0.35W0 for Zone-IV

From ANN Methods
y = 1.0587x+0.0094
R = 0.9614

1.4
1.2
1

From ANFIS Method

y = 1.0639x+0.0096
R = 0.8886

0.8
0.6

From MLR Technique

y = 1.1898x - 0.0859
R = 0.6634

0.4
0.2
0
-0.2 0

0.2

-0.4

0.4
0.6
0.8
1
L.P. by ANN, ANFIS and MLR Technique

1.2

1.4

1.6

Fig. 5.107: Comparative Study of L.P. between T&Y Method and Various Computational Methods i.e. ANN,
ANFIS and MLR for Zone IV for Model M7A0.35W0.

L.P. by T&Y Method

1.6

ANN
ANFIS
MLR

L.P. by T&Y Method for M8A0.25W0 for Zone-IV

From ANN Methods
y = 1.0041x+0.0188
R = 0.9576

1.4
1.2
1
0.8
0.6
0.4

From ANFIS Method

y = 1.0143x+0.0012
R = 0.876

0.2

From MLR Technique

y = 1.0671x - 0.0455
R = 0.6632

0
-0.2 0

0.2

0.4
0.6
0.8
1
L.P. by ANN, ANFIS and MLR Technique

1.2

1.4

1.6

Fig. 5.108: Comparative Study of L.P. between T&Y Method and Various Computational Methods i.e. ANN,
ANFIS and MLR for Zone IV for Model M8A0.35W0.

5.3.5 L.P. by ANN, ANFIS and MLR Models for Zone V

Zone-V is having 126 datasets (36 boreholes). Out of 126 datasets, 20 datasets of four boreholes
have been reserved to validate ANN, ANFIS and MLR models. The reserved boreholes are BH4, 14, 27 and 36.
5.3.5.1. Results of developed model based on Modified Seeds Method
Liquefaction potential for zone V evaluated by Modified Seeds method for 45 combinations are
summarized in Annexure B. The average chances of liquefaction potential are 41.975%
(Fig 5.49). The following combinations M8A0.35W2, M8A0.35W4 and M7A0.35W2 were showing
185

maximum chances of liquefaction in percentages which are 75.397%, 75.397% and 74.603%
respectively. AAE and RMSE were calculated between liquefaction potential obtained from
computational methods and modified Seeds method for all 45 combinations is shown in
annexure C.
The developed ANN models for above, combinations predicted similar results when compared
to actual values (Table C-37). These ANN models showed good prediction accuracy. The
network architecture N22, N22 and N34 whose prediction capabilities are good respectively for the
combinations M8A0.35W2, M8A0.35W4 and M7A0.35W2.
The minimum AAE & RMSE were 1.193% & 5.854% for

ANN model M6A0.15W4 (N39) and

maximum AAE & RMSE were for model M8A0.35W0 (N66).

Similar to ANN model prediction accuracy of ANFIS models were very good which is shown in
Table C-40. Minimum AAE & RMSE were 3.686% & 5.382% for ANFIS models M6A0.15W0
and M6A0.15W8 whereas maximum AAE & RMSE for ANFIS model M6A0.35W8 i.e. 75.015%
and 93.754%.
The developed models by MLR techniques are M8A0.35W2, M8A0.35W4 and M7A0.35W2 . Equation
of these models are as follows
For M8A0.35W2:
L.P.= 0.0676+0.8872N-0.3100m-0.0807 t -0.0366 D0.075+0.0294 wn +0.0246 D2.0-0.0043 D0.002-0.0410(37)

For M8A0.35W4:
L.P.= -0.0217+0.8932N-0.2995m-0.0995 t -0.0189 D0.075-0.0023 wn +0.1600 D2.0-0.0670 D0.002-0.0210...(38)

For M7A0.35W2:
L.P.= 0.4076+0.9685N-0.3559m-0.1186 t +0.0101 D0.075-0.0346 wn -0.1832F D2.0-0.1289 D0.002-0.0069(39)

186

The L.P. values obtained by these equation are summarized in Table C-43. All predictions were
correct but obtained AAE &RMSE are mentioned in comparative Table C-43 but AAE and
RMSE value is too high in comparison with the results obtained by ANN and ANFIS.
Table 5.6: Statistical Parameters Calculated between Modified Seeds method and Developed Models.
M8A0.35W2

M8A0.35W4

M7A0.35W2

L.P. by Modified Seed's Method

Models
ANN

ANFIS

MLR

AAE (%)

13.862

18.689

>100

14.160

13.815

>100

10.471

14.334

33.715

RMSE (%)

21.112

24.588

>100

17.319

18.158

>100

12.318

21.794

47.191

COD

0.9620

0.9208

0.7365

0.9744

0.9716

0.8405

0.9736

0.9299

0.7702

1.6
1.4

ANN

ANFIS

MLR

ANN

ANFIS

MLR

L.P. by Modified Seed's Method for M8A0.35W2 for Zone-V

ANN

From ANN Method From ANFIS Method

y = 0.852x + 0.0114 y = 0.9239x +0.0152
R = 0.9620
R = 0.9208

ANFIS

From MLR Technique

y = 0.9399x + 0.0053
R = 0.7365

1.2
1
0.8
0.6
0.4
0.2
0

0.4
0.6
0.8
1
1.2
1.4
1.6
L.P. by ANN, ANFIS and MLR Technique
Fig. 5.109: Comparative Study of L.P. between Modified Seeds Method and Various Computational Methods i.e.
ANN, ANFIS and MLR for Zone V for Model M8A0.35W2.
-0.2

0.2

187

L.P. by Modified Seed's Method for M8A0.35W4 for Zone-V

L.P. by Modified Seed's Method

From ANN Method

y = 0.9703x - 0.0035
R = 0.9744

0.9
0.8

From ANFIS Method

y = 1.0434x - 0.0067
R = 0.9716

From MLR Technique

y = 0.8122x + 0.0484
R = 0.8405

ANN
ANFIS
MLR

0.7
0.6
0.5
0.4
0.3
0.2
0.1
0

0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
L.P. by ANN, ANFIS and MLR Technique
Fig. 5.110: Comparative Study of L.P. between Modified Seeds Method and Various Computational Methods i.e.
ANN, ANFIS and MLR for Zone V for Model M8A0.35W4.

L.P. by Modified Seed's Method

0.1

L.P. by Modified Seed's Method for M7A0.35W2 for Zone-V

1.6

From ANN Method

y = 1.0899x - 0.0296
R = 0.9736

1.4
1.2

From ANFIS Method From MLR Technique

y = 1.0477x +0.0131 y = 0.7532x +0.0807
R = 0.7702
R = 0.9299

ANN
ANFIS
MLR

1
0.8
0.6
0.4
0.2
0
0

0.2

0.4
0.6
0.8
1
1.2
1.4
1.6
L.P. by ANN, ANFIS and MLR Technique
Fig.5.111: Comparative Study of L.P. between Modified Seeds Method and Various Computational Methods i.e.
ANN, ANFIS and MLR for Zone V for Model M7A0.35W2.

Results of these three were compared and are shown in Fig 5.109, 5.110 and 5.111. It can be
seen from the table that the highest R2 value were yielded by ANN models and lowest R2 value
were yielded by MLR models (Table 5.6).

188

5.3.5.2 Results of developed model based on I & B Techniques

In this case combinations of M6A0.15W4, M6A0.15W6 & M6A0.15W8 were used to develop ANN,
ANFIS and MLR models.
As per Table C-39 & C-42 ANN & ANFIS models predicted all the L.P. values in agreement
with L.P. obtained by I&B method. AAE for the models M8A0.35W0, M8A0.25W0 and M7A0.35W0
were 7.867%, 9.925% and 9.258% whereas RMSE for the models M8A0.35W0, M8A0.25W0 and
M7A0.35W0 are 10.021%, 13.777% and 13.213%. The minimum AAE & RMSE value were
1.374% & 2.481% for ANN model M6A0.15W4 (N39) and M7A0.15W8 (N4). In contrast to this
maximum AAE & RMSE for ANN model were 9.928% and 13.213% obtained from network
N22.
Minimum AAE & RMSE were 3.686% & 5.382% obtained for ANFIS models M6A0.15W0 and
M6A0.15W8. The maximum AAE is 27.178% for ANFIS model M8A0.35W6 whereas maximum
RMSE is 63.579% for ANFIS model M8A0.25W0.
The developed MLR M8A0.35W0, M8A0.25W0 and M7A0.35W0 models are shown below:
For M8A0.35W0:
L.P.= 0.0341+1.0987N-0.1500m+0.0251 t-0.0976D0.075+0.0889wn -0.0758 D2.0+0.1736 D0.002-0.0795(40)

For M8A0.25W0:
L.P.= 0.0024+1.1004N-0.1488m+0.0069 t -0.0684D0.075+0.0554wn -0.0201 D2.0+0.0811 D0.002-0.0543(41)

For M7A0.35W0:
L.P.= -0.0178+1.1014N-0.1227m+0.0105 t -0.0701 D0.075+0.0598 wn-0.0239 D2.0+0.0907 D0.002-0.0562.(42)

Results obtained through MLR techniques for model M8A0.35W0, M8A0.25W0 & M7A0.35W0 are
shown in Table C-44. AAE, RMSE and COD of the above mentioned combinations are tabulated
on Table 5.7.

189

Table 5.7: Statistical Parameters Calculated between I&B method and Developed Models.
M8A0.35W0

M8A0.25W0

M7A0.35W0

Models
ANN

ANFIS

MLR

ANN

ANFIS

MLR

ANN

ANFIS

MLR

AAE (%)

14.655

18.689

>100

9.336

13.815

>100

8.802

14.334

33.715

RMSE (%)

20.928

24.588

>100

13.664

18.158

>100

12.709

21.794

47.199

COD

0.9748

0.7392

0.9130

0.9830

0.9348

0.9163

0.9852

0.9505

0.9027

Figure 5.112, 5.113 and 5.114 it depicts the comparative study between ANN, ANFIS and MLR
models.

ANN models for the combinations M8A0.25W0 & M7A0.35W0 showed better COD

compared to other models.

L.P. by I&B Method for M8A0.35W0 for Zone-V
From ANN Methods
y = 1.0624x -0.0024
R = 0.9748

L.P. by I&B Method

0.6
0.5

From ANFIS Method

y = 0.8597x + 0.0255
R = 0.7392

ANN
ANFIS
MLR

From MLR Technique

y = 2.448x - 0.0952
R = 0.913

0.4
0.3
0.2
0.1
0
-0.1
-0.2

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

L.P. by ANN, ANFIS and MLR Method

Fig. 5.112: Comparative Study of L.P. between I&B Method and Various Computational Methods i.e. ANN, ANFIS
and MLR for Zone V for Model M8A0.35W0.

190

L.P. by I&B Method for M8A0.25W0 for Zone-V

1.6

From ANN Methds From ANFIS Method

y = 0.9403x + 0.0093 y = 1.0165x + 0.0067
R = 0.983
R = 0.9348

L.P. by I&B Method

1.4
1.2

ANN

ANFIS

MLR

From MLR Technique

y = 1.3338x - 0.0702
R = 0.9163

1
0.8
0.6
0.4
0.2
0
-0.2 0

0.2

0.4

0.6

0.8

1.2

1.4

1.6

L.P. by ANN, ANFIS and MLR Technique

Fig. 5.113: Comparative Study of L.P. between I&B Method and Various Computational Methods i.e. ANN, ANFIS
and MLR for Zone V for Model M8A0.25W0.

1.6

ANFIS

From ANN Methods From ANFIS Method From MLR Technique

y = 1.0049x -0.0036 y = 1.0049x + 0.0076
y = 1.3406x - 0.0685
R = 0.9852
R = 0.9505
R = 0.9027

1.4
L.P. by I&B Method

ANN

L.P. by I&B Method for M7A0.35W0for Zone-V

1.2

MLR

1
0.8
0.6
0.4
0.2
0

-0.2

0.2

0.4
0.6
0.8
1
1.2
L.P. by ANN, ANFIS and MLR Technique

1.4

1.6

Fig. 5.114: Comparative Study of L.P. between I&B Method and Various Computational Methods i.e. ANN, ANFIS
and MLR for Zone V for Model M7A0.35W0.

5.3.5.3 Results of developed Models based on T&Y Method

In this Zone an average chances of soil liquefaction by T&Y method is 44.515% (Fig. 5.51). The
combinations of M8A0.35W0, M7A0.35W0 and M8A0.25W0 were found to be most susceptible to
liquefaction. Therefore ANN, ANFIS and MLR models results were predicted and compared for
these combinations.

191

It can be also seen from Table C-39 & C-42 that liquefaction potential predicted by ANN and
ANFIS respectively is in accordance with T&Y method.
Minimum AAE & RMSE were 2.005% and 4.108% obtained for model M6A0.15W8 and the
network is N34 (Annexure-D). Similarly maximum AAE and RMSE is 35.351% and 42.619% for
model M8A0.15W0 obtained respectively by ANN models. AAE, RMSE and COD are shown in
the table 5.8 for the combinations M8A0.35W0, M7A0.35W0 and M8A0.25W0.
It indicates that both computational methods have strong training and predicting capability in
calculating liquefaction potential from used input parameters which is shown in Table C-42.
Table 5.8 indicates that performance of ANN was better than ANFIS in terms of statistical
parameters.
MLR models M8A0.35W0, M7A0.35W0 and M8A0.25W0 were developed whose multi linear
equations are to predict L.P. on the basis of following relationship:
For M8A0.35W0:
L.P.= -0.2395+0.5193N-0.3790m+0.0706 t +0.2666 D0.075-0.0199 wn +0.1653F-0.0209G+0.0259.(43)

For M7A0.35W0:
L.P.= -0.1846+0.5441N-0.3787m+0.0595 t +0.2891 D0.075-0.0300 wn +0.1479 D2.0-0.0164 D0.002+ 0.0107.(44)

For M8A0.25W0:
L.P.= -0.1239+0.5648N-0.3741m+0.0790 t +0.3282 D0.075-0.0473 wn +0.1223 D2.0-0.0104 D0.002+ 0.0136.(45)

Results obtained through MLR techniques for model M8A0.35W0, M8A0.25W0 & M7A0.35W0 were
developed which is shown in Table 5.45. AAE & RMSE value is much higher in comparison to
ANN and ANFIS but coefficient of determination is good.
Fig. 5.115, 5.116 and 5.117 showed comparative results of all models. Best linear fit equation
and coefficient of determination are also showed in these figures. On the basis of these results
following statistical parameters like AAE, RMSE and COD are presented in Table 5.8. It can be

192

said from these results that computational method showed better performance in terms of
statistical parameter in comparison to MLR models.
Table 5.8: Statistical Parameters Calculated between T&Y method and Developed Models.
M8A0.35W0

M7A0.35W0

M8A0.25W0

Models
ANN

ANFIS

AAE (%)

13.795

16.650

12.243

22.280

9.941

19.730

RMSE (%)

18.529

24.380

16.933

33.280

13.004

31.800

COD

0.9798

0.9198

0.9980

0.8488

0.9700

0.9405

0.2860

ANN

ANFIS

MLR

0.3145

ANN

L.P. by T&Y Method for M8A0.35W0 for Zone-V

0.45
L.P. by T&Y Method

MLR

0.35

ANFIS

MLR

0.3235

From ANN Methods

y = 0.9851x-0.0024
R = 0.9798
From ANFIS Method
y = 0.8414x +0.0044
R = 0.9198

0.25
0.15

From MLR Method

y = 0.9639x + 0.0006
R = 0.286

0.05
-0.05 0

0.1

0.2

0.3

0.4

L.P. by ANN, ANFIS and MLR Technique

-0.15

Fig.5.115: Comparative Study of L.P. between T&Y Method and Various Computational Methods i.e. ANN,
ANFIS and MLR for Zone V for Model M8A0.35W0.

L.P. by T&Y Method for M8A0.25W0 for Zone-V

L.P. by T&Y Method

0.6
0.5
0.4

From ANFIS Method

y = 0.9923x - 0.0081
R = 0.8488

0.3
0.2

From MLR Method

y = 0.8951x + 0.0071
R = 0.3145

0.1
0
0
-0.1

ANN
ANFIS
MLR
From ANN Methods
y = 1.0901x-0.025
R = 0.998

0.1

0.2
0.3
0.4
0.5
L.P. by ANN, ANFIS and MLR Technique

0.6

Fig. 5.116: Comparative Study of L.P. between T&Y Method and Various Computational Methods i.e. ANN,
ANFIS and MLR for Zone V for Model M8A0.25W0.

193

L.P. by T&Y Method

L.P. by T&Y Method for M7A0.35W0 for Zone-V

From ANN Methods
y = 1.0489x-0.0102
R = 0.97

0.4

From ANFIS Method

y = 0.9769x - 0.0045
R = 0.9405

From MLR Technique

y = 0.9391x + 0.0014
R = 0.3235

ANN
ANFIS
MLR

0.3
0.2
0.1
0
0

-0.1

0.1
0.2
L.P. by ANN, ANFIS and MLR Technique

0.3

0.4

Fig. 5.117: Comparative Study of L.P. between T&Y Method and Various Computational Methods i.e. ANN,
ANFIS and MLR for Zone V for Model M7A0.35W0.

194

5.3.6 L.P. by ANN, ANFIS & MLR Methods of Combined Zone

In this section, datasets of all zones i.e. zone-I to zone-V have been used to develop ANN,
ANFIS and MLR models to predict L.P. To train and validate the models of combined zone same
datasets were used together as it was used in case of individual zones. Total 644 datasets were
used for combined zone models .Out of 644, 529 datasets were used for training and 115 datasets
were used for validation
5.3.6.1. Results of Developed Models Based on Modified Seeds Method
Liquefaction potential for combined zone calculated by Modified Seeds method for all 45
combinations which

summarized in Annexure B. Average chances of liquefaction of soil for

combined zone is 35.1724% The maximum chance of liquefaction condition occurs for the
combination M8A0.35W2 (Fig 5.61). Three combinations with maximum chances were selected
for comparative study.
Results predicted by ANN and ANFIS were summarized in Table C-46 and Table C-49. It can be
observed that prediction in accordance with ANN results by actual liquefaction potential. ANN
showed good prediction accuracy in model M8A0.35W2 in which only 2 results i.e. serial no. 74
and 84; model M8A0.35W4 in which only 1 result i.e. serial no. 74 and model M8A0.35W6 in which
only 2 results i.e. serial no. 16 & 30 are incorrect predictions. AAE for these three models i.e.
M8A0.35W2, M8A0.35W4 and M8A0.35W6 are 13.724%, 17.533% and 8.369% and RMSE value are
21.856%, 63.419% and 13.716% respectively. The optimized networks are N40, N40 and N85
respectively.
The minimum AAE given by ANFIS model is 0.943 for the modelM6A0.15W0 and for RMSE is
2.364 also for the model M6A0.15W0. The maximum AAE given by ANFIS model is 47.078 for
the modelM6A0.35W4 and for RMSE is 133.63 for the model M8A0.35W2. Only three predictions

195

done by ANFIS models are wrong i.e. at serial no. 74 and 84 by ANFIS model M8A0.35W2 and at
serial no. 74 by ANFIS model M8A0.35W4 (Table 5.54). AAE for these three models i.e.
M8A0.35W2, M8A0.35W4 and M8A0.35W6 are 14.282%, 13.336% and 12.356% whereas, RMSE
value are 22.206%, 33.889% and 26.688% respectively.
The MLR models for these three combinations M8A0.35W2, M8A0.35W4 & M7A0.35W6 were
developed as depicted below:
For M8A0.35W2:
L.P.= 0.0257+0.8915N-0.4110m-0.0041 t +0.0547 D0.075-0.0208 wn +0.0268 D2.0 D0.002-0.0530

0.0145..(46)

For M8A0.35W4:
L.P.= 0.0918+0.9294N-0.3795m-0.0537 t +0.0329 D0.075-0.0257 wn +0.0265 D2.0-0.0756 D0.002+0.0142..(47)

For M8A0.35W6:
L.P.= 0.1979+1.1570N-0.2679m+0.0363 t -0.0298 D0.075+0.1148 wn

+0.0210 D2.00.1248 D0.002-0.0200(48)

Liquefaction potential estimated by these three equations are is summarized in Table 5.52.
According to Table 5.52, prediction in comparison to modified Seeds method is very poor.
Seven incorrect predictions were found in model M8A0.35W2, one in model M8A0.35W4 and ten in
model M8A0.35W6. Prediction accuracy is less than 90% through MLR. AAE and RMSE values
are also calculated and summarized in table 5.52.

196

L.P. by Modified Seed's Method

L.P. by Modified Seed's Method for M8A0.35W2 for Combined Zone

From ANN Method
y = 0.8872x + 0.0229
R = 0.9483

From ANFIS Method

y = 0.8953x +0.0189
R = 0.9158

From MLR Technique

y = 1.3695x + 0.0456
R = 0.8698

ANN
ANFIS
MLR

1.5

0.5

0
0

0.2

0.4
0.6
0.8
1
L.P. by ANN, ANFIS and MLR Technique

1.2

1.4

1.6

-0.5
Fig.5.118: Comparative Study of L.P. between Modified Seeds Method and Various Computational Methods i.e.
ANN, ANFIS and MLR for Combined Zone for Model M 8A0.35W2.

L.P. by Modified Seed's Method

L.P. by Modified Seed's Method for M8A0.35W4 for Combined Zone

1.8
1.6
1.4

From ANN Method

y = 0.9671x + 0.0085
R = 0.9720

ANN
ANFIS

From ANFIS Method From MLR Technique

y = 0.9418x +0.0182 y = 0.8986x + 0.0325
R = 0.9703
R = 0.9451

MLR

1.2
1
0.8
0.6
0.4
0.2
0
-0.2 0
-0.4

0.2

0.4

0.6

0.8

1.2

1.4

1.6

L.P. by ANN, ANFIS and MLR Technique

Fig. 5.119: Comparative Study of L.P. between Modified Seeds Method and Various Computational Methods i.e.
ANN, ANFIS and MLR for Combined Zone for Model M 8A0.35W4.

197

L.P. by Modified Seed's Method

L.P. by Modified Seed's Method for M8A0.35W6 for Combined Zone

From ANN Method
y = 0.9794x + 0.0048
R = 0.9778

1.6
1.4

From ANFIS Method

y = 0.9606x +0.0188
R = 0.9895

From MLR Technique

y = 0.5979x + 0.0336
R = 0.9504

1.2
1
0.8
0.6
0.4

ANN

ANFIS

MLR

0.2
0
-0.2

0.2

0.4
0.6
0.8
1
L.P. by ANN, ANFIS and MLR Technique

1.2

1.4

1.6

-0.4
Fig. 5.120: Comparative Study of L.P. between Modified Seeds Method and Various Computational Methods i.e.
ANN, ANFIS and MLR for Combined Zone for Model M 8A0.35W6.

(show comparative table here then discuss what models)

Best COD for these three models for ANN, ANFIS and MLR techniques are varies from 0.9483
to 0.9778 from ANN predictions, 0.9158 to 0.9895 from ANFIS predictions whereas 0.8689 to
0.9504 from MLR techniques.
5.3.6.2 Results of Developed Models based on I&B Method
Similarly, three combination M8A0.35W0, M8A0.35W2 & M8A0.35W4 were selected using Fig5.62
for comparative study between ANN,ANFIS and MCR model. Using actual and predicted
values, statistical parameters such as AAE and RMSE were also calculated for each
combinations of MxAyWz and the results are summarized in Table C-47 The lowest AAE =
0.339% yielded by model M6A0.15W8 whereas highest AAE = 15.077% was yielded by model
M8A0.15W0. Similarly lowest RMSE = 1.375% yielded by model M6A0.15W8 and highest RMSE
= 62.977% was yielded by model M8A0.35W4. AAE for model M8A0.35W0, M8A0.35W2 &
M8A0.35W4 are 16.024%, 13.606% and 15.334% whereas RMSE values obtained for combined
198

zone by I&B method are 26.812%, 21.274% and 23.196% (Table 5.47). These models are
optimized by the network N28, N64 and N52.Only three prediction by ANN

models were

wrongwhich is indicated in Table 5.48.

Table 5.51 is showing the prediction by ANFIS models can be seen that the prediction at serial
no. 74 by ANFIS model M8A0.35W0 13, 16, 18 and 74 for model M8A0.35W0 and M8A0.35W2
respectively are mismatch.. AAE for model M8A0.35W0, M8A0.35W2 & M8A0.35W4 are 16.418%,
14.702% and 14.843% whereas, RMSE values obtained for combined zone by I&B method are
42.681%, 23.669% and 24.031% (Table 5.50).
MLR models for the combination M8A0.35W0, M8A0.35W2 & M8A0.35W4 have been developed.
The relationship with L.P. developwed by these models are as follows:
For M8A0.35W0:
L.P.= -0.1979+1.1570N-0.2679m+0.0363t-0.0298D0.075+0.1148wn+0.0210 D2.0+0.1248 D0.002-0.0200....(49)

For M8A0.35W2: ..
L.P.= 0.0798+1.0643N-0.4023m-0.0562 t +0.0051D0.075+0.0393 wn -0.0438 D2.0+0.0299 D0.002-0.0270(50)

For M8A0.35W4:
L.P.= 0.1398+1.0175N-0.3731m-0.0983 t -0.0036 D0.075+0.0057 wn -0.0022 D2.0-0.0122 D0.002-0.0345.(51)

Liquefaction potential values calculated from these three equations are summarized in table 5.53.
Table 5.53 showed that the model M8A0.35W0, M8A0.35W2 & M8A0.35W4 having AAE value
16.085%, 13,589% and 14.451% respectively whereas RMSE value are 27.020%, 21.241% and
22.245% respectively are also shown in the table.
From Fig. 5.121, 5.122 and 5.123 it is clearly seen that for all three models i.e. M8A0.35W0,
M8A0.35W2 & M8A0.35W4 ANN showed better coefficient of determination (COD) value that
varies from 0.8773 to 0.9820 from ANN predictions, 0.9412 to 0.9859 from ANFIS predictions
whereas 0.8904 to 0.9156 from MLR techniques.

199

ANN

L.P. by I&B Method for M8A0.35W0 for Combined Zone

L.P. by I&B Method

ANFIS

From ANN Methods From ANFIS Method From MLR Technique

y = 0.9100x -0.0112
y = 1.0056x -0.0005 y = 0.9632x - 0.0075
R = 0.8904
R = 0.9820
R = 0.9859

MLR

1.5

0.5

0
0

0.2

0.4

0.6

0.8

1.2

1.4

1.6

L.P. by ANN, ANFIS and MLR Technique

-0.5
Fig. 5.121: Comparative Study of L.P. between I&B Method and Various Computational Methods i.e. ANN, ANFIS
and MLR for Combined Zone for Model M8A0.35W0.

L.P. by I&B Method for M8A0.35W2for Combined Zone

From ANN Method From ANFIS Methods
y = 0.9904x - 0.0029
y = 0.9904x -0.0029
R = 0.9707
R = 0.9412

L.P. by I&B Method

1.75

From MLR Methods

y = 0.9268x + 0.0019
R = 0.9134

ANN

ANFIS

MLR

1.25

0.75

0.25

0
-0.25

0.2

0.4
0.6
0.8
1
L.P. by ANN, ANFIS and MLR Technique

1.2

1.4

1.6

Fig. 5.122: Comparative Study of L.P. between I&B Method and Various Computational Methods i.e. ANN, ANFIS
and MLR for Combined Zone for Model M8A0.35W2.

200

L.P. by I&B Method for M8A0.35W4for Zone-Combsite

From ANN Methods
y = 0.9017x +0.0184
R = 0.8773

1.8
1.6

From ANFIS Method

y = 0.9601x - 0.0036
R = 0.9589

From MLR Methods

y = 0.9094x + 0.0203
R = 0.9156

ANN
MLR

ANFIS

1.4

1.6

L.P. by I&B Method

1.4
1.2
1
0.8
0.6
0.4
0.2
0
-0.2

0.2

0.4
0.6
0.8
1
L.P. by ANN, ANFIS and MLR Technique

1.2

Fig. 5.123: Comparative Study of L.P. between I&B Method and Various Computational Methods i.e. ANN, ANFIS
and MLR for Combined Zone for Model M6A0.35W4.

5.3.6.3 Results of developed models based on T&Y Techniques

The chances of soil liquefaction evaluated by T&Y method is 41.195%. Similar to earlier
condition three combination M8A0.35W0, M7A0.35W0 & M6A0.35W0 were selected for comparative
study.
Liquefaction potential predicted by ANN models are shown in Table C-48. Liquefaction
potential predicted by ANN models are in good agreement with T&Y method except one
prediction by ANN model M7A0.35W0 is incorrect.
In this case predictions capabilities of ANFIS models are good in compare to ANN models as no
case of wrong prediction have been found. It indicates that both computational methods have
strong training and predicting capability in calculating liquefaction potential from used input
parameters. The AAE for model M6A0.35W0 for ANN and ANFIS was 3.275% and 7.810%
respectively whereas RMSE for these is 6.408% and 12.090% respectively. Similarly the AAE

201

for model M7A0.35W0 for ANN and ANFIS was 8.479% and 11.941% respectively whereas
RMSE for these is 12.874% and 32.909% respectively. Similarly the AAE for model M 8A0.35W0
for ANN and ANFIS was 10.872% and 10.995% respectively whereas RMSE for these is
48.404% and 20.673% respectively (Table 5.48 and 5.51).

The MLR models were developed for the combination M8A0.35W0, M7A0.35W0 & M6A0.35W0 .
The developed relationship by these models are as follows:
For M8A0.35W0:
L.P.=0.2395+0.5193N-0.3790m+0.0706t+0.2666 D0.075-0.0199 wn +0.1653F-0.0209G+ 0.0259(52)

For M7A0.35W0:
L.P.=-0.1846+0.5441N-0.3787m+0.0595t+0.2891D0.075-0.0300wn+0.1479D2.0-0.0164D0.002+0.0107..(53)

For M6A0.35W0:
L.P.=-0.1239+0.5648N-0.3741m+0.0790t+0.3282D0.075-0.0473wn+0.1223D2.0-0.0104D0.002+0.0136..(54)

Liquefaction potential values, AAE and RMSE values are summarized in table 5.54 for these
three models.

202

L.P. by T&Y Method

ANN

L.P. by T&Y Method for M8A0.35W0 for Combined Zone

From ANN Methods
y = 0.9963x - 0.0002
R = 0.9978

From ANFIS Method

y = 1.0088x -0.0008
R = 0.9902

ANFIS

From MLR Technique

y = 0.867x + 0.0372
R = 0.6149

MLR

1.5

0.5

0
0

0.2

0.4

0.6

0.8

1.2

1.4

1.6

L.P. by ANN, ANFIS and MLR Technique

-0.5

Fig.5.124: Comparative Study of L.P. between T&Y Method and Various Computational Methods i.e. ANN,
ANFIS and MLR for Combined Zone for Model M 8A0.35W0.

L.P. by T&Y Method for M7A0.35W0 for Combined Zone

From ANN Methods From ANFIS Method

y = 0.9862x + 0.0017 y = 0.9731x +0.0053
R = 0.9817
R = 0.9411

From MLR Technique

y = 0.9184x + 0.0295
R = 0.6746

L.P. by T&Y Method

1.5

0.5

0
0

0.2

0.4

0.6

0.8

1.2

1.4

1.6

L.P. by ANN, ANFIS and MLR Technique

-0.5
Fig. 5.125: Comparative Study of L.P. between T&Y Method and Various Computational Methods i.e. ANN,
ANFIS and MLR for Combined Zone for Model M 7A0.35W0.

203

L.P. by T&Y Method for M6A0.35W0 for Combined Zone

1.6

From ANN Methods

y = 0.9600x-0.009
R = 0.9413

1.4

From ANFIS Method

y = 0.9307x +0.0198
R = 0.9193

From MLR Technique

y = 0.9365x + 0.0201
R = 0.7152

L.P. T&Y Method

1.2
1
0.8
0.6
ANN

0.4

ANFIS
0.2

MLR

0
-0.2

0.2

-0.4

0.4

0.6

0.8

1.2

1.4

1.6

L.P. by ANN, ANFIS and MLR Technique

Fig. 5.126: Comparative Study of L.P. between T&Y Method and Various Computational Methods i.e. ANN,
ANFIS and MLR for Combined Zone for Model M 6A0.35W0.

Fig. 5.124, 5.125 and 5.126 showed comparative analysis between results obtained by T&Y
method and results predicted by different developed models. Best linear fit equation and
coefficient of determination are also showed in these figures. From these figures it could be seen
that COD value is better in ANN than ANFIS and MLR. MLR showed poor COD value for these
three developed models.

It can be referred from these results that computational method showed higher performance in
terms of statistical parameters in comparison to MLR models. However regression models have
their own advantages such as fast convergence and generalization etc.

204

Chapter 6

CONCLUSIONS
6.1 CONCLUSIONS
In this paper, liquefaction potential in terms of FOS were evaluated by considering based on the
spatial variation like three different levels of potential earthquake magnitude, five different level
of depth of water table and three different types of horizontal acceleration. Semi empirical
methods like modified Seeds method, Idriss and Boulanger (I&B) method and Tokimatsu and
Yoshimi (T&Y) method used which are based on the available SPT profile of soil. The obtained
FOS was used for the development of different computational models like based on ANN,
ANFIS and MLR techniques. On the basis of present research work following points have been
concluded:
The parametric variations considered for conventional methods laid the important role in
evaluation of liquefaction potential like magnitude, water table and acceleration.
It has been observed that higher magnitude with other variations is critical after
M8A0.25W0 conditions for all the zones in case of liquefaction.
Liquefaction potential evaluated by Modified Seeds method, I&B Method and T&Y
method highlights the different approach with vulnerable conditions. Out of these
methods I&B and T&Y methods are more conservative compare to modified Seeds
method Therefore

I&B or T&Y approach can be used for evaluating liquefaction

potential.
The ANN model develop to predict liquefaction potential on the basis of these methods
are depicting better predictive capability when optimised models are considered for input
205

values i.e. SPT-N value (N), depth (d), bulk density (f), particle size finer than 2.00 mm,
0.002 mm and 0.075 mm , natural/field moisture content (wf), and angle of internal
friction.
Comparing the statistical parameters of different ANN models the best combinations of
models developed for zone I is M8A0.35W2 whose COD is 0.9652 by T&Y approach for
network N52, for zone II is M6A0.35W0 whose COD is 0.9979 by T&Y approach for
network N85, for zone III is M8A0.35W2 whose COD is 0.9919 by I&B approach for
network N4, for zone IV is M8A0.35W0 whose COD is 0.998 by T&Y approach for
network N45, for zone V is M8A0.35W4 whose COD is 0.9980 by I&B approach for
network N85., for combined zone is M8A0.35W2 whose COD is 0.9986 by I&B approach
for network N64.
Developed models through ANFIS represents the combined approach which is
comparable with ANN models with the similar input variables.
As per the statistical parameters, the optimised ANFIS models for zone I is M8A0.35W2
having COD 0.9651 by I&B approach. Similarly, for zone II M7A0.35W0 and M8A0.35W0
having COD 0.9946 by T&Y approach, for zone III M8A0.35W4 having COD 0.9802 by
I&B approach, for zone IV M8A0.25W0 having COD 0.9897 by I&B approach, for zone V
M7A0.35W0 having COD is 0.9505 by I&B approach and for combined zone M8A0.35W6
having COD 0.9906 by modified Seeds approach.
Comparing the statistical parameters of different MLR models, the best combinations of
developed models for zone I is M8A0.35W6 whose COD is 0.9591 by modified Seeds
approach, zone II is M8A0.35W2 whose COD is 0.9849 by modified Seeds approach, zone
206

III is M8A0.35W6 whose COD is 0.9713 by I&B approach, for zone IV is M8A0.35W4
whose COD is 0.9567 by modified Seeds approach, for zone V is M8A0.25W0 whose
COD is 0.9163 by I&B approach and the best combinations of models for developed for
combined zone is M8A0.35W6 whose COD is 0.9438 by I&B approach.
Prediction capability for ANN approach and ANFIS approach are too good for individual
zones and combined zones i.e. >99% for more vulnerable zones. Whereas, prediction
capability by MLR techniques for individual zones are better than combined zones.
The developed models by various approach have been compared for individual and
combined zones which attributed that ANN and ANFIS techniques are reliable compare
to regression technique. Hence, soft computing methods of model development can be
applied for such studies.
The overall assessment of Allahabad city for liquefaction condition zone V is the worst
affected area where as zone IV is the least affected area. Therefore, anti-liquefaction
measures should be taken up properly for vulnerable areas or zones.

6.2 Suggestions for further research

Although the work presented in this thesis give the impression of an extensive and
comprehensive study on the Liquefaction Potential of Soil using Soft Computing Methods
for Allahabad city. Many areas can be identified and suggested for future research which
can improve the confidence level and reliability of the present results.

207

The present work adopts the attenuation relation model viz., It should be mentioned here that the
present analysis has used semi empirical methods like modified Seeds method, Idriss and
Boulanger (I&B) method and Tokimatsu and Yoshimi (T&Y) method which are based on the

208