Professional Documents
Culture Documents
Runge-Kutta Type Methods For Directly Solving Spec
Runge-Kutta Type Methods For Directly Solving Spec
Runge-Kutta Type Methods For Directly Solving Spec
net/publication/281097141
CITATIONS READS
16 246
3 authors:
Norazak Senu
Universiti Putra Malaysia
194 PUBLICATIONS 1,011 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Kasim Hussain on 01 December 2015.
Research Article
Runge-Kutta Type Methods for Directly Solving Special
Fourth-Order Ordinary Differential Equations
Copyright © 2015 Kasim Hussain et al. This is an open access article distributed under the Creative Commons Attribution License,
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
A Runge-Kutta type method for directly solving special fourth-order ordinary differential equations (ODEs) which is denoted by
RKFD method is constructed. The order conditions of RKFD method up to order five are derived; based on the order conditions,
three-stage fourth- and fifth-order Runge-Kutta type methods are constructed. Zero-stability of the RKFD method is proven.
Numerical results obtained are compared with the existing Runge-Kutta methods in the scientific literature after reducing the
problems into a system of first-order ODEs and solving them. Numerical results are presented to illustrate the robustness and
competency of the new methods in terms of accuracy and number of function evaluations.
of the order conditions of RKFD method is presented. In approach depends on the derivation of order conditions for
Section 3, the zero-stability of RKFD method is given. In the Runge-Kutta method proposed in Dormand [14]. The
Section 4, three-stage RKFD methods of order four and order RKFD method in (3)–(6) can be written as follows:
five are constructed. In Section 5, numerical examples are
given to show the effectiveness and competency of the new
RKFD methods as compared with the well known Runge- 𝑦𝑛+1 = 𝑦𝑛 + ℎ𝜓 (𝑥𝑛 , 𝑦𝑛 ) ,
Kutta methods from the scientific literature. Conclusions are
given in Section 6. 𝑦𝑛+1 = 𝑦𝑛 + ℎ𝜓 (𝑥𝑛 , 𝑦𝑛 ) ,
(10)
𝑦𝑛+1 = 𝑦𝑛 + ℎ𝜓 (𝑥𝑛 , 𝑦𝑛 ) ,
2. Derivation of the RKFD Method
𝑦𝑛+1 = 𝑦𝑛 + ℎ𝜓 (𝑥𝑛 , 𝑦𝑛 ) ,
The general form of RKFD method with 𝑠-stage for directly
solving special fourth-order ODEs (1) can be written as
follows: where the increment functions are
2 3 𝑠
ℎ ℎ 4
𝑦𝑛+1 = 𝑦𝑛 + ℎ𝑦𝑛 + 𝑦 + 𝑦 + ℎ ∑𝑏𝑖 𝑘𝑖 , (3)
2 𝑛 6 𝑛 ℎ ℎ2 𝑠
𝑖=1
𝜓 (𝑥𝑛 , 𝑦𝑛 ) = 𝑦𝑛 + 𝑦𝑛 + 𝑦𝑛 + ℎ3 ∑𝑏𝑖 𝑘𝑖 ,
2 6
ℎ2 3 𝑠 𝑖=1
𝑦𝑛+1 = 𝑦𝑛 + ℎ2 𝑦𝑛 + 𝑦 + ℎ ∑𝑏𝑖 𝑘𝑖 , (4)
2 𝑛 ℎ 𝑠
𝑖=1
𝜓 (𝑥𝑛 , 𝑦𝑛 ) = 𝑦𝑛 + 𝑦𝑛 + ℎ2 ∑𝑏𝑖 𝑘𝑖 ,
𝑠 2 𝑖=1
𝑦𝑛+1 = 𝑦𝑛 + ℎ𝑦𝑛 + ℎ2 ∑𝑏𝑖 𝑘𝑖 , (5) 𝑠
(11)
𝑖=1
𝜓 (𝑥𝑛 , 𝑦𝑛 ) = 𝑦𝑛 + ℎ∑𝑏𝑖 𝑘𝑖 ,
𝑠 𝑖=1
𝑦𝑛+1 = 𝑦𝑛 + ℎ∑𝑏𝑖 𝑘𝑖 , (6) 𝑠
𝑖=1
𝜓 (𝑥𝑛 , 𝑦𝑛 ) = ∑𝑏𝑖 𝑘𝑖 ,
where 𝑖=1
𝑘1 = 𝑓 (𝑥𝑛 , 𝑦𝑛 ) , (7)
where 𝑘𝑖 is given in (8).
2 3 The first few elementary differentials for the scalar equa-
ℎ 2 ℎ 3
𝑘𝑖 = 𝑓 (𝑥𝑛 + 𝑐𝑖 ℎ, 𝑦𝑛 + ℎ𝑐𝑖 𝑦𝑛 + 𝑐𝑦 + 𝑐𝑦 tion are
2 𝑖 𝑛 6 𝑖 𝑛
(8)
4
𝑖−1 𝐹1(4) = 𝑦(𝑖V) = 𝑓,
+ ℎ ∑ 𝑎𝑖𝑗 𝑘𝑗 ) . 𝑖 = 2, 3, . . . , 𝑠.
𝑗=1 𝐹1(5) = 𝑓𝑥 + 𝑓𝑦 𝑦 ,
All parameters 𝑏𝑖 , 𝑏𝑖 , 𝑏𝑖 , 𝑏𝑖 , 𝑎𝑖𝑗 , and 𝑐𝑖 of the RKFD method 2
𝐹1(6) = 𝑓𝑥𝑥 + 2𝑓𝑥𝑦 𝑦 + 𝑓𝑦 𝑦 + 𝑓𝑦𝑦 (𝑦 ) , (12)
are used for 𝑖, 𝑗 = 1, 2, . . . , 𝑠 and are supposed to be real. The
RKFD method is an explicit method if 𝑎𝑖𝑗 = 0 for 𝑖 ≤ 𝑗 and 2
is an implicit method if 𝑎𝑖𝑗 ≠ 0 for 𝑖 ≤ 𝑗. The RKFD method 𝐹1(7) = 𝑓𝑥𝑥𝑥 + 3𝑓𝑥𝑥𝑦 𝑦 + 3𝑓𝑥𝑦𝑦 (𝑦 ) + 3𝑓𝑥𝑦 𝑦
can be represented by Butcher tableau as follows: 3
+ 3𝑓𝑦𝑦 𝑦 𝑦 + 𝑓𝑦𝑦𝑦 (𝑦 ) + 𝑓𝑦 𝑦 .
𝑐 𝐴
𝑏𝑇
We assume that the Taylor series increment function is Δ.
𝑏𝑇 (9) The local truncation errors of 𝑦(𝑥), 𝑦 (𝑥), 𝑦 (𝑥), and
𝑦 (𝑥) can be obtained after substituting the exact solution
𝑏𝑇
of (1) into (11) as follows:
𝑏𝑇
𝑠
The Taylor series increment functions of 𝑦(𝑥), 𝑦 (𝑥), 𝑦 (𝑥),
∑𝑏𝑖 𝑘𝑖
and 𝑦 (𝑥) can be expressed as follows: 𝑖=1
1 1 1 1 4 (5)
Δ = 𝑦 + ℎ𝑦 + ℎ2 𝑦 + ℎ3 𝐹1(4) +
ℎ 𝐹1 𝑠 𝑠
2 6 24 120 = ∑𝑏𝑖 𝑓 + ∑𝑏𝑖 𝑐𝑖 (𝑓𝑥 + 𝑓𝑦 𝑦 ) ℎ
𝑖=1 𝑖=1
1 5 (6)
+ ℎ 𝐹1 + 𝑂 (ℎ6 ) , 𝑠
720 1 2
+ ∑𝑏𝑖 𝑐𝑖2 (𝑓𝑥𝑥 + 2𝑓𝑥𝑦 𝑦 + 𝑓𝑦 𝑦 + 𝑓𝑦𝑦 (𝑦 ) ) ℎ2
1 1 1 2 𝑖=1
Δ = 𝑦 + ℎ𝑦 + ℎ2 𝐹1(4) + ℎ3 𝐹1(5)
2 6 24
1 4 (6) 1 5 (7) + 𝑂 (ℎ3 ) ,
+ ℎ 𝐹1 + ℎ 𝐹1 + 𝑂 (ℎ6 ) ,
120 720 𝑠
(14) ∑𝑏𝑖 𝑘𝑖
1 1 1
Δ = 𝑦 + ℎ𝐹1(4) + ℎ2 𝐹1(5) + ℎ3 𝐹1(6) 𝑖=1
2 6 24
𝑠 𝑠
1 4 (7) 1 5 (8) 1 𝑠
+ ℎ 𝐹1 + ℎ 𝐹1 + 𝑂 (ℎ6 ) , = ∑𝑏𝑖 𝐹1(4) + ∑𝑏𝑖 𝑐𝑖 ℎ𝐹1(5) + ∑𝑏𝑖 𝑐𝑖2 ℎ2 𝐹1(6)
120 720 𝑖=1 𝑖=1 2 𝑖=1
1 1 1
Δ = 𝐹1(4) + ℎ𝐹1(5) + ℎ2 𝐹1(6) + ℎ3 𝐹1(7) + 𝑂 (ℎ3 ) ,
2 6 24
1 4 (8) 𝑠
+ ℎ 𝐹1 + 𝑂 (ℎ5 ) . ∑𝑏𝑖 𝑘𝑖
720 𝑖=1
Substituting (12) into (11), the increment functions 𝜓, 𝜓 , 𝜓 , 𝑠 𝑠
and 𝜓 for RKFD method become as follows: = ∑𝑏𝑖 𝑓 + ∑𝑏𝑖 𝑐𝑖 (𝑓𝑥 + 𝑓𝑦 𝑦 ) ℎ
𝑠 𝑖=1 𝑖=1
∑𝑏𝑖 𝑘𝑖
1 𝑠 2
𝑖=1 + ∑𝑏𝑖 𝑐𝑖2 (𝑓𝑥𝑥 + 2𝑓𝑥𝑦 𝑦 + 𝑓𝑦 𝑦 + 𝑓𝑦𝑦 (𝑦 ) ) ℎ2
𝑠 𝑠
2 𝑖=1
= ∑𝑏𝑖 𝑓 + ∑𝑏𝑖 𝑐𝑖 (𝑓𝑥 + 𝑓𝑦 𝑦 ) ℎ
𝑖=1 𝑖=1 + 𝑂 (ℎ3 ) ,
1 𝑠 2 𝑠
+ ∑𝑏𝑖 𝑐𝑖2 (𝑓𝑥𝑥 + 2𝑓𝑥𝑦 𝑦 + 𝑓𝑦 𝑦 + 𝑓𝑦𝑦 (𝑦 ) ) ℎ2 ∑𝑏𝑖 𝑘𝑖
2 𝑖=1
𝑖=1
3
+ 𝑂 (ℎ ) , 𝑠 𝑠
1 𝑠
= ∑𝑏𝑖 𝐹1(4) + ∑𝑏𝑖 𝑐𝑖 ℎ𝐹1(5) + ∑𝑏𝑖 𝑐𝑖2 ℎ2 𝐹1(6)
𝑠 𝑖=1 𝑖=1 2 𝑖=1
∑𝑏𝑖 𝑘𝑖
𝑖=1 + 𝑂 (ℎ3 ) .
𝑠 𝑠
1 𝑠 (15)
= ∑𝑏𝑖 𝐹1(4) + ∑𝑏𝑖 𝑐𝑖 ℎ𝐹1(5) + ∑𝑏𝑖 𝑐𝑖2 ℎ2 𝐹1(6) + 𝑂 (ℎ3 ) ,
𝑖=1 𝑖=1 2 𝑖=1
𝑠
Using (11) and (14), the local truncation errors (13) can be
∑𝑏𝑖 𝑘𝑖 written as follows:
𝑖=1
𝑠
𝑠 𝑠 1 (4) 1
𝜏𝑛+1 = ℎ4 [∑𝑏𝑖 𝑘𝑖 − ( 𝐹 + ℎ𝐹(5) + ⋅ ⋅ ⋅)] ,
= ∑𝑏𝑖 𝑓 + ∑𝑏𝑖 𝑐𝑖
(𝑓𝑥 + 𝑓𝑦 𝑦 ) ℎ 24 1 120 1
𝑖=1
𝑖=1 𝑖=1
𝑠
1 𝑠 2
𝜏𝑛+1
1 1
= ℎ3 [∑𝑏𝑖 𝑘𝑖 − ( 𝐹1(4) + ℎ𝐹1(5) + ⋅ ⋅ ⋅)] ,
+ ∑𝑏𝑖 𝑐𝑖2 (𝑓𝑥𝑥 + 2𝑓𝑥𝑦 𝑦 + 𝑓𝑦 𝑦 + 𝑓𝑦𝑦 (𝑦 ) ) ℎ2 6 24
2 𝑖=1 𝑖=1
𝑠
+ 𝑂 (ℎ3 ) , 1 1 (16)
𝜏𝑛+1 = ℎ2 [∑𝑏𝑖 𝑘𝑖 − ( 𝐹1(4) + ℎ𝐹1(5) + ⋅ ⋅ ⋅)] ,
𝑠 𝑖=1 2 6
∑𝑏𝑖 𝑘𝑖
𝑖=1 𝜏𝑛+1
𝑠 𝑠 𝑠
1 𝑠 1 1
= ∑𝑏𝑖 𝐹1(4) + ∑𝑏𝑖 𝑐𝑖 ℎ𝐹1(5) + ∑𝑏𝑖 𝑐𝑖2 ℎ2 𝐹1(6) + 𝑂 (ℎ3 ) , = ℎ [∑𝑏𝑖 𝑘𝑖 − (𝐹1(4) + ℎ𝐹1(5) + ℎ2 𝐹1(6) + ⋅ ⋅ ⋅)] .
𝑖=1 𝑖=1 2 𝑖=1 𝑖=1 2 6
4 Mathematical Problems in Engineering
By offsetting (15) into (16) and expanding as a Taylor series third order:
expansion using computer algebra package MAPLE (see 𝑠
[15]), the local truncation errors or the order conditions for 𝑠- 1
∑𝑏𝑖 𝑐𝑖2 = , (28)
stage fifth-order RKFD method can be written as follows. 𝑖=1 3
The order conditions for 𝑦 are
fourth order: fourth order:
𝑠 𝑠
1
1
∑𝑏𝑖 = , (17) ∑𝑏𝑖 𝑐𝑖3 = , (29)
24 𝑖=1 4
𝑖=1
1
𝑏3 = , (5)
𝑛𝑝 +1
36𝑐33 − 48𝑐32 + 18𝑐3 𝜏
2
= √ ∑ (𝜏𝑖(5) ) ,
2
𝑖=1
6𝑐2 − 6𝑐3 + 1
𝑏1 = 3 , (5)
𝜏𝑔
6𝑐3 (−3 + 4𝑐3 ) 2
−1 + 𝑐3 where 𝜏(5) , 𝜏(5) , 𝜏(5) , and 𝜏(5) are the local truncation error
𝑏3 = − ,
36𝑐33 − 48𝑐32 + 18𝑐3 norms for 𝑦, 𝑦 , 𝑦 , and 𝑦 , respectively, and 𝜏𝑔(5) is the global
error.
−48𝑏1 𝑐3 + 72𝑏1 𝑐32 − 2 + 11𝑐3 − 12𝑐32 Consequently, we find the error norms of 𝑦, 𝑦 , and 𝑦 ,
𝑏2 = − ,
12 (3 − 8𝑐3 + 6𝑐32 ) respectively, as follows:
2 2
(5) 1 √ (−360𝑏3 + 11 − 720𝑐3 𝑏3 + 480𝑏1 𝑐3 − 360𝑏1 + 960𝑏3 𝑐3 − 14𝑐3 )
𝜏 = ,
2 240 (−2 + 3𝑐3 )
2
2 2 2
(5) 1 √ (−50𝑐3 + 480𝑏1 𝑐3 − 360𝑏1 𝑐3 + 48𝑐3 − 7)
𝜏 = , (35)
2 240 (−2 + 3𝑐3 )
2
2 2
(5) 1 √ (3 − 12𝑐3 + 10𝑐3 )
𝜏 = .
2 120 (−2 + 3𝑐3 )
2
6 Mathematical Problems in Engineering
Our goal is to choose the free parameters 𝑐3 , 𝑏1 , 𝑏1 , and 𝑏3 such RKFD method are written in Butcher tableau and denoted by
that the error norms of fifth-order conditions have minimal RKFD4 method as follows:
value. By plotting the graph of ‖𝜏(5) ‖2 versus 𝑐3 and choosing
a small value of 𝑐3 in the interval [0.7, 3], we find that 𝑐3 =
4 1
17/20 is the optimal value which yields a minimum value for −
‖𝜏(5) ‖2 = 3.787878788 × 10−4 . Substituting the value of 𝑐3 = 11 5
17/20 into ‖𝜏(5) ‖2 and ‖𝜏(5) ‖2 we get 17 19 19
20 125 125
(5) 1 √ 17 7 1
𝜏 = 2 −
2 440 (3 − 160𝑏1 + 214𝑏3 ) , 200 75 20
(36) (39)
(5) 1 √ 1 209 5
𝜏 = 2
2 1760 (−31 + 544𝑏1 ) . 18 1926 1926
47 847 100
Also through plotting the graph of ‖𝜏(5) ‖2 against 𝑏1 and 𝑏3 408 2568 1819
in the interval [−0.1, 0.5] and choosing a small value of 𝑏3 ,
we get that 𝑏3 = 1/20 is the optimal value which gives 𝑏1 = 47 1331 2000
17/200 and ‖𝜏(5) ‖2 = 2.272727273 × 10−4 . 408 2568 5457
Now, utilizing the same technique where we draw the
graph of ‖𝜏(5) ‖2 versus 𝑏1 in the interval [−1, 1], we find that 4.2. A Three-Stage RKFD Method of Order Five. In this
𝑏1 = 1/18 is the best choice, and with this value of 𝑏1 , we get section, a three-stage RKFD method of order five will be
‖𝜏(5) ‖2 = 4.419191919 × 10−4 . derived. The algebraic order conditions up to order five ((17)-
Therefore, the error equation of the fifth-order condition (18), (19)–(21), (22)–(25), and (26)–(30)) need to be solved.
of 𝑦 is as follows: The resulting system of equations consists of fifteen nonlinear
equations, solving the system simultaneously which results
in a solution with three free parameters 𝑏1 , 𝑎21 , and 𝑎31 as
(5) 1
𝜏 = follows:
2 4802160 (1604749085
2
+ 6194971660900𝑎21 + 8761174400𝑎21 𝑎31 3 √6
𝑐2 = + ,
+ 8761174400𝑎21 𝑎32 − 199207202920𝑎21 (37) 5 10
2 3 √6
+ 3097600𝑎31 + 6195200𝑎31 𝑎32 − 1408633600𝑎31 𝑐3 = − ,
5 10
2 1/2
+ 3097600𝑎32 − 1408633600𝑎32 ) . 1
𝑏1 = ,
9
Consequently, the global error is 4 √6
𝑏2 = − ,
9 36
(5) 1 4 √6
𝜏𝑔 =
2 144064800 (1452377332189 𝑏3 = + ,
9 36
2
+ 55754744948100𝑎21 + 78850569600𝑎21 𝑎31 1
𝑏1 = ,
+ 78850569600𝑎21 𝑎32 − 17928648262800𝑎21 9
(38) 7 √6
2
+ 27878400𝑎31 + 55756800𝑎31 𝑎32 𝑏2 = − ,
36 18
2
− 12677702400𝑎31 + 27878400𝑎32 7 √6
𝑏3 = + ,
1/2 36 18
− 12677702400𝑎32 ) .
1
𝑏1 = ,
18
By minimizing the error norm in (37) and global error in (38)
with respect to the free parameters 𝑎21 , 𝑎31 , and 𝑎32 , we get 1 √6
𝑏2 = − ,
𝑎21 = −1/5, 𝑎31 = 19/125, and 𝑎32 = 19/125, which produces 18 48
‖𝜏(5) ‖2 = 3.7878787879×10−4 and ‖𝜏𝑔(5) ‖2 = 7.3068870183× 1 √6
10−4 . Finally, all the coefficients of three-stage fourth-order 𝑏3 = + ,
18 48
Mathematical Problems in Engineering 7
1 √6 1 √6 (6) 1 √ 2 √
𝜏 =
𝑏2 = − + (− + ) 𝑏1 , 2 3600 (362 − 11760𝑎21 6 + 103200𝑎21 6
48 72 2 2
+ 600000𝑎21 √6𝑎31 − 36840𝑎21 − 34440𝑎31
1 √6 1 √6
𝑏3 = + −( + ) 𝑏1 ,
48 72 2 2 + 120√6 − 11840√6𝑎31 + 2100000𝑎21 𝑎31
131 16√6 12 3√6 2 2 2 1/2
𝑎32 = (− + ) 𝑎21 − 𝑎31 + − . + 1330800𝑎31 + 448800√6𝑎31 + 1498800𝑎21 ) .
125 125 625 2500
(43)
(40)
Also, the global error can be written as
Thus, these free parameters can be chosen by minimizing the
(6) 1 √ 2 √
𝜏𝑔 =
2 3600 (−15520𝑎21 6 + 26400𝑎21 6
local truncation error norms of the sixth-order conditions.
The error norms and the global error of the sixth-order
conditions are given by + 600000𝑎21 √6𝑎31 − 46680𝑎21 − 34440𝑎31 + 511
(44)
(6)
𝑛𝑝 +1
2
+ 162√6 − 11840√6𝑎31 + 2100000𝑎21 𝑎31
𝜏 = √ ∑ (𝜏𝑖(6) ) ,
2 2 2 2
𝑖=1
+ 1330800𝑎31 + 448800√6𝑎31 + 2127600𝑎21 ).
𝑛𝑝 +1
(6) 2 Now, minimizing the error coefficients in (43) and (44) with
𝜏 = √ ∑ (𝜏𝑖(6) ) ,
2 respect to the free parameters 𝑎21 , 𝑎31 , we obtain 𝑎21 =
𝑖=1
4059/187793 and 𝑎31 = −1502/532215 which gives 𝑎32 =
𝑛𝑝 +1 1826/569317. Thus, the error equations for 𝑦, 𝑦 , 𝑦 , and 𝑦
(6) 2
𝜏 = √ ∑ (𝜏𝑖(6) ) , are computed and given by
2 (41)
𝑖=1
(6)
𝜏 = 0,
𝑛𝑝 +1 2
(6) 2
𝜏 = √ ∑ (𝜏𝑖(6) ) , 𝜏(6) = 8.333333333 × 10−4 ,
2 2
𝑖=1
(45)
(6)
𝜏𝑔 𝜏(6) = 1.666666668 × 10−3 ,
2 2
(6)
𝑛𝑝 +1 𝑛𝑝 +1 𝑛𝑝 +1 𝑛𝑝 +1
𝜏 = 1.666666667 × 10−3 ,
2 2 2 2
= √ ∑ (𝜏𝑖(6) ) + ∑ (𝜏𝑖(6) ) + ∑ (𝜏𝑖(6) ) + ∑ (𝜏𝑖(6) ) , 2
𝑖=1 𝑖=1 𝑖=1 𝑖=1
and global error norm is
(6)
where 𝜏(6) , 𝜏(6) , 𝜏(6) , and 𝜏(6) are the local truncation 𝜏𝑔 = 2.499999999 × 10−3 . (46)
2
error norms for 𝑦, 𝑦 , 𝑦 , and 𝑦 of the RKFD method,
respectively. 𝜏𝑔(6) is the global error. The error equation of Therefore, the parameters of the three-stage fifth-order RKFD
method denoted by RKFD5 can be represented in Butcher
sixth-order condition for 𝑦 with respect to the free parameter
tableau as follows:
𝑏1 is as follows:
3 √6 4059
(6) 1 √ +
2 5 10 187793
𝜏 =
2 3600 (−19 + 1080𝑏1 ) .
(42)
3 √6 1502 1826
− −
5 10 532215 569317
The error equation ‖𝜏(6) ‖2 has a minimum value equal to
zero at 𝑏1 = 19/1080 ≈ 0.01759259259 which leads to 𝑏2 = 19 13 11√6 13 11√6
− +
13/1080 − 11√6/2160 and 𝑏3 = 13/1080 + 11√6/2160. The 1080 1080 2160 1080 2160 (47)
truncation error norms of the sixth-order condition of 𝑦, 𝑦 , 1 1 √6 1 √6
𝑦 , and 𝑦 are calculated as follows: − +
18 18 48 18 48
(6) 1 7 √6 7 √6
𝜏 = 0, − +
2 9 36 18 36 18
(6) 1 1 4 √6 4 √6
𝜏 = − +
2 1200 , 9 9 36 9 36
(6) 1 2 2 √ 5. Numerical Examples
𝜏 =
2 3600 (139 + 628800𝑎21 − 76800𝑎21 6
1/2
In this section, some numerical examples will be solved to
− 9840𝑎21 − 3760𝑎21 √6 + 42√6) , show the efficiency of the new RKFD methods of order four
8 Mathematical Problems in Engineering
0 −1
−1 −2
−2 −3
−4
−3
Log10 (max error)
2.5 3 3.5 4 4.5 2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 3.8 4
Log10 (function evaluations) Log10 (function evaluations)
Figure 1: The efficiency curves for Example 1 with ℎ = 0.1/2𝑖 , 𝑖 = Figure 3: The efficiency curves for Example 3 with ℎ = 0.1/2𝑖 , 𝑖 =
0, 2, 3, 4. 0, 2, 3, 4.
−3
−3
−4
−4
−5
−5
Log10 (max error)
−6
Log10 (max error)
−6
−7
−7
−8
−8
−9
−9
−10
−10
−11
−11
1.4 1.6 1.8 2 2.2 2.4 2.6 2.8 3 3.2 3.4
2.5 3 3.5 4 4.5
Log10 (function evaluations)
Log10 (function evaluations)
RKFD5 RK5Ns6
RKFD5 RK5Ns6 RKFD4 RK4s4
RKFD4 RK4s4 RK5Bs6 RK3/8
RK5Bs6 RK3/8
Figure 4: The efficiency curves for Example 4 with ℎ = 0.1/2𝑖 , 𝑖 =
Figure 2: The efficiency curves for Example 2 with ℎ = 0.1/2𝑖 , 𝑖 = 0, 2, 3, 4.
0, 2, 3, 4.
−1
−1
−2 −2
−3 −3
Log10 (max error)
−4 −4
Figure 5: The efficiency curves for Example 5 with ℎ = 0.1/2𝑖 , 𝑖 = Figure 7: The efficiency curves for Example 7 with ℎ = 0.025/2𝑖 ,
0, 2, 3, 4. 𝑖 = 0, 1, 2, 3.
−3
−4 Example 2. The nonhomogeneous nonlinear problem is as
follows:
−5
−6
Log10 (max error)
RKFD5 RK5Ns6
RKFD4 RK4s4
𝑦(𝑖V) = (16𝑥4 − 48𝑥2 + 12) 𝑦,
RK5Bs6 RK3/8 (50)
𝑦 (0) = 1, 𝑦 (0) = 0, 𝑦 (0) = −2, 𝑦 (0) = 0.
Figure 6: The efficiency curves for Example 6 with ℎ = 0.1/2𝑖 , 𝑖 =
0, 2, 3, 4.
2
The exact solution is given by 𝑦(𝑥) = e−𝑥 . The problem is
integrated in the interval [0, 3].
(vi) RK3/8: the four-stage fourth-order 3/8 rule Runge-
Kutta method given in Butcher [4]. Example 4. The nonlinear problem is as follows:
The exact solution is given by 𝑦(𝑥) = e𝑥 sin(𝑥). The problem The exact solution is given by 𝑦(𝑥) = arcsin(𝑥). The problem
is integrated in the interval [0, 10]. is integrated in the interval [0, 𝜋/4].
10 Mathematical Problems in Engineering
Example 5. The linear system is as follows: Example 7. The nonlinear system is as follows:
(𝑖V) 3𝑥
𝑦 = e 𝑢, 𝑧2
𝑦(𝑖V) = ,
𝑤
𝑦 (0) = 1, 𝑦 (0) = −1, 𝑦 (0) = 1, 𝑦 (0) = −1.
𝑦 (0) = 1, 𝑦 (0) = 1, 𝑦 (0) = 1, 𝑦 (0) = 1.
(𝑖V) −𝑥
𝑧 = 16e 𝑦,
𝑤2
𝑧 (0) = 1, 𝑧 (0) = −2, 𝑧 (0) = 4, 𝑧 (0) = −8. 𝑧(𝑖V) = 16 ,
𝑢
(52)
𝑤(𝑖V) = 81e−𝑥 𝑧, 𝑧 (0) = 1, 𝑧 (0) = 2, 𝑧 (0) = 4, 𝑧 (0) = 8.
(56)
𝑤 (0) = 1, 𝑤 (0) = −3, 𝑤 (0) = 9, 𝑤 (0) = −27. 𝑢2
𝑤(𝑖V) = 81 5 ,
𝑦
𝑢(𝑖V) = 256e−𝑥 𝑤.
𝑤 (0) = 1, 𝑤 (0) = 3, 𝑤 (0) = 9, 𝑤 (0) = 27.
𝑢 (0) = 1, 𝑢 (0) = −4, 𝑢 (0) = 16, 𝑢 (0) = −64.
𝑢(𝑖V) = 256𝑦4 .
The exact solution is given by
𝑦 = e−𝑥 , 𝑢 (0) = 1, 𝑢 (0) = 4, 𝑢 (0) = 16, 𝑢 (0) = 64.
The exact solution is given by
𝑧 = e−2𝑥 ,
(53) 𝑦 = e𝑥 ,
𝑤 = e−3𝑥 ,
𝑧 = e2𝑥 ,
𝑢 = e−4𝑥 . (57)
The problem is integrated in the interval [0, 2]. 𝑤 = e3𝑥 ,
𝑧 = sin (𝑥) ,
(55) References
𝑤 = cos (2𝑥) ,
[1] P. Onumanyi, U. W. Sirisena, and S. N. Jator, “Continuous finite
𝑢 = sin (2𝑥) . difference approximations for solving differential equations,”
International Journal of Computer Mathematics, vol. 72, no. 1,
The problem is integrated in the interval [0, 2]. pp. 15–27, 1999.
Mathematical Problems in Engineering 11
International
Journal of Journal of
Mathematics and
Mathematical
Discrete Mathematics
Sciences