See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/345909135

A New Modified Version Of Gauss-Seidel Iterative Method Using Grouping

Relaxation Approach

Article  in  International Journal of Scientific & Technology Research · November 2020

CITATIONS READS

0 870

4 authors, including:

Baher A. Haleem Ihab M. El Aghoury

The University of Calgary Ain Shams University
2 PUBLICATIONS   0 CITATIONS    30 PUBLICATIONS   110 CITATIONS   

SEE PROFILE SEE PROFILE

Hisham A El-Arabaty
Ain Shams University
5 PUBLICATIONS   2 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Behaviour of Tapered Beams and Columns View project

“Behavior and Design of Ultra-Large Capacity End-Plate Bolted Connections” View project

All content following this page was uploaded by Baher A. Haleem on 15 November 2020.

The user has requested enhancement of the downloaded file.

 INTERNATIONAL JOURNAL OF SCIENTIFIC & TECHNOLOGY RESEARCH VOLUME 9, ISSUE 10, OCTOBER 2020 ISSN 2277-8616

A New Modified Version Of Gauss-Seidel

Iterative Method Using Grouping Relaxation
Approach
Baher A. Haleem, Ihab M. El Aghoury, Bahaa S. Tork, Hisham A. El-Arabaty

Abstract: Systems of linear equations appear in many areas either directly as in modeling physical situations or indirectly as in the numerical solutions
of other mathematical models. The solution of the linear equations’ system is probably the most important issue in numerical methods like the finite
element method (FEM). Using the finite element method in modeling various structures, with either simple or complicated configuration of elements, in
structural engineering became prevalent many years ago. There are two main types of solvers depending on whether the used method is direct or
iterative (indirect) method. In contrast to the iterative techniques, the direct techniques provide almost exact solutions, however they are not convenient
for some situations, including but not limited to huge systems of equations. In such situations, the iterative solvers are favored as they have privileges
concerning solving speed and storage requirements. In addition, indirect solvers are simpler to program. This research focuses on using the Classical
(Stationary) iterative techniques for solving linear systems of equations. The main objective of this research is to develop a new modified version of the
well-known Gauss-Seidel (GS) iterative technique which is adapted to solving problems in structural engineering. The proposed technique remarkably
outperforms GS technique regarding the required number of iterations and the convergence speed. In this paper, the differences between the direct and
iterative approaches have been discussed, along with a quick overview of some of the methods underlying these two classes. Then, the idea and
algorithm of the new proposed ―Modified Gauss-Seidel‖ (MGS) technique have been elucidated. Afterward, the algorithm has been programmed and
used to solve some 2D Practical Examples, besides using the conventional Jacobi and GS techniques. Finally, the obtained results have been compared
to assess the proposed MGS; it outperformed both Jacobi and GS.

Index Terms: Systems of linear equations, Finite element method, Direct techniques, Indirect techniques, Stationary iterative methods, Jacobi, Gauss-
Seidel.
——————————  ——————————
1 INTRODUCTION
Systems of linear equations appear in many areas either problems in many engineering departments, as previously
directly as in modeling physical situations or indirectly as in the mentioned, however this research focuses on structural
numerical solutions of other mathematical models. These analysis problems in civil engineering only. Using the finite
applications arise in various fields, including but not limited to element method in modeling various structures, with either
engineering, biological, physical, social science, etc [1]. The simple or complicated configuration of elements, in structural
solution of the linear equations’ system is probably the engineering became prevalent many years ago. In this paper,
most important issue in numerical methods. The a new modified approach of Gauss-Seidel iterative technique
simplest models of the physical field are linear, hence is introduced for solving the systems of linear equations
linear equations appear frequently in physical problems. resulting from applying the stiffness method to structural
Not only that, but even the most complicated cases are models using the finite element analysis. Generally, the used
frequently approximated through using a linear model as techniques for solving systems of linear equations are mainly
an initial step. Moreover, the solution of a nonlinear classified into Direct techniques and Iterative techniques.
equations’ system is achieved by an iterative process involving
the solution of a sequence of linear systems, where each 1.1 Direct methods
of them approximates the nonlinear equations. Likewise, The direct methods get the exact solution within a finite
the solution of differential along with integral equations by amount of computation, so long as there is not any roundoff
discretizing the system using normally the finite difference error [2]. But in reality, the exact solution is obtained where
technique leads to a system of nonlinear or linear there is no error apart from the round off error [1]. Among the
equations [2]. The linear system problem can be expressed in direct methods are Gaussian elimination, Gauss-Jordan
matrix form such that [A]{x}={b}, where [A] is an n×n elimination, Cholesky decomposition, LU decomposition, etc.
nonsingular matrix so-called coefficient matrix, {b} is an n-
vector of known constants, and the problem is to find the 1.1.1 Gaussian elimination method
values in the n-vector {x} which is the vector of the unknowns The Gauss elimination method is one of the oldest direct
[1]. This linear system usually appears in numerical analysis methods where the process starts by the so-called ―forward
elimination‖ where some row operations are carried out in
———————————————— order to reach a triangular form of equations in the coefficient
 Baher A. Haleem, Teaching Assistant, Department of Structural
Engineering, Faculty of Engineering, Ain Shams University, Cairo,
matrix leaving zeros in place of the off-diagonal entries below
Egypt. E-mail: [email protected] the main diagonal. Afterward, this is followed by the process
 Ihab M. El Aghoury, Associate Professor, Department of Structural known as ―back-substitution‖ in which the unknown of the last
Engineering, Faculty of Engineering, Ain Shams University, Cairo, row (Xn), which was obtained directly from the last equation as
Egypt. E-mail: [email protected] it has not any other unknown, is then substituted back in the
 Bahaa S. Tork, Professor, Department of Structural Engineering, previous equation to get the value of the previous unknown
Faculty of Engineering, Ain Shams University, Cairo, Egypt. E-mail:
[email protected]
(Xn-1), and so on until all the values of all unknowns are
 Hisham A. El-Arabaty, Professor, Department of Structural attained eventually [3].
Engineering, Faculty of Engineering, Ain Shams University, Cairo,
Egypt. E-mail: [email protected]

141
IJSTR©2020
www.ijstr.org
 INTERNATIONAL JOURNAL OF SCIENTIFIC & TECHNOLOGY RESEARCH VOLUME 9, ISSUE 10, OCTOBER 2020 ISSN 2277-8616

1.1.2 Gauss-Jordan Elimination method unknowns; a, b, and x refer to the terms of the coefficient
In the Gauss-Jordan Elimination method, only the ―forward matrix, vector of known constants, and vector of unknowns
elimination‖ process is applied, since all the off-diagonal respectively.
elements, below and above the main diagonal, are eliminated 1 n
at the same time. Hence, the ―back-substitution‖ process is no xi k 1  (bi   aij x j k ),
longer needed [3]. aii j 1 (1)
j i
1.2 General merits and demerits of both classes of i  1, 2,...., n
methods
Direct techniques have some difficulties. For instance, the Nevertheless, the Jacobi technique is seldom used in practice
problem of the method of Gaussian elimination lies in the as the Gauss-Seidel technique is remarkably faster and is
accumulation of rounding errors. This has stimulated many implemented easier on a sequential computer [2].
authors to scrutinize the solutions of systems of linear
equations by both direct and indirect techniques [1]. Direct 1.3.2 Gauss-Seidel (GS) iterative method
techniques are almost habitually used for solving The Gauss-Seidel (GS) technique was devised by both Carl
equations of filled matrices, whereas iterative techniques Friedrich Gauss (1777–1855) and Philipp L. Seidel (1821–
are sometimes favored for sparse matrices. It may not 1896). GS is the modification of the Jacobi iterative technique.
be possible to exploit sparsity of sparse matrices while The difference between both the Jacobi and GS
using direct techniques because the elimination process techniques is that in the GS method each component of
can make the zero entries nonzero, which does not x is used immediately in the iteration as soon as it is
happen in the iterative techniques. Hence, the storage computed. Accordingly, Gauss-Seidel technique is
requirements and number of arithmetic operations may occasionally named the method of successive
be the same for filled and sparse matrices which may be displacements. GS method is more appropriate for
prohibitive for significantly large matrices. Thus, it may programming as the successive approximations do not
be worthy to resort to the iterative techniques in those need to be stored in two separate arrays; the new values
cases of large systems as they are less time - can be overwritten immediately on the old values [2];
consuming and require less storage on a computer. which in turn requires less storage on a computer and
Furthermore, some matrices, usually resulting from that will be significantly effective in case of huge
solution of partial differential equations, may have large systems, for instance when solving for millions of
bandwidth besides a large fraction of zero entries within unknowns [4]. However, it should be noted that the GS
the band. The iterative techniques may be considered iteration will not be really useful on a parallel computer at
also for such matrices [2]. With significant advances in which all the computation of a Jacobi iteration can be
methods for treatment of sparse matrices, the usage of direct performed simultaneously whilst for Gauss-Seidel any
techniques is increasing, even in more general cases. unknown cannot be computed until the preceding unknown is
However, the situation may be changed in favor of iterative calculated; it is a sequential process [3].
techniques when it comes to machines that are capable of Iterative formula of the Gauss-Seidel procedure is given by
parallel processing since the iterative algorithm can be (2):
i 1 n
1
(bi   aij x j k 1   aij x j k ),
effortlessly adapted for parallel processing as the equations
can be solved independently [2]. xi k 1 
aii j 1 j i 1 (2)
1.3 Iterative methods i  1, 2,...., n
The iterative methods provide approximate solutions where A comparison between Jacobi technique and GS technique
there is some error within the desired tolerance value. was presented by A. I. Bakari and I. A. Dahiru [5]. Their results
Basically, they give a sequence of approximations that showed that GS outperforms Jacobi technique as far as the
converges to the final solution [1]. The iterative methods are accuracy and the number of iterations till convergence are
divided into two main categories of methods: Classical concerned; GS is faster than Jacobi. The convergence
(Stationary) methods and Modern methods. In this research, conditions of Jacobi technique and GS technique were shown
only two Classical methods—Jacobi and Gauss-Seidel by Liu Hongxia and Feng Tianxiang [6]. Davod Khojasteh
methods—upon which the newly developed iterative technique Salkuyeh [7] introduced a generalization of Jacobi technique
is based, will be discussed. and Gauss-Seidel technique and studied their convergence
properties. The numerical results of his research proved that
1.3.1 Jacobi iterative method his new generalized procedures have been more efficient than
The Jacobi iterative technique was devised by a German conventional Jacobi technique and Gauss-Seidel technique. It
mathematician whose name is Carl Gustav Jacob Jacobi has been shown that if the coefficient matrix [A] is irreducibly
(1804–1851). The Jacobi iterative technique is also known as diagonally dominant or if it is strictly diagonally dominant
the method of simultaneous displacements; this is for the fact (SDD), the associated iterations of Jacobi technique and
that the value of every entry of the solution (deformations) Gauss-Seidel technique converge for any initial guess [8].
vector is changed, however before any of the new values of Also, if the coefficient matrix is symmetric positive definite
the entries are used in that iteration. Thus, both vectors of {xi} (SPD), the Gauss-Seidel technique also converges for any
and {xi-1} need to be stored separately [2]. Iterative formula of initial guess xo [9]. Moreover, the matrix is considered strictly
the Jacobi procedure is given by (1); where k is the iteration’s diagonally dominant providing the condition presented by (3) is
number; i and j are the row’s order and column’s order valid [10]:
respectively in the coefficient matrix; n is the total number of
142
IJSTR©2020
www.ijstr.org
 INTERNATIONAL JOURNAL OF SCIENTIFIC & TECHNOLOGY RESEARCH VOLUME 9, ISSUE 10, OCTOBER 2020 ISSN 2277-8616

a
j i
ij  aii (3)

2 DEVELOPMENT OF THE MODIFIED GAUSS-SEIDEL

(MGS) TECHNIQUE
The Modified Gauss-Seidel (MGS) method is an iterative
technique adapted for solving structural problems. MGS is
based on the well-known Gauss-Seidel method which is
considered as an iterative method among the Classical
(Stationary) techniques. What makes the MGS method
different from the GS method is the addition of a certain
relaxation mechanism that boosts the performance of this new
method and makes it remarkably better than GS method. In
fact, the MGS method includes some matrix inversion
processes upon which the relaxation mechanism depends.
However, the inverted matrices are significantly smaller than
the global stiffness matrix of the entire structure. Thus, the
disadvantages of the direct solvers, represented in matrix
inversion process or matrix decomposition, are greatly
alleviated, in addition to exploiting the merits of the iterative
methods.

2.1 Beam theories Fig. 2. A 2D frame of a single bay and three stories
There are two main beam theories which are Euler-Bernoulli
beam theory and Timoshenko beam theory. Timoshenko beam As previously mentioned, the MGS consists of two different
theory accounts for shear deformation effects, as such it gives iteration cycles that are repeated mutually in a successive
more accurate results. So, Timoshenko beam theory is manner till convergence. In this research, the developed
adopted in this research. technique is applied to solve the system of linear equations
resulting from the finite element analysis in matrix form.
2.2 2D Sign convention Accordingly, the MGS technique will be illustrated in the
The physical significance of the signs in 2D that is used in this subsequent sections in matrix form to be easily used by
research for the loads (forces and moments) and deformations computers as a program for solving those systems of
(displacements and rotations) are shown in Fig. 1. equations.

2.3.1 The standard iteration cycle (Cycle of uncoupled

DOFs)
In this iteration cycle, the concept of the conventional GS
method is applied. If this cycle were to follow the Jacobi
iterative procedure, then the vector of the residual loads would
be updated once at the beginning of each iteration cycle
through this calculation: {R}18x1 = {F0}18x1 – [K]18x18 * {Δ}18x1,
where {R} is the vector of the residual nodal loads, {F0} is the
vector of the initially applied nodal loads, [K] is the global
stiffness matrix of the entire structure, {Δ} is the vector of nodal
deformations. Afterward, the residual load value of each DOF
would be divided by the corresponding diagonal term in the
Fig. 1. The considered positive directions in 2D stiffness matrix to get the change in the value of the
deformation of this DOF in this iteration. Then, these changes
2.3 The algorithm of MGS in the deformation values are added to the previous
The algorithm of MGS is divided into two iteration cycles that deformation values to get the new values. The idealized
are distinct from each other. They are repeated one after stiffness matrix shown in Fig. 3 describes this relaxation
another until convergence is realized. The first iteration cycle procedure where only the diagonal terms are involved, whilst
follows the same concept of Gauss-Seidel iteration cycle. the other stiffness terms are assumed to be zeros temporarily
However, the other cycle is the innovative one where the in this step. Nevertheless, this iteration cycle follows the
developed relaxation mechanism is applied to certain degrees Gauss-Seidel concept, hence the value of the residual load
of freedom (DOFs), while the other DOFs are relaxed following (whether force or moment) of each DOF is updated
the idea of the conventional GS method. The Example shown individually just before the calculation of the corresponding
in Fig. 2 will be adopted to illustrate the idea of MGS deformation using this residual value. All DOFs are considered
technique. This Example is a 2D frame of single bay and three in this manner in sequence, then this iteration is finished.
stories with fixation supports. The nodes are numbered as
shown in Fig. 2. This 2D Example has 18 DOFs, where 6
nodes * 3 DOFs per node = 18 DOFs.

143
IJSTR©2020
www.ijstr.org
 INTERNATIONAL JOURNAL OF SCIENTIFIC & TECHNOLOGY RESEARCH VOLUME 9, ISSUE 10, OCTOBER 2020 ISSN 2277-8616

bandwidth [2], which is the case in those submatrices. Fig. 4

shows the relaxation mechanism of lateral displacements only
in this modified iteration cycle for the 1st, 2nd, and 3rd floors.
Besides, the extracted submatrices of the three floors are also
provided in Fig. 4.

Fig. 4. Grouping of lateral displacements

The other DOFs, nodal vertical displacements (δY) and nodal

rotation about Z-axis (ϴZ), are relaxed individually as in the
standard iteration cycle.

Fig. 3. Idealized stiffness matrix of the standard iteration 3 ASSESSMENT OF THE PERFORMANCE OF MGS IN
cycle 2D ANALYSIS
In this section, the newly developed Modified Gauss-Seidel
2.3.2 The modified iteration cycle (Cycle of coupled (MGS) iterative technique has been compared to the
DOFs) conventional Jacobi and Gaus-Seidel (GS) iterative methods
In this iteration cycle, a new relaxation mechanism is adopted to appraise the efficiency of the developed procedure. The
as some DOFs are chosen to be coupled (grouped) together Jacobi, GS, and MGS techniques have been programmed for
during the relaxation process, while the other DOFs follow the this research work using the Visual Basic for Applications
same concept of relaxation adopted in the standard iteration (VBA) programming language through the Microsoft Excel
cycle. It has been found through this research that coupling 2016 program on a Dell Inspiron N5110, Core i7 laptop.
the lateral displacements of the nodes (δX) on the same floor
to be relaxed together at the same instant results in faster 3.1 Convergence criterion and tolerances
convergence. This mechanism is achieved through extracting The convergence is attained when all the values of the
an idealized submatrix for each floor from the global stiffness residual loads are within the tolerance range. For instance, a
matrix of the entire structure. Each submatrix is inverted once tolerance value of 0.001 means that all the residual loads’
at the beginning of the analysis process to be multiplied by the values should be reduced gradually with iterations until they lie
subvector of the residual lateral forces of the corresponding within the range of -0.001 to +0.001 for the solver to stop
nodes every modified iteration cycle to get the change in the going into further iterations. It has been found through this
values of these nodal lateral displacements, where each research that a tolerance value of 0.001 is sufficient to obtain
subvector is updated just before this multiplication in each deformation values with almost the same accuracy as that of
modified iteration cycle. Then, these changes in the the results obtained from SAP2000 commercial software.
deformation values are added to the previous deformation
values to get the new values. However, this matrix inversion 3.2 2D Examples for the performance appraisal of MGS
process is not as exhausting as inverting the global stiffness In this section, some Practical Examples of 2D single-story
matrix of the structure since the order of each submatrix and multi-story buildings are solved using Jacobi, Gauss-
equals the number of nodes in this floor. Hence, in this Seidel (GS), and Modified Gauss-Seidel (MGS) iterative
Example there are 3 submatrices each of order 2*2, however techniques. The objective of those Examples is to compare the
the order of the global stiffness matrix is 18*18. Moreover, the results—the number of iterations and elapsed time in
direct techniques are generally preferred to iterative analysis—of these techniques in order to assess the
techniques providing the matrix is banded with small performance of MGS and to show how it greatly outperformed
144
IJSTR©2020
www.ijstr.org
 INTERNATIONAL JOURNAL OF SCIENTIFIC & TECHNOLOGY RESEARCH VOLUME 9, ISSUE 10, OCTOBER 2020 ISSN 2277-8616

the GS method. All the Examples have the same material It can be noticed from Table 1 and Table 2 that although some
properties. The considered material is Reinforced Concrete Examples with a relatively large number of DOFs have a
with Young’s modulus (E) = 22*105 t/m2 and with Poisson’s smaller number of iterations than other Examples with a
ratio (ʋ) = 0.2, whilst the shear modulus (G) depends on both relatively small number of DOFs, the latter Examples have
E and ʋ such that G = E / [2 * (1 + ʋ)]. In any Example, the shorter analysis time. This is because the elapsed time of the
entire frame structure is simulated (modeled) using only beam single iteration increases by increasing the number of
elements (frame elements)—just for simplicity. The cross unknowns. By comparing Example 4 with 6, it is obvious that
sections of the beams and columns are rectangular. The shear Example 6 has a larger number of iterations and analysis time,
area coefficient of these rectangular sections is taken equal to whether using GS or MGS, although both Examples have the
6/5 [11]. The beams and columns are considered massless; all same number of DOFs. This indicates that increasing the
the loads are externally applied to the nodes. The deformation number of floors has a worse effect on the solving speed than
values are initially set to zeros as an initial guess. The beams’ increasing the number of bays.
and columns’ lengths, dimensions of rectangular cross
sections, and applied nodal loads of all the considered
Examples are as follows:
Beams: Span = 10 m, Width = 0.3 m, and Depth = 1 m
Columns: Height = 5 m, Width = 0.3 m, and Depth = 0.9 m
Nodal loads: FXo = 15 tonf, FYo = -70 tonf, and MZo = -100
mtonf
Where the applied nodal nodes are applied to all nodes except
that FXo is applied to the nodes at the left edge only of any
building, as shown in Fig. 5 and Fig. 6. It should be noted that
it has been tried through this research to use the developed
MGS technique however without the inclusion of the standard
iteration cycle; only the modified iteration cycle has been
repeated successively till convergence. This form of MGS has
been given the acronym of MGS* as shown in Table 1 and
Table 2. Table 1 and Table 2 show the number of DOFs of nine
2D Practical Examples, in addition to the results (Number of
iterations and analysis time) using Jacobi, GS, MGS, and
MGS* methods. However, the Jacobi method was used in the
first three Examples only because it is already known from the
previous researches in the literature that Jacobi is slower than
GS method, but it was inserted here in order to just emphasize
this fact (Fig. 7). Fig. 5 and Fig. 6 show the configuration and
number of bays and floors of the nine Examples mentioned in
Table 1 and Table 2. Fig. 7 is a bar chart that shows the
analysis time in seconds using both Jacobi and GS methods
for Examples 1 to 3. Fig. 8 is a bar chart that shows the
analysis time in seconds using both GS and MGS methods for
Examples 1 to 9. Fig. 9 is a bar chart that shows the analysis
time in seconds using both MGS and MGS* methods for
Examples 1 to 9.

Fig. 5. The configuration of the members of Examples 1 to

5
Fig. 6. The configuration of the members of Examples 6 to 9

145
IJSTR©2020
www.ijstr.org
 INTERNATIONAL JOURNAL OF SCIENTIFIC & TECHNOLOGY RESEARCH VOLUME 9, ISSUE 10, OCTOBER 2020 ISSN 2277-8616

It is clear that the MGS method significantly outperformed the

GS method in all the nine Practical Examples (Fig. 8).
Moreover, as the number of DOFs increases—especially by
increasing the number of floors—the MGS becomes much
more powerful than GS technique as MGS takes a
considerably shorter analysis time. It can be noticed from Fig.
9 that the MGS* is remarkably faster than MGS, especially in
case of a large number of DOFs as in Example 9.

Table 1 Number of iterations of the 2D Practical Examples

Practical No. of Number of iterations
Example DOFs Jacobi GS MGS MGS*

Example 1 9 385 197 18 10

Example 2 12 422 208 16 9
Example 3 15 430 214 16 8
Example 4 18 --- 700 72 36
Example 5 24 --- 724 66 33 Fig. 8. The analysis time of GS and MGS for Examples 1 to 9
Example 6 18 --- 1414 200 101
Example 7 27 --- 1541 172 86
Example 8 36 --- 1575 158 80
Example 9 48 --- 2761 288 147

Table 2 Analysis time of the 2D Practical Examples

Practical No. of Analysis time (Sec)
Example DOFs Jacobi GS MGS MGS*

Example 1 9 1.22 0.61 0.16 0.11

Example 2 12 1.66 0.86 0.15 0.11
Example 3 15 2.12 1.11 0.17 0.11
Example 4 18 --- 4.20 0.6 0.39
Example 5 24 --- 5.98 0.73 0.45
Fig. 9. The analysis time of MGS and MGS* for Examples 1 to
9
Example 6 18 --- 8.46 1.61 1.05
Example 7 27 --- 14.56 2 1.19 4 CONCLUSION
To recap, a new modified approach of Gauss-Seidel iterative
Example 8 36 --- 20.95 2.45 1.44
technique has been introduced in this paper for solving the
Example 9 48 --- 53.20 6.27 3.6 systems of linear equations resulting from applying the
stiffness method for structural models using the finite element
analysis. The algorithm of the new proposed ―Modified Gauss-
Seidel‖ (MGS) technique has been elaborated. Then, to
assess the performance and efficiency of MGS compared to
other Stationary techniques, some 2D Practical Examples
have been solved using Jacobi, Gauss-Seidel (GS), and MGS.
The results of those Examples led to the following conclusions:
 The jacobi iterative method is not comparable to the
Gauss-Seidel (GS) method or the developed Modified
Gauss-Seidel (MGS) method.
 The large number of floors has a worse impact on the
solver’s speed than the large number of floors. The more
the number of floor increases, the slower both GS and
Fig. 7. The analysis time of Jacobi and GS for Examples 1 to 3
MGS solvers become.
 Generally, the developed MGS technique is much more
powerful than the traditional GS procedure.
 Furthermore, the idea of using the modified iteration cycle
only that resulted in MGS* is advantageous as MGS* has
been proved to be significantly faster than MGS especially
146
IJSTR©2020
www.ijstr.org
 INTERNATIONAL JOURNAL OF SCIENTIFIC & TECHNOLOGY RESEARCH VOLUME 9, ISSUE 10, OCTOBER 2020 ISSN 2277-8616

in case of a large number of unknowns.

 At a significantly large number of degrees of freedom
(unknowns), especially due to increasing the number of
floors, the GS becomes no more comparable to the
developed MGS technique.

REFERENCES
[1] A. H. Laskar and S. Behera, ―Refinement of Iterative
Methods for the Solution of System of Linear
Equations Ax=b,‖ IOSR J. Math., vol. 10, no. 3, pp. 70–
73, 2014.
[2] H. M. Antia, Numerical methods for scientists and
engineers. Hindustan Book Agency, 2012.
[3] P. R. Turner and D. Towers, Guide to Numerical
Analysis. Macmillan Education, Limited, 1989.
[4] D. S. Watkins, Fundamentals of matrix computations,
3rd ed. WILEY, 2010.
[5] A. Bakari and I. Dahiru, ―Comparison of Jacobi and
Gauss-Seidel Iterative Methods for the Solution of
Systems of Linear Equations,‖ Asian Res. J. Math., vol.
8, no. 3, pp. 1–7, 2018.
[6] H. Liu and T. Feng, ―Study on the convergence of
solving linear equations by gauss-seidel and jacobi
method,‖ Proc. - 2015 11th Int. Conf. Comput. Intell.
Secur. CIS 2015, pp. 100–103, 2016.
[7] D. K. Salkuyeh, ―Generalized Jacobi and Gauss-Seidel
Methods for Solving Linear System of Equations,‖ vol.
16, no. 2, pp. 164–170, 2007.
[8] S. Y., Iterative Methods for Sparse Linear Systems.
New York: PWS Press, 1995.
[9] B. N. Datta, Numerical Linear Algebra and
Applications. Brooks/Cole Publishing Company, 1995.
[10] S. M. H. Mirfallah and M. Bozorgnasab, ―A New
Jacobi-based Iterative Method for the Classical
Analysis of Structures,‖ Lat. Am. J. Solids Struct., vol.
12, pp. 2581–2617, 2015.
[11] H. P. Gavin, ―Structural Element Stiffness , Mass , and
Damping Matrices.‖ CEE 541. Structural Dynamics-
Department of Civil and Environmental Engineering-
Duke University, pp. 1–35, 2018.

147
IJSTR©2020
View publication stats
www.ijstr.org