key: cord-0660532-5zqkxtms
authors: Gurbuz, Burcu
title: A computational approximation for the solution of retarded functional differential equations and their applications to science and engineering
date: 2021-02-26
journal: nan
DOI: nan
sha: 83ee39a24832bbb36fce51da910692bbad2c36f4
doc_id: 660532
cord_uid: 5zqkxtms

Delay differential equations are of great importance in science, engineering, medicine and biological models. These type of models include time delay phenomena which is helpful for characterising the real-world applications in machine learning, mechanics, economics, electrodynamics and so on. Besides, special classes of functional differential equations have been investigated in many researches. In this study, a numerical investigation of retarded type of these models together with initial conditions are introduced. The technique is based on a polynomial approach along with collocation points which maintains an approximated solutions to the problem. Besides, an error analysis of the approximate solutions is given. Accuracy of the method is shown by the results. Consequently, illustrative examples are considered and detailed analysis of the problem is acquired. Consequently, the future outlook is discussed in conclusion.

In general aspect, functional differential equations (FDEs) describe a wide range of real life phenomena which are extremely valuable in scientific research fields. Specifically, delay differential equations (DDEs) and their applications perform widely in many different branches of medicine, engineering and life sciences. These fields have operations based on time-dependent processes which can be easily described by DDEs [1] , [2] . Economic models are easily described by using FDEs. Some arguments in the continuous systems are helpful to establish control problems and give us better understanding for the market models by using the time-delay contribution in equations. Multiple decision tasks, capable formulations are defined by FDEs with the idea of overlapping in time. Besides, competitive market scenarios are established with regard to the DDE models. In the theory of modern economic growth, continuous-time modeling by FDEs derives an outstanding task. For instance, the Kaldor-Kalecki model, the Cournot oligopoly model and the Goodwin model are also well-known models in economy which are described by the differential equations with time-delay [3] - [7] .

In computer science, FDEs play an essential role. Recently, machine learning studies are very attractive in many applications. Time-delay differential equations establish a 1/19 arXiv:2103.09046v1 [math.NA] 26 Feb 2021 base for machine learning applications. Specifically, reservoir computing has been introduced recently as a machine learning pattern [8] . Its implementations with hardware accomplish processing of empirical data [9] . Besides, in machining processes chatter identification is described by FDEs [10] . In this area, the importance of FDEs has been described in current studies.

In electrodynamics, a device which is called as pantograph exists in an electric locomotive. This device collects the electric current and power is provided to the locomotive by a system from above cables. So, this system of electricity collection to trains or trams provides a contact with regard to the pantograph design [11] - [12] . This system is a physical model in electrodynamics and it is related to the time dependence. The modeling of this structure is formulated by DDEs. In Fig. 1 , the pantograph above a tram is shown. This basic design of the pantograph is a physical motion and accomplishes DDEs with regard to the the vertical displacements of the pantograph at frame and head [13] - [15] In the model, we can also consider some constants such as the Figure 1 . The pantograph of a tram (İstiklal Street,İstanbul) [16] .

damping coefficients for describing the motions from the head to the frame and other way around respectively with µ E and µ D . A simple representation of a pantograph and its trolley wire system is shown in Fig. 2 which is formulated as

Here, k E and k D values are the constants which are presenting bounce between frame and head of the pantograph, m E and m D symbolise the masses of head and frame, respectively, y E (t) and y D (t) are related to the displacement of head and frame, respectively, λ y is the vertical component force, F u is the upward force and g is the gravity of earth [17] .

In biology, models which are described by the functional -differential equations have been established for determining a dynamics of some biological systems, and thermal science and mechanics systems. This wide usage of the FDEs supports mathematical 2/19 Figure 2 . A representation of a pantograph and its trolley wire system [17] .

biology studies and researchers in the field. For instance, recent studies on these type of models introduce a tool for explaining malaria, Dengue Fever epidemics and Covid-19 virus spreading models which include very frequently time delays [18] - [21] . Furthermore, Mackey and Glass [22] proposed two possible models to describe the change of density of Hematopoietic cells in the blood that is circulating in the human body [23] .

Particularly, in population dynamics DDEs are often used. Continuous-time models including delay term describe mainly structure of a population. In this research field, the rate of producing egg in a population, susceptibility to parasitism or a bacterial population and their relations with related to the time dependence and delay term have been used in many studies [24] - [25] . Also, population dynamics is described by DDEs with the examples of investigation on baleen whale populations, delay model for introducing the dynamics of human immunodeficiency virus (HIV) infection etc. [26] - [27] .

Diversely, finding the solutions of these types of models analitically can be difficult. Due to this reason, numerical methods such as Adomian decompostion, finite element methods, homotopic perturbation techniques, Runge-Kutta type methods, direct two-point block methods, collocation methods, variational iteration methods have been applied to find approximate solutions of these type of problems [28] - [35] .

In this study, a class of FDEs is described as following

with initial condition

Herein γ, β > 0 are given as constants, τ > 0 is described for a delay term by using the name "time delay". Assuming that the approximate solutions of our problem, which has been given at (3)-(4), defined in the form of Laguerre series as

Here Laguerre polynomials are L n (t);

and a n , n = 0, 1, ..., N are expressed the unknown coefficients of Laguerre polynomials. Moreover, N > 0 is an integer which is selected properly, i.e. N ≥ 2 [36] .

Applications of such dynamics in numerical studies are of valuable contribution to the field which enable us to understand the dynamical structure comprehensively. Here, an algorithmic approximation in the interdisciplinary concept is introduced which leads to widen related studies. The aim of the study is to give an alternative solution scheme and different perspective than the previous works in the field. In this way, it is aimed by this study that addresses the development of many fields through an important dynamic.

The organisation of the manuscript is given as follows. In Section 2, a mathematical outlook and preliminaries are given. In Section 3, the numerical method is defined with its details. Accuracy of the numerical technique is described in Section 4. In Section 5, numerical simulations are demonstrated by figures and tables. A brief conclusion and future studies are presented in Section 6.

Here some preliminaries about Laguerre polynomials and their properties on the delay differential equations has been introduced for the further applications. Besides, the recurrence relations of series and definitions with regard to the model are introduced to support the technical results.

Interdisiplinary researches of the Laguerre polynomials are well-known in many fields. The Laguerre polynomials are important in the applications of quantum theory in the hydrogen atom concept, chemical and mathematical physics [37] . They play an important role at the numerical methods in applied mathematics field such as Laguerre collocation method [38] , [39] , [40] , [41] , Laguerre spectral method [42] - [43] , Laguerre pseudospectral method [44] , and Laguerre wavelets collocation method [45] . Besides, they support the improvement of the numerical solutions by combining with other numerical techniques, i.e. Laguerre-Galerkin methods [46] , Laguerre Tau methods and so on [47] .

Moreover, Laguerre polynomials are of a close relation with Hermite polynomials. Most of their properties and applications at the research area of orthogonal polynomials may help to find out characteristic of the Hermite polynomials. The Hermite polynomials H n (t) are connected to Laguerre polynomials as following:

In Fig. 3 , we can see the relations between the first few Laguerre and Hermite polynomials with n = 0, 1, 2, 3 and 0 ≤ t ≤ 2 where Laguerre polynomials are presented by red lines and Hermite polynomials are shown by blue lines. Definition 1 The first few Laguerre and Hermite polynomials are defined, respectively, as following:

Hermite polynomials have been applied on physical models, more specifically, at the solution of simple harmonic oscillator of quantum mechanics and Laguerre polynomials are seen in wave functions of hydrogen atom [48] . Usage of Laguerre polynomials are also well defined for the series solution. In this study, Laguerre polynomials give us an alternative series solution for the delay differential equations which is supported by following lemma.

which is a piecewise continuous function. Then for

we have converges uniformly

for any positive integer q [47] .

In this study, a special class of functional differential equations has been considered. Functional differential equations are of many different types including delay differential equations or more specifically retarded, neutral and advanced functional differential equations [49] . Now, the definitions with regard to these equations are introduced in order to complete the problem settlement.

Definition 2 Here the following equation is considered:

which is given with 1st order in its derivative and difference. In this place, an equation in the form above is called as retarded type where ζ = 0.

In this section, a general scheme with regard to Laguerre collocation technique is introduced. We provide a novel collocation structure by using the Laguerre polynomial approach directly with related to some fundamental relations between the matrices.

Here, the idea of the implementation of the technique is presented together with collocation points. The truncation of the series is shown and algebraic equations are obtained. We have the series approach and show the approximate solutions with the help of these connections.

Our motivation here is to show that we expand the well-conditioned collocation method with regard to Laguerre matrix approach. The applications of the technique some real-world models give us a chance to investigate the models based on numerical concept.

In this subsection, an approximate solution with related to Laguerre polynomials form (6) is presented. For our purpose, we consider a matrix form of Eq. (5) as

in which

Then, we use the matrix relation

where

Besides, the connections are defined between X(t) and X (t). So, we write them as

where

6/19

Then from the relations in (15) and (19), we get

So that, from (17), (19) and (21), we have

where

Then we have

Moreover, t → t − τ is replaced into (15) and get

where

Then

are defined and in which

By using the equations (28)-(30), a matrix form of (3) is described as

Here 

In this subsection, collocation method is applied by using the collocation points which give us a pointwise approximation with regard to our truncation number N together with step size h. Then by replacing the collocation points

into Eq. (34) , so that the fundamental matrix equation is obtained and written as follows:

Briefly,

Correspondingly, the initial condition is written in the matrix form as

or

Therefore, the last rows of the augmented matrix (38) is reestablished by the row matrix (39) in order to get an approximate solution of Eq. (3) under the condition (4). Thus, the augmented matrix is obtained as

Laguerre coefficients can be computed by solving the augmented matrix system [50] - [51] . Hence, the approximate solution is established as in the desired form of Laguerre series (6) as

In this section, errors of approximation and solution strategy with regard to the its algorithm have been introduced. Error analysis has been proposed by using different norms in order to show the results more comprehensively. This lead us to see efficiency of our numerical technique. Besides, algorithm has been presented which is based on programming part of the construction of our method. Hereby the application of the numerical method by using a computer programming has been shared.

Accuracy is an important topic in approximation theory which has a reason to be introduced due to the errors occur in rounding and truncation of numerical solutions. When we use numerical algorithms with finite sensitivity, approximation errors, round or truncation error at the series solutions appears. Due to this reason, we should show the accuracy of the technique and how we deal with this approximation at the end of our application.

In this subsection, a short presentation of the error estimation for the solutions with regard to Laguerre approach (6) is given. Hereby accuracy of this technique is supported by this approximation. Let us first describe, E N (t), "error function" for t = t α , α = 0, 1, ...

where k is appointed and we pay attention to the truncation limit N which is an important argument for finding the approximate results. Accordingly, the difference E N (t α ) becomes smaller and, in here, this is denoted by 10 −k . Diversely, some particular norms are given in order to investigate the distinctive error functions for measuring errors. So, these are described as follows.

.

Here e l = u(t l ) − u N (t l ) also u N (t l ) and u(t l ) are approximate and exact solutions of the problem, respectively.

In this section, Steps of algorithm for the present method: 

We found an effective algorithmic approach for calculating an approximation procedure and to investigate the dynamics behind our model. This made a considerable impact on the dynamics and it has an excellent significance framework on approximation methodology.

Algorithms particularly connect the applicability with time standard. In order to increase the quality and to decrease the central processing unit (CPU) time, we may use an effective approach for coding. Due to this reason, straightforward techniques are of usable and effective to get the approximations. Then we can see in 4.2 that our numerical method may be applied with regard to this idea. Besides, the following flowchart of the algorithm shows us a standardized approach to find the approximate solutions in our problem (3) In here, some illustrative numerical examples are presented to show applicability of this given present method. These examples have been chosen from science and engineering applications to be analysed. Besides, the algorithm, which has been explained 4.2, is applied on the problems. The problem has been introduced with the general formula at Example 1. Let us first consider a model from biology which is called as "Wazewska-Czyzewska and Lasota model". This biological model describes a remainder red blood cells, specifically, in animals [31] , [56] :

and its initial condition

Here, number of total red blood cells at time, denoted by t, is expressed by u(t).

Moreover, γ which is presenting death rate of a unique red blood cell. Besides, β and ρ are denoted red blood cells generation per unit time and τ is time essential for generating a red blood cell. Specifically, l = 1, γ = 0.4, β = ρ = 1 are chosen. So that, we have (28) as

Then we have the matrix form as

By using the procedure, we obtain approximate solutions. The approximate solutions for different N values can be seen in Figure 1 . Different error norm results for N = 3 is given by Table 1 . Example 2. Here, the model is considered as in the form [57] , [58] : 

with initial conditions u 1 (t) = 1 and u 2 (t) = 1, t ≤ 0. 

In this study, numerical investigation of a class of FDEs has been considered. Besides, a technique based on matrices together with collocation points has been introduced. With the help of computer programmes, Maple and Matlab, numerical approximation of Laguerre collocation method has been compared with the techniques such as Runge-Kutta and Hermite collocation methods. Moreover, the technique together with an error analysis have been implemented on some illustrations to see the applicability as well as efficiency of the method. Then the results have been seen by figures and tables.

We have the advantages of the method such as straightforward applicability and coding come along with the proven convergency by some authors [60] , [61] , [62] . As a future plan, this technique can be improved and applied on some other mathematical models after completing some modifications.

Numerical methods for delay differential equations

Differential equations: Stability, oscillations, time lags

Generalized delay differential equations to economic dynamics and control

An application of delay differential equations to market equilibrium, The Functional and Algebraic Method for Differential Equations

Differential-difference equations in economics: on the numerical solution of vintage capital growth models

Delay differential nonlinear economic models

A stochastic maximum principle for a markov regime-switching jump-diffusion model with delay and an application to finance

Optimal nonlinear information processing capacity in delay-based reservoir computer

Time-delay reservoir computers and high-speed information processing capacity

Chatter classification in turning using machine learning and topological data analysis, IFAC-PapersOnLine

Series solution of the pantograph equation and its properties

The use of the decomposition procedure of Adomian for solving a delay differential equation arising in electrodynamics

Numerical method for calculating the dynamic behaviour of a trolley wire overhead contact system for electric railways

Third paper: Calculating the behaviour of an overhead catenary system for railway electrification, Proceedings of the institution of mechanical engineers

Pantograph motion on a nearly uniform railway overhead line

Istanbul plans third heritage-style tramway

An advanced 3D-model for the study and simulation of the pantograph catenary system

Infectious Diseases of Humans: Dynamics and Control

Mathematical Biology 1: An Introduction, 3 rd edition

A delay differential equation model for Dengue Fever Transmission in selected countries of South-East Asia, Doctoral dissertation

Solvable delay model for epidemic spreading: the case of Covid-19 in Italy, preprint, arXivmath

Oscillation chaos in physiological control systems

Fitted numerical methods for delay differential equations arising in biology, Doctoral dissertation

Delay-differential equations for structured populations, in Structured-population models in marine, terrestrial, and freshwater systems

Delay equations: functional-, complex-, and nonlinear analysis

Delay model for the dynamics if HIV infection

A delayed-recruitment model of population dynamics with an application to baleen whale populations

An Introduction to Delay Differential Equations with Applications to the Life Sciences

The Adomian decomposition method for solving delay differential equation

Stability properties of numerical methods for solving delay differential equations

Numerical dynamics of a nonstandard finite difference method for a class of delay differential equations

Solution of delay differential equations via a homotopy perturbation method

The variational iteration method for solving a neutral functional-differential equation with proportional delays

Solving second order delay differential equations using direct two-point block method

A Bessel collocation method for numerical solution of generalized pantograph equations

Tables of Laguerre Polynomials and Functions: Mathematical Tables Series

Effective Laguerre asymptotics

Modified operational matrix method for second-order nonlinear ordinary differential equations with quadratic and cubic terms

A Modified Laguerre Matrix Approach for Burgers -Fisher Type Nonlinear Equations, Numerical Solutions of Realistic Nonlinear Phenomena

Laguerre Matrix -Collocation Method to Solve Systems of Pantograph Type Delay Differential Equations

A numerical solution of parabolic-type Volterra partial integro-differential equations by Laguerre collocation method

A new generalized Laguerre spectral approximation and its applications

A modified generalized Laguerre spectral method for fractional differential equations on the half line

The Laguerre pseudospectral method for the two-dimensional Schrödinger equation with symmetric nonseparable potentials

Numerical solutions for the linear and nonlinear singular boundary value problems using Laguerre wavelets

Laguerre-Galerkin method for nonlinear partial differential equations on a semi-infinite interval

Laguerre Tau methods for solving higher-order ordinary differential equations

Mathematical methods for physicists

Differential-difference equations

Lucas collocation method for system of high-order linear functional differential equations

Laguerre polynomial approach for solving linear delay difference equations

Morgan-Voyce polynomial approach for solution of high-order linear differential-difference equations with residual error estimation

Rumour propagation: an operational research approach by computational and information theory

Mathematical problems of the dynamics of the red blood cells system

Solving Linear First Order Delay Differential Equations by MOC and Steps Method Comparing with Matlab Solver

Solving ddes in Matlab

A contribution to the mathematical theory of epidemics

The use of generalized Laguerre polynomials in spectral methods for solving fractional delay differential equations

Solving ddes in Matlab

On the attainable order of collocation methods for pantograph integro-differential equations