A numerical approach for a nonhomogeneous differential equation with variable delays

A numerical approach for a nonhomogeneous differential equation with variable delays In this study, we consider a linear nonhomogeneous differential equation with variable coefficients and variable delays and present a novel matrix-collocation method based on Morgan–Voyce polynomials to obtain the approximate solutions under the initial conditions. The method reduces the equation with variable delays to a matrix equation with unknown Morgan– Voyce coefficients. Thereby, the solution is obtained in terms of Morgan–Voyce polynomials. In addition, two test problems together with error analysis are performed to illustrate the accuracy and applicability of the method; the obtained results are scrutinized and interpreted by means of tables and figures. Keywords Morgan–Voyce polynomials  Matrix method  Collocation method  Delay differential equation Variable delay Introduction including production of blood cells [1, 14, 23, 25, 27, 28, 34, 37]. In this paper, we consider nonhomogeneous differential In the case of bounded delays, many authors using equation with variable delays in the form [3, 5, 10, 12, standard techniques [3, 10, 20, 27, 34, 37] have studied the 23, 30, 37, 38]. asymptotic behavior of solutions, the asymptotic stability m in equations and the existence of positive periodic solutions y ðÞ t ¼ P ðtÞþ P ðtÞyðtÞþ P ðtÞyt  s ðtÞ of delay equations. However, most of the mentioned type o 1 j j ð1Þ j¼2 delay equations have not analytical and numerical solu- tions; therefore, numerical methods are required to obtain under the initial condition yðaÞ¼ k, where the coefficients approximate solutions. For this purpose, by means of the P ðtÞ and the delays s are continuous functions on the j j matrix method based on collocation points which have interval 0  a  j  b and the delays are nonnegative, been given by Sezer and coworkers [2, 6, 16, 17, 21, s ðÞ t  0 for t  a. 26, 29, 36], we develop a novel matrix technique to find the Delay differential equations of the type 1 arise in a approximate solution of Eq. 1 under the initial condition variety of applications including control systems, electro- yðaÞ¼ k in the truncated Morgan–Voyce series form dynamics, mixing liquids, neutron transportation, popula- tion models, physiological processes and conditions ytðÞ ffi y ðÞ t ¼ y b ðtÞ; a  t  b ð2Þ N n n n¼0 & Mustafa Ozel where y ; n ¼ 0; 1; ...; N are coefficients to be determined; mustafa.ozel@deu.edu.tr n b ; n ¼ 0; 1; ...; N are the first kind Morgan–Voyce poly- Department of Geophysical Engineering, Faculty of nomials defined by [11, 32, 33] Engineering, Dokuz Eylul University, Tınaztepe Campus, Buca, 35160 Izmir, Turkey n þ j b ðÞ t ¼ t ; n 2 N; a  t  b n ð3Þ Department of Physics, Faculty of Science, Dokuz Eylul n  j j¼0 University, Tınaztepe Campus, Buca, 35160 Izmir, Turkey Department of Mathematics, Faculty of Art and Science, Celal Bayar University, Manisa, Turkey 123 146 Mathematical Sciences (2018) 12:145–155 and Here, the set of polynomialsfg b ðÞ t has the following 2 3 0 1 2 N properties [11, 15, 19, 32] 6 7 0 1 2 N 6 7 6    7 1. The polynomials b ðÞ t defined by 3 are recursively n 6 2 3 N þ 1 7 6 7 6 7 given by the relation 0 1 N  1 6 7 6   7 6 4 N þ 2 7 M ¼ 6 7 b ðÞ t ¼ðÞ t þ 2 b ðÞ t  b ðÞ t ; n  2 n n1 n2 6 7 0 N  2 6 7 6 7 . . . . 6 7 . . . . with b ðÞ t ¼ 1 and b ðÞ t ¼ t þ 1. 6 . . . . 7 o n 6  7 4 5 2N 2. The polynomials y ¼ b ðÞ t ; n ¼ 0; 1; ... are solutions n 00 0 of the differential equation 00 0 Besides, the relation between the matrix and its derivative ttðÞ þ 4 y þ 2ðÞ t þ 1 y  nnðÞ þ 1 y ¼ 0: X ðÞ t can be written in the form [13, 18, 22, 24] 3. The first four Morgan–Voyce polynomials of the first X ðÞ t ¼ XtðÞT ð6Þ kind are obtained from 3 as where b ðÞ t ¼ 1; b ðÞ t ¼ t þ 1; b ðÞ t ¼ t þ 3t þ 1; o 1 2 2 3 3 2 01 0 ... 0 b ðÞ t ¼ t þ 5t þ 6t þ 1; ... 6 7 00 2 ... 0 6 7 6 7 . . . . 6 7 . . . . T ¼ : . . . . 6 7 Fundamental matrix relations 6 7 4 5 00 0 ... N 00 0 ... 0 In this section, we compose the matrix relations of Eq. 1 and its solution Eq. 2. For this aim, we first write the Then, by means of the matrix relations Eqs. 4, 5, and 6,we matrix form of the finite Morgan–Voyce series Eq. 2 as obtain ytðÞ ffi y ðÞ t ¼ btðÞY ð4Þ ytðÞ ffi y ðtÞ¼ btðÞY ð7Þ so that ¼ XtðÞMY btðÞ ¼½ b ðÞ t ; b ðÞ t ; .. .; b ðÞ t o 1 N and 0 0 0 Y ¼½ y ; y ; ...; y ; o 1 N y ðÞ t ffi y ðtÞ¼ X ðÞ t MY ð8Þ then, by using the Morgan–Voyce polynomials Eq. 3,we ¼ XtðÞTMY obtain the matrix form btðÞ as follows By putting t ! t  s ðÞ t in Eq. 7, we gain the recurrence btðÞ ¼ XtðÞM ð5Þ relation [13, 18, 22, 24] where yt  s ðtÞ ffi y ðtÞ¼ Xt  s ðtÞ MY j N j ð9Þ 2 N ¼ XtðÞL s ðtÞ MY XtðÞ ¼ 1; t; t ; ...; t so that 2    3 0 1 2 N 0 1 2 N s ðÞ t s ðÞ t s ðÞ t  s ðÞ t j j j j 6 7 0 0 0 0 6 7 6    7 6    7 1 2 N 0 1 N1 6 7 0 s ðÞ t s ðÞ t  s ðÞ t j j j 6 7 1 1 1 6 7 6 7 6 2 N 7 0 N2 L s ðÞ t ¼ 6 7 00 s ðÞ t  s ðÞ t j j 6 7 2 2 6 7 6 7 . . . . 6 7 . . . . 6 7 . . . . 6  7 4 N  5 000  s ðÞ t 123 Mathematical Sciences (2018) 12:145–155 147 Note that the matrix Xt  s ðtÞ can be written as U ¼ XðaÞM ¼½ u ; u ; ...; u 00 01 0N Xt  s ðtÞ ¼ XtðÞL s ðtÞ j j Consequently, in order to get the approximate solution of Eq. 1 subject to yðaÞ¼ k, we replace the row matrix in By substituting the relations Eqs. 7, 8, and 9 into Eq. 1,we Eq. 13 by the last row(or any row) of the augmented matrix have the matrix equation in Eq. 12; then, we obtain the result matrix "# ~ ~ ~ ~ W; P , WY ¼ P ð14Þ o o XtðÞT  P ðÞ t XtðÞ  P ðÞ t XtðÞL s ðÞ t MY 1 j j j¼2 ~ ~ ~ If rank W ¼ rank W; P ¼ N þ 1, then we can write, ¼ P ðÞ t ð10Þ ~ ~ Y ¼ W P . Thus the matrix, Y (thereby the Morgan– and by placing the collocation points defined by Voyce coefficients y ; y ; ...; y ) is uniquely determined; o 1 N thus Eq. 1 has a unique solution. b  a t ¼ a þ i; i ¼ 0; 1; ...; N: in Eq. 10, the compact form of the obtained matrix equa- Error analysis tions system "# In this section, an error analysis will be presented for the XT  P X  P ðÞ t XL s MY ¼ P ð11Þ 1 j j o Morgan–Voyce polynomial solution in Eq. 16 with the j¼2 residual error function [4, 8, 9, 13, 21, 22, 28, 31, 36]. where In addition, we will improve the Morgan–Voyce poly- 2 3 2 3 2 N nomial solution y ðÞ t with the aid of the residual error XtðÞ 1 t t  t o o o o function. 2 N 6 7 6 7 XtðÞ 1 t t  t 1 1 6 7 6 1 1 7 6 7 6 7 Firstly, we consider the operator Eq. 1, under the initial X ¼ ¼ ; 6 . 7 6 . . . . 7 . . . . . 4 5 4 5 condition yðaÞ¼ k, . . . . 2 N XtðÞ 1 t t  t N N N N Ly½ ðÞt ¼ gðtÞ 2 3 2 3 L sðÞ t P ðt Þ o o j o 6 7 6 7 Ly½ ðtÞ ¼ y ðÞ t  P ðtÞ P ðtÞyðtÞ P ðtÞyt  s ðtÞ : 6 L sðÞ t 7 P ðt Þ o 1 j j o 1 j 1 6 7 6 7 6 7 j¼2 P ¼ ; L s ¼ 6 7 o j 6 7 . . 6 7 4 5 4 5 ð15Þ P ðt Þ L sðÞ t o N j N Here, y ðtÞ is the approximate solution of the problem and P ¼ diag P ðt Þ; P ðt Þ; ...; P ðt Þ ; j ¼ 1; 2; ...; m o j o j 1 j N satisfies the problem X ¼ diag½ Xðt Þ; Xðt Þ; ...; Xðt Þ ; o 1 N L½y ðtÞ ¼ gðtÞþ R ðtÞ; a  t  b; N N ð16Þ y ðtÞ¼ k: Morgan–Voyce matrix method Also, the residual function of the Morgan–Voyce polyno- mial approximation y ðÞ t is defined as The fundamental matrix Eq. 11 of Eq. 1 can be expressed in the form R ðtÞ¼ L½y ðtÞ  gðtÞ: ð17Þ N N WY ¼ P ,½ W; P ð12Þ o o If we know the exact solution ytðÞ, then the error function where is calculated as the difference between the approximate and "# the exact solutions defined by W ¼ w ¼ XT  P X  P ðÞ t XL s M pq 1 j j e ðÞ t ¼ ytðÞ y ðÞ t : ð18Þ N N j¼2 By using the Eqs. 15, 16, 17 and 18, we get the error p; q ¼ 0; 1; ...; N: problem By using the relation Eq. 7, we obtain the corresponding L½e ðtÞ ¼ L½yðtÞ  L½y ðtÞ ¼ R ðtÞ: N N N matrix form to the initial condition yðaÞ¼ k as ð19Þ e ðÞ a ¼ 0: UY ¼ k ,½ U; Y ð13Þ such that 123 148 Mathematical Sciences (2018) 12:145–155 By solving the error problem in Eq. 19 with the method Numerical examples presented in Sect. 3, we get the approximation e ðtÞ to N;M e ðtÞ as follows N Example 1 Consider the differential equation with vari- able delay t þ 1 e ðtÞ¼ a b ðtÞ; ðM  NÞ: N;M n 0 2 2 2 2 y ðtÞ¼3t  t þ t yðtÞ yðt  t Þþ tyðt  t  1Þ n¼0 ð20Þ 0  t  1 Consequently, by means of the polynomials y ðtÞ and e ðtÞ, ðM  NÞ, we obtain the corrected Morgan–Voyce N;M polynomial solution y ðtÞ¼ y ðtÞþ e ðtÞ. Here, N;M N N;M subject to the initial condition yð0Þ¼1. The exact e ðtÞ¼ yðtÞ y ðtÞ, E ðtÞ¼ e ðtÞ e ðtÞ¼ N N N;M N N;M solution of this equation is yðtÞ¼ t  1. First of all, let us yðtÞ y ðtÞ and e ðtÞ denote the error function, the N;M N;M determine the collocation points by the formula t ¼ a þ corrected error function and the estimated error function, ba i; i ¼ 0; 1; ...; N for a ¼ 0, b ¼ 1, and m ¼ 3; N ¼ 2. respectively. Therefore the collocation points are obtained as If the exact solution of Eq. 1 can not been known, then t ¼ 0; t ¼ ; t ¼ 1. By Eq. 11, the fundamental matrix o 1 2 the absolute errorsjj e ðt Þ ¼jj yðt Þ y ðt Þ ,(a  t  b) 2 N i i N i i equation of this problem is written as are not computed. However, the absolute errors "# jj e ðt Þ ¼jj yðt Þ y ðt Þ ,(a  t  b) can be estimated by 3 N i i N i i X XT  P X  P ðÞ t XL s MY ¼ P 1 j j o using the absolute error function e ðtÞ . N;M j¼2 where 2 3 2 3 2 3 2 3 0 00 0 000 10 0 6 7 5 1 1 6 7 6 7 6 7 6 7 P ¼ ; P ¼ 0 0 ; P ¼ 0 10 ; P ¼ 0 0 4 5 4 5 4 5 o 1 2 3 4 5 4 4 2 00 1 2 00 1 001 XtðÞ ¼ 1 tt 2 3 2 3 2 3 2 3 XðÞ 0 10 0 010 111 6 7 1 1 1 6 7 6 7 6 7 6 7 T ¼4002 5; M ¼ 4013 5; X ¼ 6 X 7 ¼ 4 1 5 4 2 5 2 4 000 001 11 1 XðÞ 1 2 3 2 3 XðÞ 0 00 10 000 00 00 6 7 1 1 1 6 7 6 7 X ¼6 0 X 0 7 ¼ 400 01 00 0 5 4 2 5 2 4 00 000 01 11 00 XðÞ 1 2 3 2 3 100 1 11 6 7 6 7 010 01 2 6 7 6 7 6 7 6 7 6 7 6 7 001 00 1 6 7 6 7 6 1 1 7 6 5 25 7 6 7 6 7 1  1 6 7 6 7 4 16 4 16 6 7 6 7 6 1 7  6 5 7 LðÞ s ðt Þ¼ ; LðÞ s ðt Þ¼ : 2 i 3 i 6 7 6 7 01  01 6 7 6 7 2 2 6 7 6 7 6 001 7 6 7 00 1 6 7 6 7 6 7 6 7 1 11 1 24 6 7 6 7 6 7 6 7 4 5 4 5 01 2 01 4 001 00 1 123 Mathematical Sciences (2018) 12:145–155 149 Table 1 Numerical results of t N ¼ 3; M ¼ 4 N ¼ 4; M ¼ 5 N ¼ 12; M ¼ 13 the exact, approximate and corrected solutions of Exact Appr. Corrected Appr. Corrected Appr. Corrected Example 2 for some N values 0 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 0.1 1.22140 1.39581 1.20219 1.20219 1.22571 1.22140 1.22140 0.2 1.49182 1.75739 1.46663 1.46663 1.49689 1.49182 1.49182 0.3 1.82212 2.11993 1.79678 1.79678 1.82697 1.82212 1.82212 0.4 2.22554 2.51864 2.20151 2.20151 2.23024 2.22554 2.22554 0.5 2.71828 2.98872 2.69510 2.69510 2.72299 2.71828 2.71828 0.6 3.32012 3.56539 3.29724 3.29724 3.32478 3.32012 3.32012 0.7 4.05520 4.28383 4.03301 4.03301 4.05954 4.05520 4.05520 0.8 4.95303 5.17927 4.93292 4.93292 4.95683 4.95303 4.95303 0.9 6.04965 6.28689 6.03286 6.03286 6.05299 6.04965 6.04965 1 7.38906 7.64191 7.37414 7.37414 7.39236 7.38906 7.38906 Fig. 1 Exact, approximate and 9 corrected solutions of Example 2 for N ¼ 3 Exact Approximation Corrected 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Fig. 2 Exact, approximate and 9 corrected solutions of Example 2 for N ¼ 4 Exact Approximation Corrected 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 2 3 The augmented matrix of the fundamental matrix equation 12 4 0 is computed as 6 7 ½ W; P ¼ 4 0:25 1:75 5:46875 1:25 5 10 2 1 y(t) y(t) 150 Mathematical Sciences (2018) 12:145–155 Table 2 Comparison of the t N ¼ 3; M ¼ 4 N ¼ 4; M ¼ 5 N ¼ 12; M ¼ 13 absolute errors and corrected absolute errors of Example 2 Appr. Corrected Appr. Corrected Appr. Corrected 0 7.99361e-15 6.21725e-15 3.10862e-15 1.02141e-14 2.02061e-14 4.97380e-14 0.1 1.74411e-01 1.92085e-02 1.92085e-02 4.30490e-03 5.60576e-11 2.89457e-12 0.2 2.65564e-01 2.51946e-02 2.51946e-02 5.06186e-03 5.52192e-11 2.88503e-12 0.3 2.97810e-01 2.53429e-02 2.53429e-02 4.85123e-03 5.51801e-11 2.87947e-12 0.4 2.93098e-01 2.40357e-02 2.40357e-02 4.69433e-03 5.49143e-11 2.87148e-12 0.5 2.70441e-01 2.31845e-02 2.31845e-02 4.71136e-03 5.42593e-11 2.86038e-12 0.6 2.45269e-01 2.28805e-02 2.28805e-02 4.66159e-03 5.29470e-11 2.80753e-12 0.7 2.28632e-01 2.21878e-02 2.21878e-02 4.33963e-03 5.04476e-11 2.68496e-12 0.8 2.26233e-01 2.01130e-02 2.01130e-02 3.79602e-03 4.58762e-11 2.45759e-12 0.9 2.37243e-01 1.67888e-02 1.67888e-02 3.34318e-03 3.76135e-11 2.08455e-12 1.0 2.52855e-01 1.49208e-02 1.49208e-02 3.29924e-03 2.63132e-11 1.53477e-12 Fig. 3 Comparison of the 0.35 absolute error with the corrected 0.3 absolute error for N ¼ 3 0.25 0.2 Absolute Error 0.15 Corrected Absolute Error 0.1 0.05 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Fig. 4 Comparison of the 0.05 absolute error with the corrected Absolute Error absolute error for N ¼ 4 0.04 Corrected Absolute Error 0.03 0.02 0.01 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 and the augmented matrix for initial condition is obtained Example 2 Consider the differential equation with vari- as able delay t 0 2t 2 2t ½ U; k¼½ 111 j  1 : 2y ðtÞ tyðtÞþ te yðt  t Þ¼ 4e ; 0  t  1 By using the procedure in Sect. 3, we obtain the approxi- subject to the initial condition yð0Þ¼ 1. The exact solution 2t mate solution as of this equation is yðtÞ¼ e . The fundamental matrix equation is y ðtÞ¼ t  1 ½ XT  P X  P XLðÞ s ðÞ t MY ¼ P : 1 2 2 i o which is the exact solution. e( t) e( t) Mathematical Sciences (2018) 12:145–155 151 Table 3 Numerical results of t N ¼ 3; M ¼ 4 N ¼ 4; M ¼ 5 N ¼ 12; M ¼ 13 the exact, approximate and corrected solutions of Exact Appr. Corrected Appr. Corrected Appr. Corrected Example 3 for some N values 0 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 0.1 0.90484 0.90675 0.90500 0.90500 0.90485 0.90484 0.90484 0.2 0.81873 0.82159 0.81894 0.81894 0.81874 0.81873 0.81873 0.3 0.74082 0.74401 0.74103 0.74103 0.74083 0.74082 0.74082 0.4 0.67032 0.67350 0.67052 0.67052 0.67033 0.67032 0.67032 0.5 0.60653 0.60956 0.60672 0.60672 0.60654 0.60653 0.60653 0.6 0.54881 0.55168 0.54900 0.54900 0.54882 0.54881 0.54881 0.7 0.49659 0.49936 0.49677 0.49677 0.49660 0.49659 0.49659 0.8 0.44933 0.45209 0.44950 0.44950 0.44934 0.44933 0.44933 0.9 0.40657 0.40937 0.40673 0.40673 0.40658 0.40657 0.40657 1 0.36788 0.37069 0.36803 0.36803 0.36789 0.36788 0.36788 Fig. 5 Exact, approximate and 1 corrected solutions of Exact 0.9 Example 3 for N ¼ 3 Approximation 0.8 Corrected 0.7 0.6 0.5 0.4 0.3 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Fig. 6 Exact, approximate and corrected solutions of Exact 0.9 Example 3 for N ¼ 4 Approximation 0.8 Corrected 0.7 0.6 0.5 0.4 0.3 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 After the collocation points substituted into this matrix The absolute and corrected errors of Example 2 in equation, we solve the system and we obtain the solutions Table 2 are compared for N ¼ 3; 4 in Figs. 3 and 4, in the form 4 of Example 2 for N ¼ 3; 4; 12 in the interval respectively. ½ 0; 1 Table 1. Example 3 Consider the delay differential equation having Now, we will give the exact, approximation, and cor- variable delays lnðÞ t þ 1 and t rected solutions of Example 2 for N ¼ 3; 4 in Figs. 1 and 2, respectively. y(t) y(t) 152 Mathematical Sciences (2018) 12:145–155 Table 4 Comparison of the t N ¼ 3; M ¼ 4 N ¼ 4; M ¼ 5 N ¼ 12; M ¼ 13 absolute errors and corrected absolute errors of Example 3 Appr. Corrected Appr. Corrected Appr. Corrected 0 0 0 2.22045e-16 2.22045e-16 4.44089e-16 4.44089e-16 0.1 1.91653e-03 1.57584e-04 1.57584e-04 1.15985e-05 1.16573e-13 2.98650e-14 0.2 2.86131e-03 2.06131e-04 2.06131e-04 1.35662e-05 1.13909e-13 2.95319e-14 0.3 3.19084e-03 2.07981e-04 2.07981e-04 1.30496e-05 1.12577e-13 2.89768e-14 0.4 3.17959e-03 1.97746e-04 1.97746e-04 1.25881e-05 1.09690e-13 2.83107e-14 0.5 3.02786e-03 1.90112e-04 1.90112e-04 1.24366e-05 1.05804e-13 2.74225e-14 0.6 2.86876e-03 1.86905e-04 1.86905e-04 1.21434e-05 1.01363e-13 2.64233e-14 0.7 2.77470e-03 1.83479e-04 1.83479e-04 1.14580e-05 9.62008e-14 2.50355e-14 0.8 2.76306e-03 1.74494e-04 1.74494e-04 1.06297e-05 9.02611e-14 2.35367e-14 0.9 2.80153e-03 1.59153e-04 1.59153e-04 1.01563e-05 8.32667e-14 2.20934e-14 1 2.81276e-03 1.45931e-04 1.45931e-04 1.00352e-05 7.82152e-14 1.96509e-14 -3 Fig. 7 Comparison of the x 10 absolute error with the corrected 3.5 absolute error for N ¼ 3 2.5 1.5 Absolute Error Corrected Absolute Error 0.5 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 -4 Fig. 8 Comparison of the x 10 absolute error with the corrected absolute error for N ¼ 4 Absolute Error Corrected Absolute Error 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 2 t t 2 ½ XT  P X  P XLðÞ s ðÞ t  P XLðÞ s ðÞ t MY ¼ P : y ðtÞ¼ t þ t  1 e  tyðÞ t  lnðÞ t þ 1 e yt  t 1 2 2 i 3 3 i o þ yðtÞ; 0  t  1 subject to the initial condition yð0Þ¼ 1. The exact solution After the collocation points substituted into this matrix of this equation is yðtÞ¼ e . The fundamental matrix equation, we can solve the system and we obtain the equation is e(t) e(t) Mathematical Sciences (2018) 12:145–155 153 Table 5 Numerical results of t Exact Present Abs. error HTL Abs. error the Exact, Present and HTL solutions of Example 4 for N ¼ 0.0 1.0000000 1.0000000 5.55112e-15 1.000000067 6.70000e-08 6 value 0.2 0.8187308 0.8187339 3.10453e-06 0.807836989 1.08938e-02 0.4 0.6703200 0.6703221 2.03288e-06 0.649394632 2.09254e-02 0.6 0.5488116 0.5488127 1.09160e-06 0.529079726 1.97319e-02 0.8 0.4493290 0.4493294 4.48652e-07 0.436772222 1.25567e-02 1.0 0.3678794 0.3678793 1.14014e-07 0.361610346 6.26909e-03 1.2 0.3011942 0.3011936 5.72331e-07 0.298780813 2.41340e-03 1.4 0.2465970 0.2465960 9.17880e-07 0.249314192 2.71723e-03 1.6 0.2018965 0.2018953 1.22873e-06 0.212885438 1.09889e-02 1.8 0.1652989 0.1652975 1.43489e-06 0.173619567 8.32068e-03 2.0 0.1353353 0.1353338 1.44145e-06 0.078902496 5.64328e-02 Fig. 9 Comparison of the exact, 1.2 present and HTL solutions of Exact Example 4 for N ¼ 6 value Present 1.0 HTL 0.8 0.6 0.4 0.2 0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 Fig. 10 RMSE values for N ¼ 1 1; 2; .. .; 100 for Examples 2, 3 Example 2 -2 and 4 Example 3 Example 4 -4 -6 -8 -10 -12 -14 -16 0 102030405060708090 100 y(t) RESM 154 Mathematical Sciences (2018) 12:145–155 creativecommons.org/licenses/by/4.0/), which permits unrestricted solutions in the form 4 of Example 3 for N ¼ 3; 4; 12 in use, distribution, and reproduction in any medium, provided you give the interval½ 0; 1 Table 3. appropriate credit to the original author(s) and the source, provide a All exact, approximation, and corrected solutions of link to the Creative Commons license, and indicate if changes were Example 3 for N ¼ 3; 4 are given in Fig. 5 and 6, respec- made. tively. The absolute and corrected errors of Example 3 in Table 4 are compared for N ¼ 3; 4 in Figs. 7 and 8, References respectively. 1. Abd-Elhameed, W.M., Youssri, Y.H., Doha, E.H.: A novel operational matrix method based on shifted Legendre polyno- Example 4 Consider the following delay differential 2 mials for solving second-order boundary value problems equation [7] having variable delay lnðÞ t þ 1 involving singular, singularly perturbed and Bratu-type equa- 0 2 t 2 tions. Math. Sci. 9, 93–102 (2015) y ðtÞ¼ t þ 1 e yt  ln t þ 1  yðtÞ; 0  t  2 2. Akyu¨z, A., Sezer, M.: A Chebyshev collocation method for the solution of linear integro-differential equations. Int. J. Comput. subject to the initial condition yð0Þ¼ 1. The exact solution Math. 72, 491–507 (1999) of this equation is yðtÞ¼ e . 3. Ardjouni, A., Djoudi, A.: Fixed points and stability in linear neutral differential equations with variable delays. Nonlinear Similarly, we can solve this problem by present method Anal. Theory Methods Appl. 74, 2062–2070 (2011) and we obtain the solutions in the form 4 of Example 4 for 4. Bahs¸ı, M.M., C ¸ evik, M., Sezer, M.: Orthoexponential polynomial N ¼ 6 in the interval½ 0; 2 . Then, we also compare the solutions of delay pantograph differential equations with residual present solutions and Hybrid Taylor-Lucas Method [7] error estimation. Appl. Math. Comput. 271, 11–21 (2015) 5. Bahs¸ı, M.Mustafa, Bahs¸ı, Ays¸e Kurt, C ¸ evik, Mehmet, Sezer, (HTL) solutions in Table 5. The present method has been Mehmet: Improved Jacobi matrix method for the numerical shown to be suitable as graphically for nonhomogeneous solution of Fredholm integro-differential-difference equations. differential equation with variable delays in Fig. 9. Math. Sci. 10, 83–93 (2016) In order to better define the solution space of the prob- 6. Balcı, M.A., Sezer, M.: Hybrid Euler-Taylor matrix method for solving of generalized linear Fredholm integro-differential dif- lems of Examples 2, 3 and 4 described above, the N value ference equations. Appl. Math. Comput. 273, 33–41 (2016) has been scanned up 1 to 100. The root-mean-square error 7. Bayku, N., Sezer, M.: Hybrid Taylor–Lucas collocation method (RMSE) value of the solution obtained for each N value is for numerical solution of high-order pantograph type delay dif- calculated and shown in Fig. 10. When N [ 12, the RMSE ferential equations with variables delays. Appl. Math. Inf. Sci. 11, 1795–1801 (2017) values of the solutions of both problems are oscillated 8. C ¸ elik, I.: Approximate calculation of eigenvalues with the 10 16 between 10 and 10 . The reason of these oscillations method of weighted residuals-collocation method. Appl. Math. is the truncation errors in the calculations. Comput. 160, 401–410 (2005) 9. C ¸ elik, I.: Collocation method and residual correction using Chebyshev series. Appl. Math. Comput. 174, 910–920 (2006) 10. Dix, J.G.: Asymptotic behavior of solutions to a first-order dif- Conclusion ferential equation with variable delays. Comput. Math. Appl. 50, 1791–1800 (2005) 11. Djordjevic, G.B., Milovanovic, G.V.: Special classes of polyno- A new approach using the Morgan–Voyce polynomials to mials. Univ. Nis Fac. Technol. Leskovac 58, 11–18 (2014) solve numerically the first-order nonhomogeneous differ- 12. Elahi, Zaffer, Akram, Ghazala, Siddiqi, Shahid Saeed: Numerical ential equations with variable delays is presented in this solution for solving special eighth-order linear boundary value study. An error analysis technique based on residual problems using Legendre Galerkin method. Math. Sci. 10, 201–209 (2016) function is also developed our problems. If the exact 13. Erdem, K., Yalc¸inbas¸, S., Sezer, M.: A Bernoulli polynomial solution of the problem is not known, by using this tech- approach with residual correction for solving mixed linear nique it is possible to estimate the error function and also to Fredholm integro-differential-difference equations. J. Differ. reduce the error due to the residual function. It is seen that, Equ. Appl. 19, 1619–1631 (2013) 14. Erfanian, M., Zeidabadi, H.: Using of Bernstein spectral Galerkin the accuracy improves, when N is increased. To compute method for solving of weakly singular Volterra–Fredholm inte- our solutions and error functions, we have written a code in gral equations. Math. Sci. (2018). https://doi.org/10.1007/s40096- Matlab and calculated all computations by means of this 018-0249-1 code. 15. Fortuna, L., Frasca, M.: Generating passive systems from recur- Consequently, the present method has been shown to be sively defined polynomials. Int. J. Circuits Syst. Signal Process. 6, 179–188 (2012) convenient, reliable and effective for solving the first-order 16. Gu¨lsu, M., Sezer, M.: Approximations to the solution of linear nonhomogeneous differential equation with variable Fredholm integrodifferential-difference equation of high order. delays. J. Franklin Inst. 343, 720–737 (2006) 17. Gu¨rbu¨z, B., Sezer, M., Gu¨ler, C.: Laguerre collocation method for solving Fredholm integro-differential equations with functional Open Access This article is distributed under the terms of the Creative arguments. J. Appl. Math. 2014, 48 (2014) Commons Attribution 4.0 International License (http:// 123 Mathematical Sciences (2018) 12:145–155 155 18. Is¸ik, O.R., Gu¨ney, Z., Sezer, M.: Bernstein series solutions of 29. Sezer, M., Akyu¨z-Das¸cıoglu, A.: A Taylor method for numerical pantograph equations using polynomial interpolation. J. Differ. solution of generalized pantograph equations with linear func- Equ. Appl. 18, 357–374 (2012) tional argument. J. Comput. Appl. Math. 200, 217–225 (2007) 19. Ilhan, O., Sahin, N.: On Morgan–Voyce polynomials approxi- 30. Singh, Randhir, Wazwaz, Abdul-Majid: Numerical solutions of mation for linear differential equations. Int. J. Comput. Math. 72, fourth-order Volterra integro-differential equations by the Greens 491–507 (2014) function and decomposition method. Math. Sci. 10, 159–166 20. Jin, C., Luo, J.: Fixed points and stability in neutral differential (2016) equations with variable delays. Proc. Am. Math. Soc. 136, 31. Shahmorad, S.: Numerical solution of the general form linear 909–918 (2008) Fredholm-Volterra integro-differential equations by the Tau 21. Ku¨rkc¸u¨,O., Aslan, E., Sezer, M.: A numerical approach with method with an error estimation. Appl. Math. Comput. 167, error estimation to solve general integro-differential-difference 1418–1429 (2005) equations using Dickson polynomials. Appl. Math. Comput. 276, 32. Stoll, T., Tichy, R.: Diophantine equations for Morgan–Voyce 324–339 (2016) and other modified orthogonal polynomials. Math. Slovaca 58, 22. Ku¨rkc¸u¨,O., ASLAN, E., Sezer, M.: A novel collocation method 11–18 (2008) based on residual error analysis for solving integro-differential 33. Swamy, M.N.S.: Further properties of Morgan–Voyce polyno- equations using hybrid dickson and taylor polynomials. Sains mials. Int. J. Comput. Math. 72, 491–507 (1968) Malays. 46, 335–347 (2017) 34. Wang, H.: Positive periodic solutions of functional differential 23. Le, Anh Minh, Chau, Dang Dinh: Asymptotic equilibrium of equations. J. Differ. Equ. 202, 354–366 (2004) integro-differential equations with infinite delay. Math. Sci. 9, 35. Yu¨zbas¸ı,S¸ ., Sezer, M.: An exponential approximation for solu- 189–192 (2015) tions of generalized pantograph-delay differential equations. 24. Mollaog˘lu, T., Sezer, M.: A numerical approach with residual Appl. Math. Model. 37, 9160–9173 (2013) error estimation for eolution of high-order linear differential- 36. Yu¨zbas¸ı,S¸., S¸ ahin, N., Sezer, M.: A Bessel polynomial approach difference equations by using gegenbauer polynomials. Sains for solving linear neutral delay differential equations with vari- Malays. 46, 335–347 (2017) able coefficients. J. Adv. Res. Differ. Equ. 3, 81–101 (2011) 25. Nouri, K., Torkzadeh, L., Mohammadian, S.: Hybrid Legendre 37. Zhang, Bo: Fixed points and stability in differential equations functions to solve differential equations with fractional deriva- with variable delays. Nonlinear Anal. Theory Methods Appl. 63, tives. Math. Sci. (2018). https://doi.org/10.1007/s40096-018- 233–242 (2005) 0251-7 38. Zhao, D.: New criteria for stability of neutral differential equa- 26. Oguz, C., Sezer, M.: Chelyshkov collocation method for a class tions with variable delays by fixed points method. Adv. Differ. of mixed functional integro-differential equations. Appl. Math. Equ. 2011, 48 (2011) Comput. 259, 943–954 (2015) 27. Olach, R.: Positive periodic solutions of delay differential equa- tions. Appl. Math. Lett. 26, 1141–1145 (2013) Publisher’s Note 28. Oliveira, F.A.: Collocation and residual correction. Numer. Math. Springer Nature remains neutral with regard to jurisdictional claims in 36, 27–31 (1980) published maps and institutional affiliations. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Mathematical Sciences Springer Journals

A numerical approach for a nonhomogeneous differential equation with variable delays

Free
11 pages

Loading next page...
 
/lp/springer_journal/a-numerical-approach-for-a-nonhomogeneous-differential-equation-with-rCGJShZpWn
Publisher
Springer Journals
Copyright
Copyright © 2018 by The Author(s)
Subject
Mathematics; Applications of Mathematics
ISSN
2008-1359
eISSN
2251-7456
D.O.I.
10.1007/s40096-018-0253-5
Publisher site
See Article on Publisher Site

Abstract

In this study, we consider a linear nonhomogeneous differential equation with variable coefficients and variable delays and present a novel matrix-collocation method based on Morgan–Voyce polynomials to obtain the approximate solutions under the initial conditions. The method reduces the equation with variable delays to a matrix equation with unknown Morgan– Voyce coefficients. Thereby, the solution is obtained in terms of Morgan–Voyce polynomials. In addition, two test problems together with error analysis are performed to illustrate the accuracy and applicability of the method; the obtained results are scrutinized and interpreted by means of tables and figures. Keywords Morgan–Voyce polynomials  Matrix method  Collocation method  Delay differential equation Variable delay Introduction including production of blood cells [1, 14, 23, 25, 27, 28, 34, 37]. In this paper, we consider nonhomogeneous differential In the case of bounded delays, many authors using equation with variable delays in the form [3, 5, 10, 12, standard techniques [3, 10, 20, 27, 34, 37] have studied the 23, 30, 37, 38]. asymptotic behavior of solutions, the asymptotic stability m in equations and the existence of positive periodic solutions y ðÞ t ¼ P ðtÞþ P ðtÞyðtÞþ P ðtÞyt  s ðtÞ of delay equations. However, most of the mentioned type o 1 j j ð1Þ j¼2 delay equations have not analytical and numerical solu- tions; therefore, numerical methods are required to obtain under the initial condition yðaÞ¼ k, where the coefficients approximate solutions. For this purpose, by means of the P ðtÞ and the delays s are continuous functions on the j j matrix method based on collocation points which have interval 0  a  j  b and the delays are nonnegative, been given by Sezer and coworkers [2, 6, 16, 17, 21, s ðÞ t  0 for t  a. 26, 29, 36], we develop a novel matrix technique to find the Delay differential equations of the type 1 arise in a approximate solution of Eq. 1 under the initial condition variety of applications including control systems, electro- yðaÞ¼ k in the truncated Morgan–Voyce series form dynamics, mixing liquids, neutron transportation, popula- tion models, physiological processes and conditions ytðÞ ffi y ðÞ t ¼ y b ðtÞ; a  t  b ð2Þ N n n n¼0 & Mustafa Ozel where y ; n ¼ 0; 1; ...; N are coefficients to be determined; mustafa.ozel@deu.edu.tr n b ; n ¼ 0; 1; ...; N are the first kind Morgan–Voyce poly- Department of Geophysical Engineering, Faculty of nomials defined by [11, 32, 33] Engineering, Dokuz Eylul University, Tınaztepe Campus, Buca, 35160 Izmir, Turkey n þ j b ðÞ t ¼ t ; n 2 N; a  t  b n ð3Þ Department of Physics, Faculty of Science, Dokuz Eylul n  j j¼0 University, Tınaztepe Campus, Buca, 35160 Izmir, Turkey Department of Mathematics, Faculty of Art and Science, Celal Bayar University, Manisa, Turkey 123 146 Mathematical Sciences (2018) 12:145–155 and Here, the set of polynomialsfg b ðÞ t has the following 2 3 0 1 2 N properties [11, 15, 19, 32] 6 7 0 1 2 N 6 7 6    7 1. The polynomials b ðÞ t defined by 3 are recursively n 6 2 3 N þ 1 7 6 7 6 7 given by the relation 0 1 N  1 6 7 6   7 6 4 N þ 2 7 M ¼ 6 7 b ðÞ t ¼ðÞ t þ 2 b ðÞ t  b ðÞ t ; n  2 n n1 n2 6 7 0 N  2 6 7 6 7 . . . . 6 7 . . . . with b ðÞ t ¼ 1 and b ðÞ t ¼ t þ 1. 6 . . . . 7 o n 6  7 4 5 2N 2. The polynomials y ¼ b ðÞ t ; n ¼ 0; 1; ... are solutions n 00 0 of the differential equation 00 0 Besides, the relation between the matrix and its derivative ttðÞ þ 4 y þ 2ðÞ t þ 1 y  nnðÞ þ 1 y ¼ 0: X ðÞ t can be written in the form [13, 18, 22, 24] 3. The first four Morgan–Voyce polynomials of the first X ðÞ t ¼ XtðÞT ð6Þ kind are obtained from 3 as where b ðÞ t ¼ 1; b ðÞ t ¼ t þ 1; b ðÞ t ¼ t þ 3t þ 1; o 1 2 2 3 3 2 01 0 ... 0 b ðÞ t ¼ t þ 5t þ 6t þ 1; ... 6 7 00 2 ... 0 6 7 6 7 . . . . 6 7 . . . . T ¼ : . . . . 6 7 Fundamental matrix relations 6 7 4 5 00 0 ... N 00 0 ... 0 In this section, we compose the matrix relations of Eq. 1 and its solution Eq. 2. For this aim, we first write the Then, by means of the matrix relations Eqs. 4, 5, and 6,we matrix form of the finite Morgan–Voyce series Eq. 2 as obtain ytðÞ ffi y ðÞ t ¼ btðÞY ð4Þ ytðÞ ffi y ðtÞ¼ btðÞY ð7Þ so that ¼ XtðÞMY btðÞ ¼½ b ðÞ t ; b ðÞ t ; .. .; b ðÞ t o 1 N and 0 0 0 Y ¼½ y ; y ; ...; y ; o 1 N y ðÞ t ffi y ðtÞ¼ X ðÞ t MY ð8Þ then, by using the Morgan–Voyce polynomials Eq. 3,we ¼ XtðÞTMY obtain the matrix form btðÞ as follows By putting t ! t  s ðÞ t in Eq. 7, we gain the recurrence btðÞ ¼ XtðÞM ð5Þ relation [13, 18, 22, 24] where yt  s ðtÞ ffi y ðtÞ¼ Xt  s ðtÞ MY j N j ð9Þ 2 N ¼ XtðÞL s ðtÞ MY XtðÞ ¼ 1; t; t ; ...; t so that 2    3 0 1 2 N 0 1 2 N s ðÞ t s ðÞ t s ðÞ t  s ðÞ t j j j j 6 7 0 0 0 0 6 7 6    7 6    7 1 2 N 0 1 N1 6 7 0 s ðÞ t s ðÞ t  s ðÞ t j j j 6 7 1 1 1 6 7 6 7 6 2 N 7 0 N2 L s ðÞ t ¼ 6 7 00 s ðÞ t  s ðÞ t j j 6 7 2 2 6 7 6 7 . . . . 6 7 . . . . 6 7 . . . . 6  7 4 N  5 000  s ðÞ t 123 Mathematical Sciences (2018) 12:145–155 147 Note that the matrix Xt  s ðtÞ can be written as U ¼ XðaÞM ¼½ u ; u ; ...; u 00 01 0N Xt  s ðtÞ ¼ XtðÞL s ðtÞ j j Consequently, in order to get the approximate solution of Eq. 1 subject to yðaÞ¼ k, we replace the row matrix in By substituting the relations Eqs. 7, 8, and 9 into Eq. 1,we Eq. 13 by the last row(or any row) of the augmented matrix have the matrix equation in Eq. 12; then, we obtain the result matrix "# ~ ~ ~ ~ W; P , WY ¼ P ð14Þ o o XtðÞT  P ðÞ t XtðÞ  P ðÞ t XtðÞL s ðÞ t MY 1 j j j¼2 ~ ~ ~ If rank W ¼ rank W; P ¼ N þ 1, then we can write, ¼ P ðÞ t ð10Þ ~ ~ Y ¼ W P . Thus the matrix, Y (thereby the Morgan– and by placing the collocation points defined by Voyce coefficients y ; y ; ...; y ) is uniquely determined; o 1 N thus Eq. 1 has a unique solution. b  a t ¼ a þ i; i ¼ 0; 1; ...; N: in Eq. 10, the compact form of the obtained matrix equa- Error analysis tions system "# In this section, an error analysis will be presented for the XT  P X  P ðÞ t XL s MY ¼ P ð11Þ 1 j j o Morgan–Voyce polynomial solution in Eq. 16 with the j¼2 residual error function [4, 8, 9, 13, 21, 22, 28, 31, 36]. where In addition, we will improve the Morgan–Voyce poly- 2 3 2 3 2 N nomial solution y ðÞ t with the aid of the residual error XtðÞ 1 t t  t o o o o function. 2 N 6 7 6 7 XtðÞ 1 t t  t 1 1 6 7 6 1 1 7 6 7 6 7 Firstly, we consider the operator Eq. 1, under the initial X ¼ ¼ ; 6 . 7 6 . . . . 7 . . . . . 4 5 4 5 condition yðaÞ¼ k, . . . . 2 N XtðÞ 1 t t  t N N N N Ly½ ðÞt ¼ gðtÞ 2 3 2 3 L sðÞ t P ðt Þ o o j o 6 7 6 7 Ly½ ðtÞ ¼ y ðÞ t  P ðtÞ P ðtÞyðtÞ P ðtÞyt  s ðtÞ : 6 L sðÞ t 7 P ðt Þ o 1 j j o 1 j 1 6 7 6 7 6 7 j¼2 P ¼ ; L s ¼ 6 7 o j 6 7 . . 6 7 4 5 4 5 ð15Þ P ðt Þ L sðÞ t o N j N Here, y ðtÞ is the approximate solution of the problem and P ¼ diag P ðt Þ; P ðt Þ; ...; P ðt Þ ; j ¼ 1; 2; ...; m o j o j 1 j N satisfies the problem X ¼ diag½ Xðt Þ; Xðt Þ; ...; Xðt Þ ; o 1 N L½y ðtÞ ¼ gðtÞþ R ðtÞ; a  t  b; N N ð16Þ y ðtÞ¼ k: Morgan–Voyce matrix method Also, the residual function of the Morgan–Voyce polyno- mial approximation y ðÞ t is defined as The fundamental matrix Eq. 11 of Eq. 1 can be expressed in the form R ðtÞ¼ L½y ðtÞ  gðtÞ: ð17Þ N N WY ¼ P ,½ W; P ð12Þ o o If we know the exact solution ytðÞ, then the error function where is calculated as the difference between the approximate and "# the exact solutions defined by W ¼ w ¼ XT  P X  P ðÞ t XL s M pq 1 j j e ðÞ t ¼ ytðÞ y ðÞ t : ð18Þ N N j¼2 By using the Eqs. 15, 16, 17 and 18, we get the error p; q ¼ 0; 1; ...; N: problem By using the relation Eq. 7, we obtain the corresponding L½e ðtÞ ¼ L½yðtÞ  L½y ðtÞ ¼ R ðtÞ: N N N matrix form to the initial condition yðaÞ¼ k as ð19Þ e ðÞ a ¼ 0: UY ¼ k ,½ U; Y ð13Þ such that 123 148 Mathematical Sciences (2018) 12:145–155 By solving the error problem in Eq. 19 with the method Numerical examples presented in Sect. 3, we get the approximation e ðtÞ to N;M e ðtÞ as follows N Example 1 Consider the differential equation with vari- able delay t þ 1 e ðtÞ¼ a b ðtÞ; ðM  NÞ: N;M n 0 2 2 2 2 y ðtÞ¼3t  t þ t yðtÞ yðt  t Þþ tyðt  t  1Þ n¼0 ð20Þ 0  t  1 Consequently, by means of the polynomials y ðtÞ and e ðtÞ, ðM  NÞ, we obtain the corrected Morgan–Voyce N;M polynomial solution y ðtÞ¼ y ðtÞþ e ðtÞ. Here, N;M N N;M subject to the initial condition yð0Þ¼1. The exact e ðtÞ¼ yðtÞ y ðtÞ, E ðtÞ¼ e ðtÞ e ðtÞ¼ N N N;M N N;M solution of this equation is yðtÞ¼ t  1. First of all, let us yðtÞ y ðtÞ and e ðtÞ denote the error function, the N;M N;M determine the collocation points by the formula t ¼ a þ corrected error function and the estimated error function, ba i; i ¼ 0; 1; ...; N for a ¼ 0, b ¼ 1, and m ¼ 3; N ¼ 2. respectively. Therefore the collocation points are obtained as If the exact solution of Eq. 1 can not been known, then t ¼ 0; t ¼ ; t ¼ 1. By Eq. 11, the fundamental matrix o 1 2 the absolute errorsjj e ðt Þ ¼jj yðt Þ y ðt Þ ,(a  t  b) 2 N i i N i i equation of this problem is written as are not computed. However, the absolute errors "# jj e ðt Þ ¼jj yðt Þ y ðt Þ ,(a  t  b) can be estimated by 3 N i i N i i X XT  P X  P ðÞ t XL s MY ¼ P 1 j j o using the absolute error function e ðtÞ . N;M j¼2 where 2 3 2 3 2 3 2 3 0 00 0 000 10 0 6 7 5 1 1 6 7 6 7 6 7 6 7 P ¼ ; P ¼ 0 0 ; P ¼ 0 10 ; P ¼ 0 0 4 5 4 5 4 5 o 1 2 3 4 5 4 4 2 00 1 2 00 1 001 XtðÞ ¼ 1 tt 2 3 2 3 2 3 2 3 XðÞ 0 10 0 010 111 6 7 1 1 1 6 7 6 7 6 7 6 7 T ¼4002 5; M ¼ 4013 5; X ¼ 6 X 7 ¼ 4 1 5 4 2 5 2 4 000 001 11 1 XðÞ 1 2 3 2 3 XðÞ 0 00 10 000 00 00 6 7 1 1 1 6 7 6 7 X ¼6 0 X 0 7 ¼ 400 01 00 0 5 4 2 5 2 4 00 000 01 11 00 XðÞ 1 2 3 2 3 100 1 11 6 7 6 7 010 01 2 6 7 6 7 6 7 6 7 6 7 6 7 001 00 1 6 7 6 7 6 1 1 7 6 5 25 7 6 7 6 7 1  1 6 7 6 7 4 16 4 16 6 7 6 7 6 1 7  6 5 7 LðÞ s ðt Þ¼ ; LðÞ s ðt Þ¼ : 2 i 3 i 6 7 6 7 01  01 6 7 6 7 2 2 6 7 6 7 6 001 7 6 7 00 1 6 7 6 7 6 7 6 7 1 11 1 24 6 7 6 7 6 7 6 7 4 5 4 5 01 2 01 4 001 00 1 123 Mathematical Sciences (2018) 12:145–155 149 Table 1 Numerical results of t N ¼ 3; M ¼ 4 N ¼ 4; M ¼ 5 N ¼ 12; M ¼ 13 the exact, approximate and corrected solutions of Exact Appr. Corrected Appr. Corrected Appr. Corrected Example 2 for some N values 0 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 0.1 1.22140 1.39581 1.20219 1.20219 1.22571 1.22140 1.22140 0.2 1.49182 1.75739 1.46663 1.46663 1.49689 1.49182 1.49182 0.3 1.82212 2.11993 1.79678 1.79678 1.82697 1.82212 1.82212 0.4 2.22554 2.51864 2.20151 2.20151 2.23024 2.22554 2.22554 0.5 2.71828 2.98872 2.69510 2.69510 2.72299 2.71828 2.71828 0.6 3.32012 3.56539 3.29724 3.29724 3.32478 3.32012 3.32012 0.7 4.05520 4.28383 4.03301 4.03301 4.05954 4.05520 4.05520 0.8 4.95303 5.17927 4.93292 4.93292 4.95683 4.95303 4.95303 0.9 6.04965 6.28689 6.03286 6.03286 6.05299 6.04965 6.04965 1 7.38906 7.64191 7.37414 7.37414 7.39236 7.38906 7.38906 Fig. 1 Exact, approximate and 9 corrected solutions of Example 2 for N ¼ 3 Exact Approximation Corrected 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Fig. 2 Exact, approximate and 9 corrected solutions of Example 2 for N ¼ 4 Exact Approximation Corrected 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 2 3 The augmented matrix of the fundamental matrix equation 12 4 0 is computed as 6 7 ½ W; P ¼ 4 0:25 1:75 5:46875 1:25 5 10 2 1 y(t) y(t) 150 Mathematical Sciences (2018) 12:145–155 Table 2 Comparison of the t N ¼ 3; M ¼ 4 N ¼ 4; M ¼ 5 N ¼ 12; M ¼ 13 absolute errors and corrected absolute errors of Example 2 Appr. Corrected Appr. Corrected Appr. Corrected 0 7.99361e-15 6.21725e-15 3.10862e-15 1.02141e-14 2.02061e-14 4.97380e-14 0.1 1.74411e-01 1.92085e-02 1.92085e-02 4.30490e-03 5.60576e-11 2.89457e-12 0.2 2.65564e-01 2.51946e-02 2.51946e-02 5.06186e-03 5.52192e-11 2.88503e-12 0.3 2.97810e-01 2.53429e-02 2.53429e-02 4.85123e-03 5.51801e-11 2.87947e-12 0.4 2.93098e-01 2.40357e-02 2.40357e-02 4.69433e-03 5.49143e-11 2.87148e-12 0.5 2.70441e-01 2.31845e-02 2.31845e-02 4.71136e-03 5.42593e-11 2.86038e-12 0.6 2.45269e-01 2.28805e-02 2.28805e-02 4.66159e-03 5.29470e-11 2.80753e-12 0.7 2.28632e-01 2.21878e-02 2.21878e-02 4.33963e-03 5.04476e-11 2.68496e-12 0.8 2.26233e-01 2.01130e-02 2.01130e-02 3.79602e-03 4.58762e-11 2.45759e-12 0.9 2.37243e-01 1.67888e-02 1.67888e-02 3.34318e-03 3.76135e-11 2.08455e-12 1.0 2.52855e-01 1.49208e-02 1.49208e-02 3.29924e-03 2.63132e-11 1.53477e-12 Fig. 3 Comparison of the 0.35 absolute error with the corrected 0.3 absolute error for N ¼ 3 0.25 0.2 Absolute Error 0.15 Corrected Absolute Error 0.1 0.05 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Fig. 4 Comparison of the 0.05 absolute error with the corrected Absolute Error absolute error for N ¼ 4 0.04 Corrected Absolute Error 0.03 0.02 0.01 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 and the augmented matrix for initial condition is obtained Example 2 Consider the differential equation with vari- as able delay t 0 2t 2 2t ½ U; k¼½ 111 j  1 : 2y ðtÞ tyðtÞþ te yðt  t Þ¼ 4e ; 0  t  1 By using the procedure in Sect. 3, we obtain the approxi- subject to the initial condition yð0Þ¼ 1. The exact solution 2t mate solution as of this equation is yðtÞ¼ e . The fundamental matrix equation is y ðtÞ¼ t  1 ½ XT  P X  P XLðÞ s ðÞ t MY ¼ P : 1 2 2 i o which is the exact solution. e( t) e( t) Mathematical Sciences (2018) 12:145–155 151 Table 3 Numerical results of t N ¼ 3; M ¼ 4 N ¼ 4; M ¼ 5 N ¼ 12; M ¼ 13 the exact, approximate and corrected solutions of Exact Appr. Corrected Appr. Corrected Appr. Corrected Example 3 for some N values 0 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 0.1 0.90484 0.90675 0.90500 0.90500 0.90485 0.90484 0.90484 0.2 0.81873 0.82159 0.81894 0.81894 0.81874 0.81873 0.81873 0.3 0.74082 0.74401 0.74103 0.74103 0.74083 0.74082 0.74082 0.4 0.67032 0.67350 0.67052 0.67052 0.67033 0.67032 0.67032 0.5 0.60653 0.60956 0.60672 0.60672 0.60654 0.60653 0.60653 0.6 0.54881 0.55168 0.54900 0.54900 0.54882 0.54881 0.54881 0.7 0.49659 0.49936 0.49677 0.49677 0.49660 0.49659 0.49659 0.8 0.44933 0.45209 0.44950 0.44950 0.44934 0.44933 0.44933 0.9 0.40657 0.40937 0.40673 0.40673 0.40658 0.40657 0.40657 1 0.36788 0.37069 0.36803 0.36803 0.36789 0.36788 0.36788 Fig. 5 Exact, approximate and 1 corrected solutions of Exact 0.9 Example 3 for N ¼ 3 Approximation 0.8 Corrected 0.7 0.6 0.5 0.4 0.3 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Fig. 6 Exact, approximate and corrected solutions of Exact 0.9 Example 3 for N ¼ 4 Approximation 0.8 Corrected 0.7 0.6 0.5 0.4 0.3 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 After the collocation points substituted into this matrix The absolute and corrected errors of Example 2 in equation, we solve the system and we obtain the solutions Table 2 are compared for N ¼ 3; 4 in Figs. 3 and 4, in the form 4 of Example 2 for N ¼ 3; 4; 12 in the interval respectively. ½ 0; 1 Table 1. Example 3 Consider the delay differential equation having Now, we will give the exact, approximation, and cor- variable delays lnðÞ t þ 1 and t rected solutions of Example 2 for N ¼ 3; 4 in Figs. 1 and 2, respectively. y(t) y(t) 152 Mathematical Sciences (2018) 12:145–155 Table 4 Comparison of the t N ¼ 3; M ¼ 4 N ¼ 4; M ¼ 5 N ¼ 12; M ¼ 13 absolute errors and corrected absolute errors of Example 3 Appr. Corrected Appr. Corrected Appr. Corrected 0 0 0 2.22045e-16 2.22045e-16 4.44089e-16 4.44089e-16 0.1 1.91653e-03 1.57584e-04 1.57584e-04 1.15985e-05 1.16573e-13 2.98650e-14 0.2 2.86131e-03 2.06131e-04 2.06131e-04 1.35662e-05 1.13909e-13 2.95319e-14 0.3 3.19084e-03 2.07981e-04 2.07981e-04 1.30496e-05 1.12577e-13 2.89768e-14 0.4 3.17959e-03 1.97746e-04 1.97746e-04 1.25881e-05 1.09690e-13 2.83107e-14 0.5 3.02786e-03 1.90112e-04 1.90112e-04 1.24366e-05 1.05804e-13 2.74225e-14 0.6 2.86876e-03 1.86905e-04 1.86905e-04 1.21434e-05 1.01363e-13 2.64233e-14 0.7 2.77470e-03 1.83479e-04 1.83479e-04 1.14580e-05 9.62008e-14 2.50355e-14 0.8 2.76306e-03 1.74494e-04 1.74494e-04 1.06297e-05 9.02611e-14 2.35367e-14 0.9 2.80153e-03 1.59153e-04 1.59153e-04 1.01563e-05 8.32667e-14 2.20934e-14 1 2.81276e-03 1.45931e-04 1.45931e-04 1.00352e-05 7.82152e-14 1.96509e-14 -3 Fig. 7 Comparison of the x 10 absolute error with the corrected 3.5 absolute error for N ¼ 3 2.5 1.5 Absolute Error Corrected Absolute Error 0.5 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 -4 Fig. 8 Comparison of the x 10 absolute error with the corrected absolute error for N ¼ 4 Absolute Error Corrected Absolute Error 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 2 t t 2 ½ XT  P X  P XLðÞ s ðÞ t  P XLðÞ s ðÞ t MY ¼ P : y ðtÞ¼ t þ t  1 e  tyðÞ t  lnðÞ t þ 1 e yt  t 1 2 2 i 3 3 i o þ yðtÞ; 0  t  1 subject to the initial condition yð0Þ¼ 1. The exact solution After the collocation points substituted into this matrix of this equation is yðtÞ¼ e . The fundamental matrix equation, we can solve the system and we obtain the equation is e(t) e(t) Mathematical Sciences (2018) 12:145–155 153 Table 5 Numerical results of t Exact Present Abs. error HTL Abs. error the Exact, Present and HTL solutions of Example 4 for N ¼ 0.0 1.0000000 1.0000000 5.55112e-15 1.000000067 6.70000e-08 6 value 0.2 0.8187308 0.8187339 3.10453e-06 0.807836989 1.08938e-02 0.4 0.6703200 0.6703221 2.03288e-06 0.649394632 2.09254e-02 0.6 0.5488116 0.5488127 1.09160e-06 0.529079726 1.97319e-02 0.8 0.4493290 0.4493294 4.48652e-07 0.436772222 1.25567e-02 1.0 0.3678794 0.3678793 1.14014e-07 0.361610346 6.26909e-03 1.2 0.3011942 0.3011936 5.72331e-07 0.298780813 2.41340e-03 1.4 0.2465970 0.2465960 9.17880e-07 0.249314192 2.71723e-03 1.6 0.2018965 0.2018953 1.22873e-06 0.212885438 1.09889e-02 1.8 0.1652989 0.1652975 1.43489e-06 0.173619567 8.32068e-03 2.0 0.1353353 0.1353338 1.44145e-06 0.078902496 5.64328e-02 Fig. 9 Comparison of the exact, 1.2 present and HTL solutions of Exact Example 4 for N ¼ 6 value Present 1.0 HTL 0.8 0.6 0.4 0.2 0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 Fig. 10 RMSE values for N ¼ 1 1; 2; .. .; 100 for Examples 2, 3 Example 2 -2 and 4 Example 3 Example 4 -4 -6 -8 -10 -12 -14 -16 0 102030405060708090 100 y(t) RESM 154 Mathematical Sciences (2018) 12:145–155 creativecommons.org/licenses/by/4.0/), which permits unrestricted solutions in the form 4 of Example 3 for N ¼ 3; 4; 12 in use, distribution, and reproduction in any medium, provided you give the interval½ 0; 1 Table 3. appropriate credit to the original author(s) and the source, provide a All exact, approximation, and corrected solutions of link to the Creative Commons license, and indicate if changes were Example 3 for N ¼ 3; 4 are given in Fig. 5 and 6, respec- made. tively. The absolute and corrected errors of Example 3 in Table 4 are compared for N ¼ 3; 4 in Figs. 7 and 8, References respectively. 1. Abd-Elhameed, W.M., Youssri, Y.H., Doha, E.H.: A novel operational matrix method based on shifted Legendre polyno- Example 4 Consider the following delay differential 2 mials for solving second-order boundary value problems equation [7] having variable delay lnðÞ t þ 1 involving singular, singularly perturbed and Bratu-type equa- 0 2 t 2 tions. Math. Sci. 9, 93–102 (2015) y ðtÞ¼ t þ 1 e yt  ln t þ 1  yðtÞ; 0  t  2 2. Akyu¨z, A., Sezer, M.: A Chebyshev collocation method for the solution of linear integro-differential equations. Int. J. Comput. subject to the initial condition yð0Þ¼ 1. The exact solution Math. 72, 491–507 (1999) of this equation is yðtÞ¼ e . 3. Ardjouni, A., Djoudi, A.: Fixed points and stability in linear neutral differential equations with variable delays. Nonlinear Similarly, we can solve this problem by present method Anal. Theory Methods Appl. 74, 2062–2070 (2011) and we obtain the solutions in the form 4 of Example 4 for 4. Bahs¸ı, M.M., C ¸ evik, M., Sezer, M.: Orthoexponential polynomial N ¼ 6 in the interval½ 0; 2 . Then, we also compare the solutions of delay pantograph differential equations with residual present solutions and Hybrid Taylor-Lucas Method [7] error estimation. Appl. Math. Comput. 271, 11–21 (2015) 5. Bahs¸ı, M.Mustafa, Bahs¸ı, Ays¸e Kurt, C ¸ evik, Mehmet, Sezer, (HTL) solutions in Table 5. The present method has been Mehmet: Improved Jacobi matrix method for the numerical shown to be suitable as graphically for nonhomogeneous solution of Fredholm integro-differential-difference equations. differential equation with variable delays in Fig. 9. Math. Sci. 10, 83–93 (2016) In order to better define the solution space of the prob- 6. Balcı, M.A., Sezer, M.: Hybrid Euler-Taylor matrix method for solving of generalized linear Fredholm integro-differential dif- lems of Examples 2, 3 and 4 described above, the N value ference equations. Appl. Math. Comput. 273, 33–41 (2016) has been scanned up 1 to 100. The root-mean-square error 7. Bayku, N., Sezer, M.: Hybrid Taylor–Lucas collocation method (RMSE) value of the solution obtained for each N value is for numerical solution of high-order pantograph type delay dif- calculated and shown in Fig. 10. When N [ 12, the RMSE ferential equations with variables delays. Appl. Math. Inf. Sci. 11, 1795–1801 (2017) values of the solutions of both problems are oscillated 8. C ¸ elik, I.: Approximate calculation of eigenvalues with the 10 16 between 10 and 10 . The reason of these oscillations method of weighted residuals-collocation method. Appl. Math. is the truncation errors in the calculations. Comput. 160, 401–410 (2005) 9. C ¸ elik, I.: Collocation method and residual correction using Chebyshev series. Appl. Math. Comput. 174, 910–920 (2006) 10. Dix, J.G.: Asymptotic behavior of solutions to a first-order dif- Conclusion ferential equation with variable delays. Comput. Math. Appl. 50, 1791–1800 (2005) 11. Djordjevic, G.B., Milovanovic, G.V.: Special classes of polyno- A new approach using the Morgan–Voyce polynomials to mials. Univ. Nis Fac. Technol. Leskovac 58, 11–18 (2014) solve numerically the first-order nonhomogeneous differ- 12. Elahi, Zaffer, Akram, Ghazala, Siddiqi, Shahid Saeed: Numerical ential equations with variable delays is presented in this solution for solving special eighth-order linear boundary value study. An error analysis technique based on residual problems using Legendre Galerkin method. Math. Sci. 10, 201–209 (2016) function is also developed our problems. If the exact 13. Erdem, K., Yalc¸inbas¸, S., Sezer, M.: A Bernoulli polynomial solution of the problem is not known, by using this tech- approach with residual correction for solving mixed linear nique it is possible to estimate the error function and also to Fredholm integro-differential-difference equations. J. Differ. reduce the error due to the residual function. It is seen that, Equ. Appl. 19, 1619–1631 (2013) 14. Erfanian, M., Zeidabadi, H.: Using of Bernstein spectral Galerkin the accuracy improves, when N is increased. To compute method for solving of weakly singular Volterra–Fredholm inte- our solutions and error functions, we have written a code in gral equations. Math. Sci. (2018). https://doi.org/10.1007/s40096- Matlab and calculated all computations by means of this 018-0249-1 code. 15. Fortuna, L., Frasca, M.: Generating passive systems from recur- Consequently, the present method has been shown to be sively defined polynomials. Int. J. Circuits Syst. Signal Process. 6, 179–188 (2012) convenient, reliable and effective for solving the first-order 16. Gu¨lsu, M., Sezer, M.: Approximations to the solution of linear nonhomogeneous differential equation with variable Fredholm integrodifferential-difference equation of high order. delays. J. Franklin Inst. 343, 720–737 (2006) 17. Gu¨rbu¨z, B., Sezer, M., Gu¨ler, C.: Laguerre collocation method for solving Fredholm integro-differential equations with functional Open Access This article is distributed under the terms of the Creative arguments. J. Appl. Math. 2014, 48 (2014) Commons Attribution 4.0 International License (http:// 123 Mathematical Sciences (2018) 12:145–155 155 18. Is¸ik, O.R., Gu¨ney, Z., Sezer, M.: Bernstein series solutions of 29. Sezer, M., Akyu¨z-Das¸cıoglu, A.: A Taylor method for numerical pantograph equations using polynomial interpolation. J. Differ. solution of generalized pantograph equations with linear func- Equ. Appl. 18, 357–374 (2012) tional argument. J. Comput. Appl. Math. 200, 217–225 (2007) 19. Ilhan, O., Sahin, N.: On Morgan–Voyce polynomials approxi- 30. Singh, Randhir, Wazwaz, Abdul-Majid: Numerical solutions of mation for linear differential equations. Int. J. Comput. Math. 72, fourth-order Volterra integro-differential equations by the Greens 491–507 (2014) function and decomposition method. Math. Sci. 10, 159–166 20. Jin, C., Luo, J.: Fixed points and stability in neutral differential (2016) equations with variable delays. Proc. Am. Math. Soc. 136, 31. Shahmorad, S.: Numerical solution of the general form linear 909–918 (2008) Fredholm-Volterra integro-differential equations by the Tau 21. Ku¨rkc¸u¨,O., Aslan, E., Sezer, M.: A numerical approach with method with an error estimation. Appl. Math. Comput. 167, error estimation to solve general integro-differential-difference 1418–1429 (2005) equations using Dickson polynomials. Appl. Math. Comput. 276, 32. Stoll, T., Tichy, R.: Diophantine equations for Morgan–Voyce 324–339 (2016) and other modified orthogonal polynomials. Math. Slovaca 58, 22. Ku¨rkc¸u¨,O., ASLAN, E., Sezer, M.: A novel collocation method 11–18 (2008) based on residual error analysis for solving integro-differential 33. Swamy, M.N.S.: Further properties of Morgan–Voyce polyno- equations using hybrid dickson and taylor polynomials. Sains mials. Int. J. Comput. Math. 72, 491–507 (1968) Malays. 46, 335–347 (2017) 34. Wang, H.: Positive periodic solutions of functional differential 23. Le, Anh Minh, Chau, Dang Dinh: Asymptotic equilibrium of equations. J. Differ. Equ. 202, 354–366 (2004) integro-differential equations with infinite delay. Math. Sci. 9, 35. Yu¨zbas¸ı,S¸ ., Sezer, M.: An exponential approximation for solu- 189–192 (2015) tions of generalized pantograph-delay differential equations. 24. Mollaog˘lu, T., Sezer, M.: A numerical approach with residual Appl. Math. Model. 37, 9160–9173 (2013) error estimation for eolution of high-order linear differential- 36. Yu¨zbas¸ı,S¸., S¸ ahin, N., Sezer, M.: A Bessel polynomial approach difference equations by using gegenbauer polynomials. Sains for solving linear neutral delay differential equations with vari- Malays. 46, 335–347 (2017) able coefficients. J. Adv. Res. Differ. Equ. 3, 81–101 (2011) 25. Nouri, K., Torkzadeh, L., Mohammadian, S.: Hybrid Legendre 37. Zhang, Bo: Fixed points and stability in differential equations functions to solve differential equations with fractional deriva- with variable delays. Nonlinear Anal. Theory Methods Appl. 63, tives. Math. Sci. (2018). https://doi.org/10.1007/s40096-018- 233–242 (2005) 0251-7 38. Zhao, D.: New criteria for stability of neutral differential equa- 26. Oguz, C., Sezer, M.: Chelyshkov collocation method for a class tions with variable delays by fixed points method. Adv. Differ. of mixed functional integro-differential equations. Appl. Math. Equ. 2011, 48 (2011) Comput. 259, 943–954 (2015) 27. Olach, R.: Positive periodic solutions of delay differential equa- tions. Appl. Math. Lett. 26, 1141–1145 (2013) Publisher’s Note 28. Oliveira, F.A.: Collocation and residual correction. Numer. Math. Springer Nature remains neutral with regard to jurisdictional claims in 36, 27–31 (1980) published maps and institutional affiliations.

Journal

Mathematical SciencesSpringer Journals

Published: Jun 5, 2018

References

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off