Soft Computing
https://doi.org/10.1007/s00500-018-3529-7
METHODOLOGIES AND APPLICATION
Fuzzy linear systems via boundary value problem
Hewayda M. S. Lotfy
1
· Azza A. Taha
1
· I. K. Youssef
1
© Springer-Verlag GmbH Germany, part of Springer Nature 2018
Abstract
Reduction in storage and number of operations are considered through avoiding the representation of zeros in storage as well
as in the calculations. The importance of this approach has its effect in large problems that appear in numerical treatments
of boundary value problems in general and becomes more effective when fuzzy concepts are considered. We introduce
an extended embedding solution model named fuzzy compact storage Gauss–Seidel (FCGS) for solving linear systems of
equations with a fuzzy-based right-hand side. The model starts by applying the embedding approach to the n × n fuzzy
linear system, a compact storage technique is then applied to the resultant 2n × 2n de-fuzzification matrix, and finally, a
Gauss–Seidel method is applied to the system. The FCGS experimental results and algorithm are clarified on some numerical
examples including a fuzzy boundary value problem (FBVP). The error improvements through Gauss–Seidel iterations of
fuzzy solution computations are reported. The fuzzy solutions at α-cuts are shown and compared to the exact solutions. FCGS
achieved a reduction of at least 50% of storage by using the compact storage concepts and consequently obtain a reduction
in the mathematical operations and accordingly the running time especially in FBVP applications.
Keywords Compact storage · Iterative methods · Fuzzy system of linear equations
1 Introduction
Systems of linear equations play important roles in sev-
eral domains such as in the commercial, scientifical, and
engineering domains. Usually, problems in these domains
get pruned and reduced into linear systems of equations.
When the problem structure becomes imprecise, the linear
system of equations becomes no more crisp. Vague values
of the system parameters can be demonstrated using prob-
ability distributions, intervals, or fuzzy values. Fuzziness
deals with imprecise, ambiguous, and unclear inputs that is
why it is extremely important to create mathematical models
and numerical methods for fuzzy linear system of equations
(FLS).
Communicated by V. Loia.
B Azza A. Taha
Azzataha@sci.asu.edu.eg
Hewayda M. S. Lotfy
hewayda_lotfy@sci.asu.edu.eg
I. K. Youssef
kaoud22@sci.asu.edu.eg
1
Department of Mathematics, Faculty of Science, Ain Shams
University, Abbassia, Cairo 11566, Egypt
A general model for solving an arbitrary n × n FLS whose
coefficients matrix are crisp and its right-hand side column
is an arbitrary fuzzy number vector is given (Friedman et al.
1998). They proposed an embedding approach where the
original n × n FLS is replaced by 2n × 2n crisp linear
system. Moreover, another embedding approach (Allahvi-
ranloo and Hashemi 2014) replaces the original n × n FLS
by two n × n crisp linear systems. In both of the mentioned
embedding techniques, the size of the computational work is
doubled. Solving such crisp large linear systems is a prob-
lem in itself, and therefore, the use of iterative techniques
becomes extremely useful. The iterative numerical solutions
for fuzzy linear system are studied (Allahviranloo 2005; Feng
2008; Yin and Wang 2009). Compact storage is a technique
for storing sparse matrices. The main idea is to store only
the nonzero elements and to be able to perform the neces-
sary matrix operations. Compact storage schemes allocate
contiguous storage in memory for the nonzero elements of
the matrix. This requires also a scheme for knowing the
places of the elements in the original matrix. There are many
alternatives for the compact storage schemes, mentioned for
instance (Saad 2003; Day 1977; Barrett et al. 1994).
In this paper, we propose a general model for solving an
n × n FLS whose coefficients are crisp with a fuzzy right-
123