You are on page 1of 16

Chapter 2

LITERATURE REVIEW

2.1 Literature Review

Generation Of Artificial Earthquake Motion Using Observed Earthquake Motions -


Masahiko NAKAMURA, Fumio SASAKI, Kazuki YOKOYAMA, Tetsuo TAMAOKI,
Akira TANABE, Wataru MIZUMACHI and Michio YAMADA [1]
In this paper, they propose an integrated approach to seismic behavior that satisfies the target
response spectrum by observing seismic motion. The integrated approach to seismic action does
not use any parametric model, such as the regression model. The method utilizes temporal
frequency characteristics and amplitude characteristics, including observing seismic motion.
Here, they show that the integrated method of seismic action can carry out artificial seismic
motion satisfying the target response spectrum by using 30 observations of seismic motion. In
addition, the energy contribution of the artificial earthquake motion produced by this method is
not affected by the energy contribution of the particular observation seismic motion.

Generation of Artificial Earthquake Wave Compatible with Design Spectrum - Dae-Han


Jun, Pyeong-Doo Kang and Jae-Ung Kim [2]
In this paper, they explain that the seismic safety of a building structure can be assessed by the
nonlinear behavior of the structure caused by ground motion. The input ground motion
characteristics and structural dynamic characteristics are important factors influencing the
structural seismic response. The ground motion used in the structural seismic response includes
various characteristics, depending on the fault mechanism of the earthquake, the propagation of
the waves and the enlargement of the soil type. Therefore, quantitative inspection of all the
factors is a difficult task. In this study, simulated ground motion is developed for the design of
the spectrum, and the feasibility of the input ground motion is evaluated by the nonlinear seismic
response analysis of the multi-layer frame structure.
New Method For Generation Of Artificial Earthquake Record - M. Fadavi Amiri, A.
Akbari and G. Ghodrati Amiri [3]
This paper presents the generation and simulation of acceleration records. By comparing the
results, it can be concluded that the generated acceleration record contains the maximum
frequency content recorded by the actual acceleration and the response spectrum of the artificial
acceleration record is appropriately adjusted to the response spectrum of the actual acceleration
record. By comparing the data recorded manually and the actual acceleration, we can conclude
that the proposed model has some degree of ability to simulate the number of records. Also by
comparing the response spectra, we note that the average of the response spectra of the generated
records is appropriately adjusted to the average of the response spectra of the generated records.
Thus, it can be assumed that the proposed model can simulate earthquakes from the perspective
of domain and frequency content.

Generation of artificial earthquakes for matching target response unsmooth spectrum via
wavelet package transform - Kang PENG, Ze-wei WANG, Jin-jin SUN [4]
In this paper, they explain the calculation of wavelet packet components, wavelet packet
decomposition of the components of the seismic recording of the frequency band center
frequency and wavelet packet response spectrum peak frequency between the existence of a
certain deviation. Five strategies for selecting iterative points are proposed. The example shows
that the peak response frequency of the wavelet packet component is chosen as the iteration point
to match the target spectrum well and is stable. So, this strategy is recommended as the best
choice. Numerical examples also confirm that the peak frequency of the response spectrum is
stable during the iteration process. Thus, the peak frequency achieved in the first signal
decomposition can be chosen as the iteration point for the next iteration and the peak frequency
can be avoided repeatedly.

Predicting the Seismic Performance of Cylindrical Steel Tanks Using Artificial Neural
Networks (ANN) - Mehran S. Razzaghi, Alireza Mohebbi [5]
In this paper, we propose an iterative scheme for stochastic model design, which will produce the
artificial ground acceleration time course which fits the predetermined ground response
spectrum. The plan applies to China's Lianyungang nuclear power plant will be built seismic
data. Research on Matching Time of Design Acceleration Time of Lianyungang Nuclear Power
Plant and Design Spectrum of HAF 0101. Shows compliance with the HAF 0101 design
spectrum according to ASCE 4-86 and NUREG 0800 rule design time history.

Scatter Of Dynamic Response And Damage Of An Arch Dam Subjected To Artificial


Earthquake Accelerograms – Jin Ting Wang, Ai-Yun Jin, Xiu-Li Du, Ming- Xin Wu [6]
The aim of this study is to study the possibility of using artificial neural network to predict the
seismic performance of liquid storage tanks. To this end, selected 240 earthquake-affected tanks
of the database. Design and train five different propagation ANNs, 190 data for different
architectures. The results of this study show that the artificial neural network has the feasibility
of predicting the seismic performance of liquid storage tanks. According to the ANN structure
evaluation results given the best correlation factor (R2 = 0.7793) was obtained from the [6 6 6 1]
model with the smallest mean square error. For the PGA less than 0.3 g, the ANN model
accurately predicts the damage state. For 0.6 g <PGA <0.7 g, the ANN prediction is also
acceptable, but for PGA> 0.7 g, due to the lack of training data, In most cases the predicted value
is higher than the actual damage state, so there is no accurate prediction. It is worth mentioning
that a seismic event with a PGA higher than 0.7 g is a very strong earthquake and usually has a
longer return period.

Preliminary numerical analysis of bridge piers under randomly generated artificial


earthquake-like ground motion - VB Nguyen [7]
The artificial acceleration map can also be modeled by using a stochastic model with the
advantage of obtaining a large amount of time history. This is particularly useful for probabilistic
and reliability analysis. In addition, the main characteristics of the acceleration graph can be
controlled and adjusted during simulation. In particular, the strong duration of exercise is a
parameter of the random ground motion simulation model, which can be selected according to
the situation. The variability is explained by modeling the strong motion duration as a lognormal
random variable consistent with the nearest GMPE (eg, Kempton & Steward 2006). When the
spectral shape is the only criterion for ground motion selection, the strong duration of motion can
not be controlled. At the same time, the simulated acceleration map has the advantage that a
desired acceleration map (e.g., for higher reliability and sensitivity analysis) can be obtained. In
addition, by using the same set of simulated target spectra for analysis, the comparison between
artificial and natural ground motions can be made more meaningful.

A Study On The Importance Of Seismic Parameter Selection For The Vulnerability


Assessment Of Mid-Rise Reinforced Concrete Structures - Nikolaos Nanos [8]
This paper introduces the construction of random ground motion model with time-frequency
nonstationarity based on WPT. The time-varying and frequency characteristics of the real ground
motion are modeled using the wavelet packet coefficients. The proposed model requires 13
parameters to describe the given ground motion. Then, the 13 model parameters are related to
seismic variables such as seismic volume and distance. Through two-stage regression analysis,
the mean, variability and correlation trends of these parameters can be recorded in the strong
ground motion database of large records.

Numerical Modelling of Reinforced Concrete Bridge Pier under Artificially Generated


Earthquake Time-Histories - Van Bac Nguyen [9]
The purpose of this paper is to develop an effective neural network-based model for generating
artificial earthquakes and response spectra. Five models based on neural networks are proposed
to replace the traditional process. Based on the artificial neural network model I (ANN-I) instead
of the empirical model of the parameter identification process, used to generate Fourier
amplitude spectrum. Training ANN-II and ANN-III respectively obtain the parameters of the
power spectral density function and the intensity function. ANN-IV directly generates an
acceleration response spectrum with basic information such as magnitude, epicenter distance,
site condition, and depth of focus. Finally, ANN-V reverses ANN-IV so that it can be applied to
the generation of synthetic ground motion acceleration maps that are compatible with the target
response spectrum. This study shows that the process of using neural network-based models is
suitable for generating artificial earthquakes and response spectra. Several numerical examples
are given to validate the developed model.
Generation of artificial earthquake records with a nonstationary Kanai–Tajimi model -
F.R. Rofooei, Aghababaii Mobarake , G. Ahmadi [10]
The generalized nonstationary Kanai-Tajimi model is used to describe and simulate the ground
motion time history. Amplitude and frequency instability are included in the model. The mobile
time window technique is used to evaluate the time-varying parameters of the model using actual
seismic records. The model is applicable to several Iranian earthquakes Naghan (1977), Tabas
(1978) and Manjil (1990). This shows that the model and recognition algorithm can accurately
capture the non-stationary characteristics of these seismic accelerometers. The statistical
characteristics of the spectral response curve of the simulated acceleration graph are compared
with the statistical properties of the actual record to prove the effectiveness of the model.

New Methodology Of Generating Multiple Spectrum Compatible Earthquake


Accelerogram For Taiwan Area Using Neural Networks - Chu-Chieh Jay Lin, Zhi-Min Liu
And You-Chuan Lin [11]
In this article, they explain Taiwan, located in the area where the earthquake occurred. The
geological constitution of Taiwan is constantly changing. There are at least 42 events in Taiwan.
Since the beginning of the twentieth century, eight catastrophic earthquakes have hit Taiwan; in
1999, Taichung earthquake was 7.3 times, causing serious irreversible damage, such as splitting
the ground, the bridge of the crash, the bridge of collapse, and more than 2,300 Of life. At the
same time, a large number of seismic acceleration maps of 921 odd earthquakes and their
aftershocks were recorded. Therefore, we need a method of systematically designing and
analyzing these large number of recorded acceleration maps systematically. The seismic
response spectrum is often used for structural analysis and design.

Artquake, A New Software For Generation Of Artificial Time Histories - Sergio


Hampshire C. Santos, Silvio Souza Lima And Rodrigo M. R. Rodrigues [12]
The proposed example represents an artificial earthquake that conforms to the desired response
spectrum and satisfies the constraints of velocity and displacement history after integration. The
initial value of the frequency component as a function of the power spectral density function
compatible with the given response spectrum; b) the speed and displacement values are adjusted
to zero at the beginning and the end of the time history, using the coefficients adjusted with the
least squares value process C) Since the acceleration values are manually adjusted at the
beginning and end of the time course after the process, they are manually adjusted near these
points; d) after which the conventional baseline correction is applied using the offset for the
displacement The fifth order polynomial function. By this step, the necessary adjustment of step
d) is greatly reduced, and the distortion caused by the high frequency component is actually
arranged. The proposed method is automated in a program called ARTQUAKE. An example of
the generated time course is described to illustrate the application of the program under
consideration.

Scatter of dynamic response and damage of an arch dam subjected to artificial earthquake
accelerograms - Jin-Ting Wang, Ai-Yun Jin, Xiu-Li Du , Ming-Xin Wu [13]
This paper aims to study the seismic response and damage sensitivity of arch dam to artificial
ground motion. Based on the design of the acceleration response spectrum, a total of 32
acceleration graphs were synthesized. The artificial dynamic graphs are studied, and the
nonlinear dynamic response of dam dams in southwest dam is analyzed. Dynamic response,
including displacement and joint opening and damage distribution for statistical analysis. The
results show that even if the artificial acceleration map has the same response spectrum, peak
ground acceleration, velocity and displacement, for a variety of acceleration maps, seismic
response and dam damage may be significant scattering.

Numerical simulation analyses of the collapse response of masonry structures under


earthquake - Yu Wenzheng, Tao Zhong, Luan Liwen, Zhu Ying & Chen Mingming [14]
In this paper, numerical simulation analysis of the collapse reaction of masonry structures under
earthquake in China is briefly summarized. The results show that the bottom plate and the wall
between the underground masonry windows are weak parts of the earthquake. The results show
that the collapse of the two - story masonry structure under ANSYS / LS - DYNA earthquake is
the result. From the simulation results, the bottom of the masonry structure is easy to become
part of the damage, so improve the strength of the bottom of the masonry structure is appropriate.
Not much improvement, otherwise it will cause other parts of the weak links, causing damage.
The walls between the windows are also weak in the masonry houses; therefore, in the wall
design between the windows, the width should be ensured to increase its strength to develop its
carrying capacity and to avoid damage caused by the walls between the windows The whole
building collapsed.

A Numerical Evaluation of Anti-Vibration Mechanisms Applied to Frame Structures


Under Earthquake - Luciano M. Bezerra, Rodolfo C. Carneiro [15]
In this article, we present the main earthquakes around the world and their respective damage.
Discuss the possibility of induced sensitivity. Describes the main seismic mechanism. The
advantages and disadvantages of the main anti-vibration devices, functional sketches and control
equations are also proposed. Examines an example of a man-made earthquake equivalent to the
typical building frame of El Centro. Check the frame when rigid is fixed at base and when
certain types of HRDB are used at the base. The structural response of the frame in the
displacement and shear force of the base is discussed. The HRDB isolation system used
effectively relieves the structural floor displacement of the ground motion. At the lower or higher
level of the frame structure, the relative displacement of the isolated system is significantly
reduced. The HRDB isolator absorbs most of the seismic energy. The advantage of using HRDB
in terms of the shearing force on the base makes the shear force reduced as compared to the
number of shearing forces in the case where the frame is rigidly connected to the base. In some
cases, the use of isolators is insignificant, for example, in structures with low natural frequencies,
outside the frequency range of the earthquake. In this case, there is no advantage (or
disadvantage) in the use of seismic isolation mechanisms. Horizontal displacement of the floor,
especially between the top and bottom, can be safe and comfortable.

Generation of Artificial Earthquake Accelerogram Compatible with Spectrum using the


Wavelet Packet Transform and Nero-Fuzzy - Peyman Shadman Heidari, Mohammad
Khorasani [16]
In this study, a new method of wavelet packet transform and adaptive neural network fuzzy
inference system (ANFIS) is used to generate artificial seismic records compatible with response
spectra.The proposed method is based on the development of an ANFIS network which takes the
recorded response spectrum as input and generates the wavelet packet coefficients of the wavelet
packet transform of the wavelet packet, the output of the ANFIS network, and then generates the
input of the inverse wavelet packet recording. In an illustrative example, the proposed method is
applied to the seismic acceleration map recorded by Iran. For the training of the trained ANFIS
network, when the response spectrum is given as an input, the resulting ANFIS network
generates a record that is very similar to its training set; a response spectrum with a near input, or
it synthesizes a new and realistic look Up the record, shows that the time domain and response
spectra of the generated record are similar to the original recorded acceleration graphs.

Response Spectrum Analysis for Bhuj Earthquake Recorded at Ahmadabad Station -


Kishankumar Umeshbhai Patel, Surendrasingh Kesarsingh Sindal, Sangril Harshadkumar
Patel, Tejan Niranjanbhai Patel [17]
In this paper, the study of the Bhuj earthquake response spectrum recorded at the Ahmadabad
station is studied. Meteorological data collected by the Indian Meteorological Department
recorded by Ahmadabad. The graphical relationship of the maximum acceleration, velocity and
deformation response of the Bhuj earthquake is shown here. The response of 3000 elastic single-
degree-of-freedom systems with different natural period values (Tn) using Newmark's Beta
method is calculated. This paper presents a comparison of the acceleration response spectra of
the Bhuj earthquake with the design spectral acceleration plot given in IS-CODE 1893 (Part 1):
2002. In this paper, the design response spectrum of the medium soil of Ahmadabad is reviewed
and further suggestions are put forward.

The Application of Artificial Neural Networks in Predicting Structural Response of


Multistory Building in The Region of Sumatra Island - Reni Suryanita, Hendra Jingga,
and Enno Yuniarto [18]
In this study, ANN is used to predict the structural response (story drift) of multi - story
reinforced concrete structures under the earthquake load in Sumatra region. The modal response
spectrum analysis is performed to simulate the seismic load and produce structural response data,
which is further used for ANN. The ANN architecture consists of three layers: the input layer,
the hidden layer, and the output layer. The seismic load parameters, soil conditions and
architectural geometries of 11 sites in Sumatra Island were chosen as input parameters, and the
story drift was chosen as the output parameter of ANN. Up to 1080 data sets are used to train
ANNs and 405 datasets for testing. Well-trained ANNs can predict story drift at seismic loads at
95% predicted rate and calculated mean square error (MSE) as low as 1.6.10-4. Fault drift
prediction accuracy of up to 90% or more, can greatly help engineers according to the seismic
load to quickly identify the situation of the building, and planning for building maintenance.
After 5000 iterations of ANN training using the 1080 data set, the prediction rate is as high as
95% and the MSE is 1.6 x 10-4.

New method for simulation earthquake records by using adapted wavelet - G. Ghodrati
Amiri and A. Asadi [19]
The study has been completed for some of Iran's selected records (Ramezi 1997) (Naghan 1977,
Tabas 1978). The peak ground acceleration of these records is 1g. In this section, the proposed
method has been applied to MATLAB software (MATLAB 1999), which uses wavelet packet
basis functions for seismic recording.In this study, a method of applying a wavelet packet
transform to generate an artificial acceleration map from a pseudo-acceleration response
spectrum has been developed. This method shows the optimal tree calculation for a given
entropy, and calculates the optimal wavelet packet to balance the amount of compression and the
retention of energy. By using this method you can optimize the results. The feasibility and
reliability of the proposed method has been verified by different earthquakes in Iran.

A BP Artificial Neural Network Model for Earthquake Magnitude Prediction in


Himalayas, India - S. Narayanakumar, K. Raja [20]
In this paper, they propose a BP artificial neural network, which is a very effective model for
nonlinear modeling, analysis and prediction of seismic magnitude in the Himalayan region. Has
many advantages such as simple learning, memory and self-adaptation, but there are very
complex multi-dimensional curves with multi-local extreme points. The results show that the BP
neural network model provides a higher prediction accuracy, and the BP ANN model is better
than other seismic models proposed below the predicted earthquake range. This is due to the fact
that BP ANN can capture non-linear relationships compared to statistical methods and other
proposed approaches.
Seismic Time History And Non-Linear Analysis Of Large-Scale Power House Structure -
ZHANG Huidong, WANG Yuanfeng, LI Zheng and WANG Rixuan [21]
In this paper, the seismic performance of large hydropower stations is studied by time history
method, and the influence of material nonlinearity is combined. Comparisons from linear and
non-linear results can conclude that the use of elasticity analysis methods and elastomeric
plastics can yield considerable different results. The elastic response method has a significant
effect on the dynamic response of the structure than the elastic plasticity analysis method. The
method of elastic-plastic analysis can show the location of cracks, the time of occurrence and the
behavior of damage. The time history method can describe the whole process of dynamic
response change with time.From the time history of the earthquake and the analysis of the
nonlinear analysis of the large dynamic space structure,concluded that: (1) the peak time history
can take full account of the peak acceleration value and its duration. Thus, it can describe the
whole process of dynamic response changes over time. (2) According to the comparison of linear
and nonlinear principal stresses, the whole structure can be considered as elastic phase. Although
the dynamic response is small, but from the dynamic point of view, the structure under the cyclic
load fatigue failure should be more attention. (3) From the linear and nonlinear analysis, it can be
seen that the structural nonlinear analysis is a more accurate description of the actual state of the
structure under load.

New development of artificial record generation by wavelet theory - G. Ghodrati Amiri , P.


Ashtari and H. Rahami [22]
The main purpose of this paper is to propose a new method based on wavelet theory to generate
more artificial seismic records consistent with the target spectrum. The wavelet can break the
time series into several levels, each covering a specific frequency range. If the acceleration map
is transformed into a frequency domain by Fourier transform, the wavelet is considered to be a
transform in the time domain scale domain, which has been changed to a scale in the nearest
field. Since the wavelet theory separates each signal, it is possible to generate many manual
records with the same target spectrum. This paper presents a simple and practical method for
generating many realistic manuscripts with a given spectrum. In addition, the method uses a
more reliable real record than other random methods. The feasibility and reliability of the
proposed method has been validated with different Iranian examples.
Artificial Neural Network Modeling for Dynamic Analysis of a Dam Reservoir Foundation
System - Pr. Dr. Rafa H. S. Al-Suhaili, Pr. Dr. Ahmed A. M. Ali, Shamil A. K. Behaya
[23]
In this study, the artificial neural network (ANN) model was established and validated to quickly
estimate the maximum stress, strain and hydrodynamic pressures generated by seismic excitation
in the gravity dam. The developed model can accurately estimate these values. There is no
explicit equation that associates the input with the output variable. Considering the interaction of
the three media, the dam (concrete), the reservoir (water) and the foundation (soil) need to solve
a system of simultaneous local differential equations. In this study, 900 different databases used
ANSYS software to build case input and output. Each of these input variables is assigned a range
from which the values are selected. These scopes are based on the recommendations of the
authorization source associated with this issue. After converting the input and output variables
into dimensionless form, the artificial neural network ANN model is constructed using the
statistical software SPSS of the above database. The results show that the model can predict the
output value (stress and hydrodynamic pressure) with high accuracy. The correlation between the
observed output value and the predicted value model is between 97.8% and 99.7%. The
MATLAB programming language is used to write a program to apply an artificial neural
network model to obtain any set of input stress and hydrodynamic pressure variables rather than
using a longer ANSYS modeling process. In order to further examine the performance of the
model, it is applied to three different data cases that do not exist in the database used to build the
model. The comparison between the three cases obtained by the artificial neural network model
and the three cases obtained by using ANSYS software shows that the model can predict the
output with high precision. The correlation coefficients of these three cases are only horizontal
acceleration, vertical acceleration and double acceleration (horizontal and vertical), respectively
99.6%, 99.9% and 99.8%.

New Method for Generation of Artificial Earthquake Record - G. Ghodrati Amiri, A.


Bagheri [24]
In this paper, a new method for generating artificial acceleration graphs from pseudo-velocity
response spectra using wavelet transform and neural network is developed. The first DWT is
used to decompose the seismic acceleration map into several levels, each covering a specific
frequency range, and then for each level, the RBF neural network is trained to learn to correlate
the response spectrum with the wavelet coefficients. Finally, the generated acceleration graph
using inverse discrete wavelet transform is obtained. In an illustrative example, the proposed
method is applied to a sample of 106 recorded seismic acceleration maps. In the training of
trained neural networks, a pseudo-velocity response spectrum is given as input, and the resulting
neural network generates an acceleration map that is very similar to its training set; a pseudo-
velocity response spectrum with a nearly input, or Synthesize a new and realistic acceleration
map. The proposed method is also tested by using the design spectrum as an input and generating
an acceleration map compatible with these spectra.

Appraisal of the evolutionary-based methodologies in generation of artificial earthquake


time histories - A. Shahjouei, G. Ghodrati Amirib [26]
In this paper, we focus on the application of evolutionary algorithms in AEA generation
methods. In this regard, we have counted two novel methods; based on genetic algorithm and
hybrid evolution neural network method. The main feature of this paper is to provide two kinds
of proposed method of statistical information, to assess the future model and algorithm provides
some quantitative criteria. Based on the spectrum compatibility of the three main functions, the
generated seismograph random diversity and computational work were evaluated. The results
show the practical advantages of evolutionary algorithms in this case.

Artificial Adaptive Systems to predict the magnitude of earthquakes - P.M. BusceMa, G.


Massini and G. Maurelli [27]
The purpose of this experimental study is to test the advanced ANN and machine learning
capabilities to estimate the size of events recorded daily. The description of each event is
characterized by the original time (UTC), latitude, longitude, depth and magnitude. In the
database, an earthquake event means an event between 0.1 and 5.9. We have tested ANN
technology for different datasets: a) USGS data from 1976 to 2002; b) USGS and ISIDe data
from 2005 to 2011; c) ISIDe data for 2005 to 2013. This paper aims to show that ANN is a
promising seismic forecasting technique that is superior to ANN training for local data as an
ANN training for global seismic data for local earthquakes. In fact, the results show that ANN
has very good performance in functional approximation when the training set represents a
sample of the world's earthquakes, ie pattern recognition: 10% of the absolute error of the
amplitude estimate and about 90% of the correct classification Class) in pattern recognition
tasks. The results of using only the ISIDe dataset in Italy are also promising, although the
information available is small, but not as accurate as before: about 99% for the event of M ≤ 2.0,
about 75% of the moderate event (2.0 < M <3.0), the correct classification rate between 30% to
40%, where M ≥ 3.0 events. The last result is not surprising, since the number of fewer events in
the Italian dataset (ISIDe) is small. These results may also be the starting point for the
development of ANN-based systems to provide day-to-day estimates of possible future
earthquake events.

Artificial Neural Network Application for Predicting Seismic Damage Index of Buildings in
Malaysia - Azlan Adnan & Patrick Liq Yee Tiong, Rozaina Ismail, Siti Mariyan
Shamsuddin [29]
In the study, it was found that when all seven initial recognition input parameters were used
together, ANN gave the highest precision, rather than eliminating any one. At the same time, the
ANN system achieves the highest accuracy by using 15 hidden neurons. The system in the test
phase designed 112 building samples to accurately generate 104 output data, the standard
accuracy of 93%. However, the accuracy of the validation phase is only 75%, of which only 39
of the 52 building samples are accurately predicted. From these percentages it can be clearly seen
that the characteristics of the input data fed into the ANN system significantly affect the
prediction results. The distribution pattern and the number of these inputs in the training phase
play a very important role in determining the overall accuracy of the ANN system. This study
focuses on the use of ANN to predict the feasibility of seismic seismic damage and the
investigation of parameters to determine the effect of different input parameters on the impact
prediction accuracy. Therefore, this study does not include the quantification of the uncertainty
of one of the key indicators used in conventional seismic risk assessment. ANN predicts the
seismic damage of the frame structure system directly according to the input training data
analyzed by IDARC-2D.
The simulation of earthquake ground motion for the generation of artificial accelerograms
- J. A. Abdalla & Y. M. Hag-Elhassan [30]
In this paper, we use random process to simulate ground motion acceleration time history. The
ground acceleration is modeled as a Gaussian stabilization process and a non-stationary
acceleration time history is obtained by applying an envelope or shape function to the stationary
process. The power spectral density used for this process is a function of seismic magnitude,
epicentral distance, and field soil conditions. The generated acceleration time history can be used
to assess the vulnerability of existing buildings to earthquakes and the seismic design of new
buildings. Although this study laid the groundwork for simulating the seismic effects of
Khartoum buildings, there are still many areas and topics that need further study and require
more data for more detailed investigation. In addition, the best way to generate an acceleration
map is to record using a widely available strong motion recorder.

Artificial Intelligence Based Techniques For Earthquake Prediction: A REVIEW - Faisal


Azam, Muhammad Sharif, Musarat Yasmin, Sajjad Mohsin [31]
This work covers the recent research and scientific analysis of intelligent technologies applied to
seismic data sets. The survey shows that by applying the results of a single method, the dynamics
and nature of potential earthquake disasters can not be fully predicted. Therefore, there is a need
for multidimensional methods with different characteristics and good quality data to predict
future patterns. In addition, for short-term regional earthquake prediction, conventional methods
and techniques may not be consistent and reliably using the same strategy for future prediction,
because the nature and dynamics of the underground crust are constantly changing.

Selection and Scaling of Real Earthquake Accelerograms to Fit the Turkish Design Spectra
- Yasin M. FAHJAN [32]
One of the key issues of this analysis is to choose to accelerate the time history to meet the
design criteria and the soil type of the particular site. In the literature, there are three sources of
acceleration time: a manual record compatible with the design response spectrum, a synthetic
record obtained from the seismic model, and an acceleration map of the actual seismic record.
Due to the increase in the available terrestrial motion database, the use and scaling of the actual
recorded acceleration map is becoming one of the most recent research topics. In this study, we
discuss and summarize the basic methods and criteria for choosing the history of strong ground
motion. The time-domain scaling procedure is used to scale the available actual records to match
the elastic design spectra presented in the different seismic and local site categories given by the
Turkish seismic code (DBYBHY, 2007). Select and classify the best ground movement time
history, taking into account the magnitude of the earthquake, the source mechanism and the site
conditions.

2.2 Findings from the Literature


The following conclusions can be drawn from the above discussed literatures.

• Seismic action synthesis method by observing seismic motion to meet the target
response spectrum.
• The response to simulated ground motion on each floor is better than the response of the
recorded ground motion.
• Determine the simulated ground motion of the designed elastic spectrum, showing that the
difference in the nonlinear response of each floor is small.
• The resulting simulated ground motion can be used as an input ground motion for non-linear
response analysis of high-rise building structures.
• The peak response frequency of the wavelet packet component, because the iterations point
well match the target spectrum and are stable.
• Artificial neural networks have an acceptable potential to predict the seismic performance of
liquid tanks.
• Linear prediction coefficient analysis is used to reduce the size of the neural network.

2.3 Gap Analysis


By the overview of literature, the areas having lack of research can be summarized as
follows.
• Use ANN and wavelet methods to develop artificial acceleration maps of different frequency
regions.
• Nonlinear time-stepped FE analysis is not considered in previous studies.
2.4 Objectives of the Present Study
The main objectives of the present study are
• To study the behavior of different class structures under artificial earthquakes.
• Comparison with behavior of structures under natural earthquakes.

2.5 Need of the Present Investigation

You might also like