ABSTRACT
A computational technique for analysing coupled system of singularly perturbed convection-diffusion delay differential equations with large delay is presented in this study. We solve this problem using non-polynomial spline technique and arrive at a system that can potentially be solved. To show the theoretical investigation, an example is solved for different perturbation parameter values, and the computational results are shown in tables. The model we have proposed is of first order convergent, and this technique indicated that the acquired findings were more accurate than earlier results.
1. Introduction
A singularly perturbed delay differential equation is one in which the highest order derivative term is multiplied by a positive small parameter and at least one delay or advance term, or both, is involved. The study of singular perturbations has come a long way in mathematics and has important potential applications in many areas of science and technology. Different types of real-world phenomena such as in the modelling of the pupil-light reflex in humans (Longtin and Milton Citation1988), predator-prey model (Gourley and Kuang Citation2004), HIV infection model (Culshaw and Ruan Citation2000), variational problem in control theory (Glizer Citation2000), determination of the behaviour of a neuron to random synaptic input (Lange and Miura Citation1982), optically bistable device (Mayer et al. Citation1995),the dynamics of two identical amplifier networks (Chen and Wu Citation2001), the first exit-time problem (Kot Citation2001), hydrodynamics of liquid helium (Kuang Citation1993), various physiological process models (Mackey and Glass Citation1977) and numerous other applied mathematical problems can be modelled mathematically by singularly perturbed delay differential equations. These differential equations are used to solve problems in which the future is dependent on the present as well as past states of the system being investigated under consideration.
In singular perturbation problems, the convection-diffusion type is referred to when the perturbation parameter equals zero and the order is lowered by one, whereas the reaction-diffusion type is referred to when the order is lowered by two. As a result, second-order singularly perturbed delay differential equations can be either convection-diffusion or reaction-diffusion in nature. In the first type of problem, the left or right-boundary layer is only determined by the sign of the coefficients in the diffusion and convection terms. On the other hand, in reaction-diffusion delay differential equations, there are two boundary layers. As a result, numerous researchers have been working to create various numerical techniques for handling these kinds of problems. When normal numerical methods are used, it is widely known that these types of problems yield inaccurate results for small values of the perturbation parameter. As a result, it is essential to develop numerical techniques that can provide improved accuracy regardless of parameter values. Another crucial aspect of these types of problems is the investigation of the effect of shift parameters on the solution’s boundary layer behaviour. Consequently, there has been a rise in interest in numerical techniques for singularly perturbed delay differential equations in recent years.
In problems involving control systems, the state of the system is observed and adjustments are incorporated to control the outcomes. Due to the impossibility of instantaneous adjustment, there is always a delay between observation and control action. Differential equations with delay arguments govern this type of system. For a single singularly perturbed delay differential equation, robust numerical techniques have been developed by Lange and Miura (Lange and Miura Citation1994) investigated a class of boundary-value problems and discussed an asymptotic method to approximately solve this type of differential equations. Phaneendra and Lalu (Phaneendra and Lalu Citation2019) had carried out a quadrature technique for the solution of this type of problems. They used a neutral delay differential equation which is equivalent to the SPDDE. Chakravarthy and Kumar (Chakravarthy and Kumar Citation2021) proposed Numerov’s method which uses a fitted operator finite difference technique. Sekar and Tamilselvan (Erdogan et al. Citation2020) suggested a piecewise Shishkin type mesh to solve the problem by using finite difference method. Erdogan and Sakar (Woldaregay and Liu Citation2022) suggested a numerical technique based on a piecewise uniform Shishkin mesh with an exponentially fitted difference scheme for each time-subinterval. Woldaregay and Durressa (Sekar and Tamilselvan Citation2019) used Taylor series approximation to approximate the delay terms, and the obtained singularly perturbed boundary value problem is solved with an exponentially fitted operator mid-point upwind finite difference method. But for systems of equations, only a small number of results have been reported. Subburayan and Ramanujam (Subburayan and Ramanujam Citation2014) proposed an asymptotic computational technique for convection diffusion type equations. The authors in (Selvi and Ramanujam Citation2017) applied a computational technique to solve a system of singularly perturbed delay differential equations of large delay that was weakly coupled. Chakravarthy and Gupta (Chakravarthy and Gupta Citation2020) focus on the use of cubic splines in tension approximation techniques to solve system of singularly perturbed convection diffusion equation.
We developed a fitted non-polynomial spline technique for the numerical solution of the system of second-order singularly perturbed delay differential equations with large delay. When the perturbation parameter is small compared to the mesh width used to discretize the given problem, it is well known that the conventional techniques fail. Our goal is to demonstrate that, when epsilon is small or large in relation to step size h, accurate numerical approximations can be obtained using a non-polynomial spline. Here, we take a non-polynomial spline approach using the fitted operator method to solve this kind of problem. In Sect. 2, we define the problem and assumptions on the parameter and stability analysis is discussed in Sect. 3. In Sect. 4 we employ the non-polynomial spline to arrive at its numerical solution by Gauss elimination method. In Sect. 5, we perform the error and convergence analysis of the method. In Sect. 6 we took some examples and compared the results with (Subburayan and Ramanujam Citation2014; Selvi and Ramanujam Citation2017; Chakravarthy and Gupta Citation2020) and their results are shown. Conclusion follows in Sect. 7.
2. Statement of the problem
Consider the following class of system of singularly perturbed delay differential equation on the domain ,
with boundary conditions,
where the perturbation parameter satisfy .
It is also assumed that the functions and are sufficiently smooth functions on together with the following conditions:
and
The differential operator in (1) can be written as:
where the coefficient matrices are given by,
Also the functions . In addition, the boundary conditions and the source term are smooth functions on , where . In this context, refers to the class of times continuously differentiable functions in . Also, the problem (1) and (2) shows a weak boundary layer at and a strong layer at .
3. Stability analysis
Theorem 3.1.
Let the function satisfying and . Then
Proof.
Define , where
Here is positive for all and . Let
Then there exists satisfying or or both and . Without loss of generality, we assume that . Hence attains its minimum. Suppose the theorem does not hold true, then .
Case 1:
Then, for ,
which is a contradiction to our assumption.
Case 2: .
Then, for ,
which is a contradiction to our assumption.
Case 3:
Then, for ,
which is a contradiction to our assumption.
From this, it is clear that in all the three cases we arrived a contradiction. Therefore is not possible. This shows that,
Theorem 3.2.
Let be any function. Then
for all
Proof. Let be a constant. Define , where,
where,
By appropriate choice of , we obtain,
Let , then by appropriate choice of and , we obtain,
Similarly we can prove for in . Therefore in . Also by proper choice of we obtain,
Then by Theorem 3.1, we obtain . Therefore,
for all
4. Derivation of the method
The objective of this section is to introduce the uniform mesh and the numerical approach that is being examined in this article. Let be the discretized domain with mesh – intervals on the domain , where is as an even positive integer. Let be the mesh points obtained while dividing , such that the step size for
4.1. Numerical algorithm
The following algorithm is proposed to obtain the numerical solution:
Step 1: Introduce the uniform mesh by discretizing the domain with mesh intervals.
Step 2: Non-polynomial spline method is applied to the statement problem to obtain the scheme.
Step 3: We use Taylor series approximations for the derivatives in the scheme obtained in Step 2 to arrive at two systems for and .
Step 4: Introduce fitting factor to the schemes obtained in Step 3, and we derive the schemes into a system of equations in different mesh intervals, incorporating history functions.
Step 5: Finally, we employ Gauss elimination method to solve the system obtained in Step 4 by using MATLAB R2022a mathematical software.
4.2. The proposed numerical scheme
The non-polynomial spline is of the following form:
for where the coefficients are unknown and is an arbitrary parameter that will be utilized to improve the methods accuracy.
In order to calculate the coefficients which are unknown in (3), we use the conditions,
to determine and
We get and in (3) as,
where .
Applying the first derivative continuity condition at , , we get,
We obtain the following equation by reducing indices in (4) by one and putting them into (5),
Multiplying with in both sides of (6), we get,
where,
As . We get by applying L’Hospital’s rule. Substituting into (1). The boundary conditions can be written as,
where .
For , we consider the notations,
From Equationequation (1)(1) (1) , we obtain,
Substituting and with and using the following approximations for the first derivative of ,
We obtain the following linear system of equations for and
4.3. Calculation of fitting parameter
An estimate for the solution of the homogeneous problem (1) is of the form:
which is derived from the theory of singular perturbations. Here is the solution of the reduced problem.
We may assume that the coefficients are locally constant because we are examining the differential equations on suitably small subintervals. By applying the Taylor’s series expansion for and about the point and restricting to their first terms, the above Equationequation (9)(9) (9) becomes,
From (10), taking the limit as , and , we obtain,
Substituting these limiting values in (1), we get the fitting parameter,
The scheme in (8) with the fitting factor can be rewritten as:
By incorporating history functions in interval conditions, the above scheme (11) can be rewritten as follows:
and
where,
and,
5. Error and convergence analysis
By Taylor series expansion,
For , the truncation error obtained is
We combine the equations derived from (11) for the system of two equations to arrive at a matrix form as follows:
where is a matrix of order such that,
and all the remaining entries of are zero and is a column vector such that
Let
be the truncation error, where
Then (12) can be written as , where
is the exact solution. It follows that , where
Let denote the row sum of the matrix . Then
It can be easily verified that the matrix is monotone and irreducible when is sufficiently small and hence exists and its elements are non-negative. From which implies
Let be the element of . Then for . This gives
Then we can conclude that, for and
Hence
6. Numerical experiments and discussions
Some numerical experiments are taken into consideration to illustrate the applicability of the proposed method. The tabulated solution of some problems with varying is presented. Our theoretical approaches have been validated by solving the following examples numerically. MATLAB R2022a is utilized to plot the estimated findings. For the problems with no exact solution, maximum absolute errors are determined by applying the double mesh principle to the provided examples.
for where and are the and components of the numerical solutions of meshes of and points, respectively. We compute the uniform error by,
Example 1
The maximum pointwise errors are presented in for different values of and the numerical results compared with the results in (Subburayan and Ramanujam Citation2014; Selvi and Ramanujam Citation2017; Chakravarthy and Gupta Citation2020). shows the maximum absolute errors for different values of . As the step size tends to zero for all values of the perturbation parameter epsilon, the maximum absolute errors decrease with the linear rate of convergence, as shown by the results presented in for different values of with .
The graph of the computed solution of two components for different values of is given in . represents the graphs of maximum absolute error of both components for different values , whereas represents the graphs of point-wise absolute errors for different values of with . The loglog plot of the maximum pointwise error is also given in .
Example 2
The maximum pointwise errors are presented in for different values of and the numerical results are compared with the results in (Selvi and Ramanujam Citation2017). As the step size tends to zero for all values of the perturbation parameter epsilon, the maximum absolute errors decrease with the linear rate of convergence, as shown by the results presented in for different values of with .
The graph of the computed solution of two components for different values of is given in . represents the graphs of maximum absolute error of both components for different values , whereas represents the graphs of point-wise absolute errors for different values of with . The loglog plot of the maximum pointwise error is also given in .
The Rate of Convergence
The computational rate of convergence is also obtained by using the double mesh principle defined as:
7. Conclusion
This work presents an effective numerical approach for handling a system of SPDDEs (convection-diffusion problems) with large delay. To demonstrate the efficacy of this technique, various values of have been applied to some model examples (which do not have an actual solution). It is apparent from the tables that the methodology provides an approximation to the solution. The numerical solutions are summarized in terms of maximum absolute errors , and it is discovered that the discussed method gives an improvement compared to (Subburayan and Ramanujam Citation2014; Selvi and Ramanujam Citation2017; Chakravarthy and Gupta Citation2020). The findings demonstrate convergence of the obtained solution by showing that the maximum absolute errors decline with decreasing grid size and the convergence is of order one with this method. The graphs represent the numerical solution of the problem. represent the rate of convergence .
Acknowledgments
The authors express their sincere thanks to the editor and reviewers for their valuable comments and suggestions that improved the quality of the manuscript.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Data availability statement
No data was used for the research described in the article.
Additional information
Funding
References
- Chakravarthy PP, Gupta T. 2020. Numerical solution of a weakly coupled system of singularly perturbed delay differential equations via cubic spline in tension. Natl Acad Sci Lett. 43:259–262. doi: 10.1007/s40009-019-00806-0.
- Chakravarthy PP, Kumar K. 2021. A novel method for singularly perturbed delay differential equations of reaction-diffusion type. Differ Equ Dyn Syst. 29(3):723–734. doi: 10.1007/s12591-017-0399-x.
- Chen Y, Wu J. 2001. The asymptotic shapes of periodic solutions of a singular delay differential system. J Differ Equ. 169(2):614–632. doi: 10.1006/jdeq.2000.3910.
- Culshaw RV, Ruan S. 2000. A delay-differential equation model of HIV infection of CD4+ T-cells. Math Biosci. 165(1):27–39. doi: 10.1016/S0025-5564(00)00006-7.
- Erdogan F, Sakar MG, Saldır O. 2020. A finite difference method on layer-adapted mesh for singularly perturbed delay differential equations. Appl Math & Nonlin Sci. 5(1):425–436. doi: 10.2478/amns.2020.1.00040.
- Glizer VY. 2000. Asymptotic solution of a boundary-value problem for linear singularly-perturbed functional differential equations arising in optimal control theory. J Optim Theory Appl. 106(2):309–335. doi: 10.1023/A:1004651430364.
- Gourley SA, Kuang Y. 2004. A stage structured predator-prey model and its dependence on maturation delay and death rate. J Math Biol. 49(2):188–200. doi: 10.1007/s00285-004-0278-2.
- Kot M. 2001. Elements of mathematical ecology. Cambridge, United Kingdom: Cambridge University Press.
- Kuang Y. 1993. Delay differential equations: with applications in population dynamics. Cambridge, Massachusetts, United States: Academic press.
- Lange CG, Miura RM. 1982. Singular perturbation analysis of boundary value problems for differential-difference equations. SIAM J Appl Math. 42(3):502–531. doi: 10.1137/0142036.
- Lange CG, Miura RM. 1994. Singular perturbation analysis of boundary-value problems for differential-difference equations. VI. Small shifts with rapid oscillations. SIAM J Appl Math. 54(1):273–283. doi: 10.1137/S0036139992228119.
- Longtin A, Milton JG. 1988. Complex oscillations in the human pupil light reflex with mixed and delayed feedback. Math Biosci. 90(1–2):183–199. doi: 10.1016/0025-5564(88)90064-8.
- Mackey MC, Glass L. 1977. Oscillation and chaos in physiological control systems. Science. 197(4300):287–289. doi: 10.1126/science.267326.
- Mayer H, Zaenker KS, An Der Heiden U. 1995. A basic mathematical model of the immune response. Chaos: An Interdisciplinary J Nonlinear Sci. 5(1):155–161. doi: 10.1063/1.166098.
- Phaneendra K, Lalu M. 2019. Numerical solution of singularly perturbed delay differential equations using gaussion quadrature method. J Phys Conf Ser. 1344(1):012013. doi: 10.1088/1742-6596/1344/1/012013.
- Sekar E, Tamilselvan A. 2019. Singularly perturbed delay differential equations of convection–diffusion type with integral boundary condition. J Appl Math Comput. 59(1–2):701–722. doi: 10.1007/s12190-018-1198-4.
- Selvi PA, Ramanujam N. 2017. An iterative numerical method for a weakly coupled system of singularly perturbed convection–diffusion equations with negative shifts. Int J Appl Comput Math. 3(S1):147–160. doi: 10.1007/s40819-017-0346-0.
- Subburayan V, Ramanujam N. 2014. An asymptotic numerical method for singularly perturbed weakly coupled system of convection–diffusion type differential difference equations. Novi Sad J Math. 44(2):53–68.
- Woldaregay MM, Liu L. 2022. Solving singularly perturbed delay differential equations via fitted mesh and exact difference method. Res Math. 9(1):2109301. doi: 10.1080/27684830.2022.2109301.