This article was downloaded by: [McGill University Library]

On: 06 February 2015, At: 12:06

Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,
37-41 Mortimer Street, London W1T 3JH, UK

International Journal of Systems Science

Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/tsys20

New integral operational matrix in block-pulse series

analysis
a a
WEN-LIANG CHEN & CHING-YEI CHUNG
a
Department of Power Mechanical Engineering , National Tsing Hua University , Hsinchu,
Taiwan, Republic of China
Published online: 31 May 2007.

To cite this article: WEN-LIANG CHEN & CHING-YEI CHUNG (1987) New integral operational matrix in block-pulse series
analysis, International Journal of Systems Science, 18:3, 403-408, DOI: 10.1080/00207728708963975

To link to this article: http://dx.doi.org/10.1080/00207728708963975

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained in the
publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations
or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any
opinions and views expressed in this publication are the opinions and views of the authors, and are not the
views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be
independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses,
actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever
caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content.

This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions
 INT. J. SYSTEMS SCI., 1987, VOL. 18, No.3, 403-408

New integral operational matrix in block-pulse series analysis

WEN-LIANG CHENt and CHING-YEI CHUNGt

The error of the block-pulse series expansion of an integral function employing the
usual integral operational matrix is evaluated and a new integral operational matrix
is obtained from the three-point interpolating polynomial in Lagrange form. It is
shown that this new matrix reduces the error.
Downloaded by [McGill University Library] at 12:06 06 February 2015

1. Introduction
In the block-pulse series analysis of linear systems derived by Sannuti (1977). Rao
and Srinivasan (1978) and Chen and Jeng (1981). one important property used
repeatedly is the evaluation of block-pulse series coefficients of an integral operational
matrix. However, using this matrix, the values of the coefficients will deviate from their
exact values and will introduce error in the block-pulse series analysis of linear
systems.
In this paper. the error due to the use of the usual integral operational matrix is
calculated and a new integral operational matrix obtained from the three-point
interpolating polynomial in Lagrange form is presented. This reduces the error of
block-pulse series coefficient of an integral function and also reduces the error of the
block-pulse series solution in the analysis of linear systems. Two numerical examples
are given at the last section to compare the results obtained from the usual integral
operational matrix and the new operational matrix.

2. Review of integral operational matrices

Let t/Jl(t). t/J2(t)• ..., t/J... (r) be the block-pulse functions and t/J(m)(t)!
[t/Jl(t) t/J2(t) ... t/Jm(t)]' where ' means transpose. It is well. known that an
absolutely integrable function 1(1) defined on [0, T] can be expanded into an m-term
block-pulse series as
1(t);;;;;flt/Jl(t)+ 12t/J2(t) + .. , + f",t/J",(t)=[fl 12 ... 1",]t/J",(t)
! Ft/Je"'l(t) (1)

where

m Iii/mIT
I. = - f(t) dt, i = 1, 2, .... m (2)
T Hi-1 l/m)T

Define

l
F(t) = If(t')dt (3)

Received 24 September 1985.

t Department of Power Mechanical Engineering, National Tsing Hua University, Hsinchu,
Taiwan, Republic of China.
 404 w.-L. Chen and C.- Y. Chung

By the block-pulse series expansion technique in evaluating the integral of a function,

F(l) can be evaluated approximately as

F(l);;- t Fl/J/m)(t')dt' = FPl/J/ml(t) (4)

Where P, an m x m matrix, is called the integral operational matrix and

1
2

T 0
P=-
t (5)
m

0 0 0 1.
2
Downloaded by [McGill University Library] at 12:06 06 February 2015

Substituting (5) into (4) yields

I TTl T
F(l) ~ [ 2;-11 ;;/1 + 2;;/2
~[Fl F2 ••• Fm]l/J(m)(t) (6)

where

(7)

By (2) and (3), F, can be simplified to

Fj = I(i-ll/mlTI(l) dt + ~ Ii~m::l/(l)dt = FC mIT) + ~[F(~T) - Fe mIT)]

= [F(i/m)T + F«i - 1)/m)T] = {(T/m)[F(i/m)T + F«i - 1)/m)T]}m (8)

2

Equation (8) means that r, is the average area of F(t) in the time interval
2

with area obtained by the trapezoidal rule. Now, let the direct block-pulse series
T

e: I T, ~ T]
expansion of F(t) be
(9)

where

- m l(i/mlT
F, = T F(t) dt, i = I, 2, ,m (10)
(i-l/mlT

Equation (10) means that the coefficient F j is the average area of F(t) in the time
i -1 i ]
interval [ ---;;;- T,;;; T .
In the block-pulse series analysis, the coefficients Fi , i = 1,2, ... , m, in (8) are used
r,
to represent the exact coefficients in (10). It is thus easy to see that the error is due to
the calculation of area inaccurately. Hence, in the following section, a new integral
operational matrix is derived and through using this matrix, the error is reduced.
 A new matrix in block-pulse series analysis 405

3. A new integral operational matrix

2
Consider the three points Fe: T), Fe: 1 T) and F(~ T). The polynomial of
degree ~2 in Lagrange form, which passes through these points is

-( ) = F(i - 2
Ft m
r) {t - «i- l)/m)T}(t - (i/m)T)
2(T/m)2

_ F(i - 1 T) (t - (i/m)T){ t - «i - 2)/m)T}

m (T/m)2

+ F(i T) {t - «(i - l)/m)T}{ t - «i - 2)/m)T} (II)

m 2(T/m)2

[i- ~ T),
Downloaded by [McGill University Library] at 12:06 06 February 2015

If F(t) is approximated by (11), that is, by F(t) over the time interval 1 T,
m m
then

.: I(i/m)T
F(t)dt ~ F(t)dt
(i-l/m)T (i-limIT

=!.[_~F(i-2T)
m 12 m
+ ~F(~T)
12 m
+ 2.F(~T)J
12 m
(12)

Dividing (12) by T'[m and using (10), we have

Fi~-/2FC:2T)+ 12 F
8
C: 1T
) + :2F(~T)£Fi (13)

Because F(t) is the integral of /(t),

i-
F( - T =
2) J
O 2/
- m )T T[m i(l/m)T
/(t)dt = - -T f(t)dt +-
m i(2 /m l T
/(t)dt + '0'
mom 0 T (1/m)T

+-
m i(i- 2/mlT
f(t)dt
]
(14)
T (i-3/m)T

Using (2), we then get for (14)

Fe:
2
T) = ~(l1 + /2 + ... + ];-2) (15)

Similarly

i- l ) T
F ( ----;;-T = ;(11 + /2 + '" + ];-d (16)

and

F -T T
i ) = -(11 + /2 + ... +];) (17)
(m m .

Substituting (15)-(17) into (13) yields

(18)
 406 W-L. Chen and c.- Y. Chung

Equation (18) is satisfied for i> 2. For i = 2

F2 ~ - /2 F(0) + 182F(~) + 152F(~ T)

=;;;T(13 5)_
1/1 + I2f2 =F2 (19)

where F(O) = 0, and for i = 1,

F_
I
~ --F
12 m 12
T)
1 ( - - +-F(O)+-F
8 5 -
12 m
(T) (20)

Since F( - Tjm) is undefined, PI cannot be calculated using this method. Here, PI is

estimated by the usual method, that is, by (7) as
Downloaded by [McGill University Library] at 12:06 06 February 2015

_ IT _
F I ~--fl =F I (2] )
2m
Substituting (18), (19) and (21) into (9), we get
F(t) = [PI F2 ••• PmJljI(m)(t)

1 13
2 12
5 13
0 - -
12 12
T
~[fl f2 ... fmJ- 5 13
m 0 0 - -]2
12

5
o 0 0 0
12

(22)
Comparing (4) with (22), P is a new integral operational matrix and is modified
slightly from P.

4. Error analysis
In this section the error of the block-pulse series coefficients of F(t) calculated from
P and j5 are estimated. From (8) and (10), the error between the exact block-pulse
series coefficient Pj and the coefficient F, obtained from P is

Pi - r, = '!!. {r(ifm)T F(t)dt _ !.. [F(ijm)T + F«i - l)jm)T]} (23)

T J(i"l/"l7lT m 2
By Atkinson (1978)

T [F(iJmT) + F«i - l)jm)T] )3 F(f1)

i (l /mlT
(i-l/m)T
F(t)dt--
m 2
1 (]
= - - -T
12 m
/I

(24)
 A new matrix in block-pulse series analysis 407

for some '1 E C: 1 T,'; T J. Hence

Fi - r, = - Ili~Y F"('1) (25)

Next, from Atkinson (1978), the error of F(t) and F(t) in (II) is

F(t) _ F(t) = t(t - (T/m))(t - (2 Tim)) F"/(e) (26)

3!
for some e. Integrating (26) from «i - l)/m)T to (ilm)T, yields
(T)4 r"(e)
I
li/m)T f(i/mlT 1
F(t)dt- F(t)dt=-24 - (27)
(i-l/m)T (i-l/m)T m
Downloaded by [McGill University Library] at 12:06 06 February 2015

Dividing (27) by Tim and using (10) and (12) or (13), we have
_ _
F, - F, = - 24 ;
1 (T)3 r'(~) (28)

Equations (25) and (28) are, respectively, the errors between the exact block-pulse
series coefficient F i and the coefficient F, obtained from the integral operational
matrix P and the coefficient Fi obtained from the new integral operational matrix P. If
the function F(t) is smooth enough, then for large m, the error r, - Fi is less than
Pi - F;, that is, the use of the new integral operational matrix P is better than the use
of the usual operational matrix P_

Example 1
Let f(t) = exp 2t. Then
F(t) =
'I
0
f(t')dt =
1
-(exp (2t) - I)
2 .
The eight-term block-pulse series expansion of f(t) and F(t) on t E [0, I) are
f(t) = [1-1361 1·4589 1·8732 2·4051 3·0882 3·9656 5·0916 6'5378Jt/t(Sl(t)
(29)
and
F(t) = [0-0681 0·2294 0'4364 0-7024 1·0439 1·4836 2-0460 2'7685Jt/t(s)(t)
(30)
The block-pulse series expansion of F(t) obtained by the integral operational matrix P,
that is, by (4) is
F(t) ~ [0-0710 0·2332 0·4415 0-7089 1·0522 1·4931 2·0592 2'7860Jt/t(s)(t)
(31)
and the block-pulse series expansion of F(t) obtained by the new integral operational
matrix P, that is, by (22) is
F(t) ~ [0,0710 0·2298 0-4372 0·7034 1·0451 1·4839 2·0474 2'7709]t/t(s)(t)
(32)
 408 A new matrix in block-pulse series analysis

Comparing (30), (31) and (32), shows that using the new integral operational matrix P
gives better results than using P.

Example 2
Consider the simple differential equation
f(t) - f(t) = 0, f(O) = 1 (33)
The solution is f(t) = exp t. The to-term block-pulse series of f(t) can be calculated as
f(t)= [1'0517 }'1623 1·2846 1·4197 1·5690 1·7340 /·9163
2'1179 2·3406 2'5868]"'(lo)(t) (34)
By block-pulse series analysis, it is well known that (33) can be changed into algebraic
Downloaded by [McGill University Library] at 12:06 06 February 2015

form as
(35)
where F, is the ith coefficient of the block-pulse series solution of f(t) and P is the
integral operational matrix. If P in (5) is used, the block-pulse series solution is
f(t) ~ []'0526 1·]634 ],2859 ]·4213 1'5709 1·7362 ]·9190
2·1210 2·3442 2'59IOJ"'(l0)(t) (36)
If the new integral operational matrix Pin (22) is used, the block-pulse series solution
is
f(t)~[I'0526 1·1625 1·2847 1·4198 1·5692 1·7342 1·9166
2·1182 2·3410 2'5890J"'(lo)(t) (37)
Comparing (34), (36) and (37), we can conclude that using the new integral operational
matrix gives a better block-pulse series solution.

5. Conclusions
In this paper, the geometric meaning of using the integral operational matrix to
calculate the block-pulse series coefficients of an integral function is obtained and it is
found that the error of the coefficients due to the use of the integral operational matrix
depends on the accuracy of the area to be calculated. Therefore any numerical method
which can evaluate the area accurately may be introduced to find a new integral
operational matrix and obtain a better result in the block-pulse series analysis. In this
paper, the three-point interpolating polynomial in Lagrange form is used.

REFERENCES
ATKINSON, K. E., 1978, An Introduction to Numerical Analysis (New York: Wiley).
CHEN, W. L., and JENG, B. S., 1981, Int. J. Systems Sci., 12,625.
RAO, G. P., and SRINIVASAN, T., 1978, Proc. lnstn elect. Engrs, 125, 1064.
SANNUTI, P., J 977, Proc. lnstn elect. Engrs, 124, 569.