Newton Method of Solving Nonlinear
Newton Method of Solving Nonlinear
Newton Method of Solving Nonlinear
r d
3 INTERNATIONALCONFERENCE ON SCIENCE
AND MATHEMATICS EDUCATION
10-12 NOVEMBER 2009
PENANG
SOLVING N O N - L I N E A R SYSTEMS BY N E W T O N ' S M E T H O D USING
SPREADSHEET EXCEL
Abstract
Tay (2006) had proposed solving numerical methods using the Casio fx-570MS calculator to overcome the
tediousness of doing recursive calculations. Here, we present another alternative, that is solving a non-linear system
using Newton's method in Microsoft Office Excel. For this, we just make use of the MULT function to do matrix
multiplication and MINVERSE function to do the matrix inverse operation. The concept presented here can be
developed into a solver where the user just needs to input the initial vector X, the corresponding formula of
Jacobian matrix J(x, y) and non-linear system vector F(x, y) The full solutions will be displayed
automatically. The solver is easy and user friendly for students and educator who needs its full solutions quickly.
Introduction
Many real-life problems are non-linear. A couple of non-linear equations form a system of non-
linear equations. Systems of non-linear equations can be solved numerically by Newton's method, fixed-
point iteration method. Solving a system of non-linear equations by Newton's method is tedious. Tay
(2006) proposed a step by step instruction on how to use Casio fx-570MS in Numerical methods, while
Guerreri-Garcia and Santos-Polomo (2008) presented a collection of keystroke sequences helping in
solving numerical methods. Tay and Kek (2008) developed a solver to solve the most dominant eigenvalue
by the power method through Microsoft Office Excel. On the other hand, Kek and Tay (2008) developed a
solver to solve system of linear equations using Excel too. None of the literature works dealt with a system
of non-linear equations. Hence, in this paper, we have developed a solver for solving system of non-linear
equations by using Newton's method through Excel 2003.
Newton's Method
f(x,y) = 0,
g(x,y) = 0.
we would like to find the values x = x and y = y such that f(x*,y') = 0 and g(x*,y*) = 0.
Assume that an initial approximate solution to equation (1) is known as (x.,y ) . Express t f(x,y)
and g(x,y) in two-variables Taylor series about (x,,.y,), we get
f{xM ,y )M = f(x ,y )
l l + —f(x ,
t y )(x -x )
M i + — f(x, ,y,)(y -y,)
M + ... = 0,
ox dy
(2)
g(x ,y )
M l+l = g(x ,y )l l + — g(x ,y,-x )
l t + — g(x,,y)(yM -y,.) + ... = 0
ox oy
1
f(x ,y )
M M — 0 and g ( x , , y ) = 0 because will be the roots. Truncating Eq. (2) after the first
+] i+l
Writing Eq. r(3) in matrix notation yields the small changes below:
d „.
- /(W,)
.a
—f{x„y,)
ox oy
—g(x„y,) Mu
(4)
-g(x,,y,).
-rgix„y,)
ox oy
J(x„y,)AX =-Y(x„y,), i
l
AX.=-J(x„y )- F(x ,y,), i t
y M =y,+tyi,
fx } V (Ax;
+
or equivalently
X, =X,+AX,..
+1
Procedure
dg(x„y,) oyd—g(x„y,)
ox
a
( fix v S\
Step 2 Obtain F(x,, y,,) =
\g(.x„y,)J
Step 3: Solve Jix„y )AX i i = -F(x„y,) for AX,.
Step 4: Calculate X = X , + AX,. i+1
Numerical Example
\ .
\ \
f
\
/'
/
1
3
\
\
X "
Figure 1: Graph of / ( x , y) and g(x, y)
Solve the system of non-linear equation above by using Newton's method with initial guess (0, -1) and
0,1).
Solution
Stepl
df
Ax,y) =
dh
dx dy
dh
- 2x
-1 2y
dx dy_
2
~x -y - l "
F{x,y) = 2 _
y X
Step 3
By using X . = X + A X where A X = -J{x ,y )~ Y{x ,y )
+1 ( f ; j i
x
, we obtain the numerical solution
j j
Table 1
Initial guess (0, -1)
i X, F(x„y )
J.'x .V.) i AX,
0 0 0 1
-1 iw' v. # * - 2 ' "• 1 0
3
1 1 2 -1 1 -0.4
-1 -1 -2 0 0.2
:
2 0.6 mx& •. p g i • 0.16 -0.074
-0.8 W,6' 0.04 0.0712
3 0.526 1.052055 -1 0.0055 -0.001
-0.729 -1 -1.45753 0.0051 0.0043
4 0.5249 1E-06 6E-06
-0.725 :.JS^lli49' 2E-05 8E-06
5 0.5249 1.049777 -1 4E-11 2E-12
-0.724 -1 -1.44898 7E-11 4E-11
6 0.5249 $6497p. 0 0
-0.724 -1.44898 0 0
Since | X -X 1=
6 5 0.0000 < £ = 0.0005 , so one of the root is (0.5249, -0.7240).
Table 2
Initial guess (1, 1)
i X, J(XiVi) F(X|,Yi) AX,
0 1 v. :
-1.. -1 0.6667
1 • . 2 H- 0 0.3333
1 1.6667 3.333333 -1 0.4444 -0.164
1.3333 -1 2.666667 0.1111 -0.103
2 1.5023 3 004695 0.027 -0.012
1.23 2 460094 0.0107 -0.009
3 1.4903 2.980573 -1 0.0001 -7E-05
1.2208 -1 2.441616 9E-05 -6E-05
4 1.4902 2 980432'' "r - 1 5E-09 -3E-09
1.2207 2.441488 4E-09 -3E-09
Since | X -X
4 3 \= 0.0001 < £ = 0.0005, so the other root is (1.4902, 1.2207).
Steps In Using The Solver For Solving The System Of Non-Linear Equation
4
n Microsoft Excel - COSMED09 02
Efe Edit \flew Insert Format Tools Qgta Window Help
Arial 10 / .D I i
- Ufa
A B C D E F H
1 Example ~Bf_ $
2
3
/(„)=^-,-i=o J U y ) m dx c 2x
4 Bh B -1 2v.
g(xzy) = y2 - x = 0
5 dx c
6
7 i x f J(x..yi) AX,
8 0 0 0 1
9 -1 ' 1 0
10 1 1 2 -1 1 -0.4
r
11 -1 ' -2 ' 0 0.2
12 2 0.6 -1 -\ «- 0.16 -0.074
13 -0.8 ' 0.04 0.0712
14 3 0.526 1.052055 -1 0.0055 -0.001
15 -0.729 ' -1 '0.0051 0.0043
16 4 0.5249 1.(89754 r• -1.45753
M 1E-06 6E-06
17 -0.725 " -1.449 '2E-05 8E-06
18 5 0.5249 1.049777 -1 4E-11 2E-12
19 -0.724 ' -1 '-1.44898 '7E-11 4E-11
20 6 0.5249 1.0497?7 -1 ' 0 0
21 -0.724 -1.44898 ' 0 0
22 7 0.5249 1.049777 -1 0 0
23 -0.724 ' -1 '-1.44898 ' 0 0
24 8 0.5249 1.W9777 -1 '. 0 0
25 -0.724 r
i-J,448S8 ' 0 0
26 9 0.5249 1.049777 -1 0 0
27 -0.724 ' -1 ' 0 0
28 10 0.5249 1.049777. r-1.44898
-1 0 0
29 -0.724
:
% £1.44898 0 0
Step 1: We select cells F8:F9 and type in = -MMULT(MINVERSE(C8:D9), E8:E9), then press SHIFT,
CRTL and ENTER at the same time to calculate AX . 0
Step 2: We select cells B10:B11 and type in = B8:B9+F8:F9 and then press SHIFT, CRTL and ENTER
at the same time to calculate X.
x
5
Advantages and Shortcomes of The Solver
This solver helps educators and students who need the full numerical solutions of a system of non-
linear equations by inputting the initial vector X , Jacobian matrix J(x ,y )
0 0 0 and initial vector F(x ,y ).
0 0
Even though it is designed for a system of two non-linear equations, the EXCEL command involved is not
that difficult if compared to the MATHCAD, MAPLE, MATLAB, or C Programming. Hence users can
extend the concept presented here to a system of n non-linear equations. One of the limitations here is the
solver is not yet fully automated so that the users need to control the number of iterations themselves
manually by dragging to increase the number of iterations or select only the desired number of iterations
from the given ten iterations. However, this manual option is suitable for the purpose of teaching and
learning in the class room and evaluation in the examination hall, where students need to know where to
stop the iterations.
Conclusion
A solver is developed to get the full numerical solutions of a system of two non-linear equations by
Newton's method through Excel. This solver helps educators and students who need full numerical
solutions of a system of two non-linear equations by just inputting the initial vector X , Jacobian matrix
0
J(x ,y )
0 0 and initial vector F(x ,y ),
0 0 but the solver is not yet fully automated where user needs to
control the number of iterations themselves manually. So, we intend to study the Microsoft Excel VBA
language to overcome this shortage in future.
Acknowledgement(s)
We wish to acknowledge registrar of Universiti Tun Hussein Onn Malaysia for supporting the budget
rd
of the presentation of this paper in 3 CoSMED 2009 in Penang.
References
Guerrero-Garcia, P. and Santos-Palomo, A. (2008): Squeezing the Most Out of the Casio fx-570ms for
Electrical/Electronics Engineers. Proceedings of the International Conference on Computational and
Mathematical Methods in Science
Kek, S. L. & Tay, K. G (2008). Solver For System of Linear Equations. Prosiding Seminar Kebangsaan
Aplikasi Sains dan Matematik 2008 (SKASM2008). Batu Pahat: Penerbit UTHM. Pg 605-615.
Tay, K.G. (2006). How To Use Calculator Casio FX-570MS In Numerical Methods, Batu Pahat: Penerbit
KUiTTHO.
Tay, K. G & Kek, S. L. (2008). Approximating The Dominant Eigenvalue Using Power Method Through
Spreedsheet Excel. Prosiding Seminar Kebangsaan Aplikasi Sains dan Matematik 2008 (SKASM
2008). Batu Pahat: Penerbit UTHM. Pg 599-604.
Copyright Statement
Copyright © 2009 <Tay Kim Gaik>, <Kek Sie Long>& <Rosmila Abdul Kahar>. The authors grant a non-
rd
exclusive license to the organisers of the 3 CosMED International Conference 2009, SEAMEO RECSAM
to publish this document in the Conference Book. Any other usage is prohibited without the consent or
permission of the author(s).
6