You are on page 1of 96

Mathematical Model Analysis on Smoking Cigarette with

Optimal Control

MSc. Thesis

By: Bezabih Bent

Hawassa University

Hawassa, Ethiopia
November, 2022
A Thesis
on

Mathematical Model Analysis on Smoking Cigarette with


Optimal Control

MSc. Thesis

By: Bezabih Bent

Advisor: Zerihun Kinfe (PhD)(Assistant Professor)

Co-Advisor: Tadele Tesfa (PhD)


Submitted to:
Department of Mathematics
College of Natural and Computational Science
Hawassa University

In Partial Fulfillment of the Requirement for the Degree of


Master of Science
in Computational Mathematics

Hawassa, Ethiopia
November, 2022
Declaration

I hereby declare that “Mathematical Model Analysis on Smoking Cigarette with Optimal Con-
trol” is my own work, that it has not been done for any degree in any other universities, and that
all the sources I have used have been indicated and acknowledged by complete references.

Name Signature Date

Hawassa, Ethiopia
November, 2022

i
Approval Sheet-1
This is to officially state that the study entitled “Mathematical Model Analysis on Smoking
Cigarette with Optimal Control” is an original work carried out by Bezabih Bent, ID No.
GPCoMaR/0002/12 under my guidance and supervision. This is a genuine work that has been
done by Bezabih Bent for the partial fulfillment of the award of the Degree of Master of Science
in Computational Mathematics from Hawassa University. Daily acknowledgments are done
during his course of investigation. Therefore, I recommend that it would be accepted as fulfilling
the thesis requirements.

Advisor Name Signature Date

Co-Advisor Name Signature Date

ii
Approval Sheet-2

We, the undersigned, members of the Board of Examiners of the final open defense by Bezabih
Bent have read and evaluated his thesis entitled“Mathematical Model Analysis on Smoking
Cigarette with Optimal Control” and examined the candidate. This is therefore to certify that
the thesis has been accepted in partial fulfillment of the requirement of the Degree of Master of
Science in Computational Mathematics

Name of Internal Examiner 1 Signature Date

Name of External Examiner Signature Date

Name of Internal Examiner 2 Signature Date

Name of Advisor Signature Date

Name of Co-Advisor Signature Date

Department Head Signature Date

SGS Approval Signature Date

iii
Acknowledgment

First of all,I would like to thanks almighty God for caring my life and blessing my activities in
advance of the completion of my research work at everywhere.

My special gratitude and appreciation goes to my advisor, Dr.Zerihun Kinfe (Assistant Pro-
fessor) and my co-advisor,Dr.Tadele Tesfa for their great assisting,useful suggestion,correcting
and giving comments by devoting adeal of their time.They gave me many suggestion and com-
ments,adjust the time to follow up,valuable advice and guidance that contributed to me the suc-
cessful and realization from the beginning of the title selection,the proposal and to the concluding
level of the thesis.I greatly appreciate them not only with their knowledge and unlimited construc-
tive advices in every aspect of my work,but also with their professional and personal ethic.They
are role model for my life in work profession.

In sponsoring,I would like to express my greatest thanks to Sidama National Regional State
for sponsoring me by covering educational and research expenses financially,support and giving
achance to follow up on this education program.

I am extremely grateful to express my gratitude to all the Hawassa University of the staff of
mathematics department for their knowledge sharing,especially I want to give great thanks to
Dr.Mohammed Yiha and Dr.Kiros Gebrearegawi for their precious suggestions and comments
during my studies-May God bless them. And also I would like thank you my class mates for their
cooperative work through our study time and doing this thesis.

Finally,I would like to thanks to all my family special my mother Tadelech Chea and my wife
Abaynesh Abebe and her Father and Mother for their patience,tolerance and supporting in
completing the program.

iv
List of Abbreviations

Abbreviations Meaning
WHO World Health Organization
FMHACA Food,Medicine and Health care Administration and Control Authority
SFE Smoking-free equilibrium
SPE Smoking present equilibrium point
IVP Initial-value problem
ODE Ordinary differential equation
MATLAB Matrix Labratory
RHC Routh-Hurwitz criteria
R0 Reproduction number
OC Optimal control
PMP Pontryagin’s Maximum Principle

v
Table of contents

Declaration i

Acknowledgment iv

List of Abbreviations v

List of Tables ix

List of Figures x

Abstract xi

1 Introduction 1
1.1 Background of the study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Statement of the problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3 Objective of the study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3.1 General objective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3.2 Specific objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.4 Significance of the study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.5 Limitation of the Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.6 Organization of the thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

2 Literature Review 9

3 Methodology(Or Mathematical Modelling Tools) 12


3.1 System of ordinary differential equations . . . . . . . . . . . . . . . . . . . . . . . 12
3.1.1 Stability analysis of equilibrium points . . . . . . . . . . . . . . . . . . . 14
3.1.2 Routh-Hurwitz stability criterion . . . . . . . . . . . . . . . . . . . . . . . 16
3.2 Optimal control theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.2.1 The cost function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

vi
3.2.2 Pontryagin’s Maximum Principle (PMP) . . . . . . . . . . . . . . . . . . 18
3.2.3 Numerical computation of optimal control problems . . . . . . . . . . . . 19
3.3 Basic concepts in epidemiological modeling . . . . . . . . . . . . . . . . . . . . . 20
3.3.1 Smoking Generation Number/ Basic Reproduction Number . . . . . . . . 20
3.3.2 Sensitivity Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

4 Mathematical Model Formulation and Analysis 22


4.1 The Existing Mathematical Model . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4.2 The Modified Mathematical Model . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.2.1 Mathematical Model Formulation . . . . . . . . . . . . . . . . . . . . . . 24
4.3 Scaling of Modified Model Equation . . . . . . . . . . . . . . . . . . . . . . . . . 28
4.4 Qualitative Analysis of the Modified Model . . . . . . . . . . . . . . . . . . . . . 29
4.4.1 Well-posedness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
4.4.2 Equilibrium Point . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
4.4.3 Basic Reproduction Number . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.4.4 Stability Analysis of the Smoking Free Equilibrium . . . . . . . . . . . . . 38
4.4.5 Stability of Smoking Free Equilibrium Point . . . . . . . . . . . . . . . . 38
4.4.6 Local Stability of Smoking Free Equilibrium . . . . . . . . . . . . . . . . 38
4.4.7 Global Stability of Smoking Free Equilibrium . . . . . . . . . . . . . . . . 42
4.4.8 Smoking Present Equilibrium Point(SPE) . . . . . . . . . . . . . . . . . . 44
4.4.9 Local Stability of Smoking Present Equilibrium Point . . . . . . . . . . . 45
4.4.10 Sensitivity Analysis(Or Sensitivity Of the Basic Reproduction
Number) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
4.5 Numerical Simulations and Discussion . . . . . . . . . . . . . . . . . . . . . . . . 54

5 Extension of the Modified Model into an Optimal Control 58


5.1 Characterization of the Optimal Control . . . . . . . . . . . . . . . . . . . . . . . 58
5.1.1 Existence of Optimal Control Solution . . . . . . . . . . . . . . . . . . . . 60
5.1.2 The Hamiltonian and Optimality System . . . . . . . . . . . . . . . . . . 62
5.1.3 Numerical Simulations of Optimal Control Problem . . . . . . . . . . . . 67

6 Conclusion and Recommendation 70


6.1 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
6.2 Recommendation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71

References 72

vii
Appendices 77

viii
List of Tables

4.1 The state variables and their descriptions. . . . . . . . . . . . . . . . . . . . . . . 25


4.2 Parameters of the modified mathematical model and their descriptions. . . . . . . . 27
4.3 The Sensitivity index value of modified model . . . . . . . . . . . . . . . . . . . . 52
4.4 The parameter values of the modified model. . . . . . . . . . . . . . . . . . . . . . 54

ix
List of Figures

4.1 Flow diagram of the existing model. . . . . . . . . . . . . . . . . . . . . . . . . . 23


4.2 The flow chart for the modified mathematical model of smoking dynamics. . . . . 25
4.3 (a) Time sries of state variables for R0 = 0.41390. (b) Time sries of state variables
for R0 = 13.13692. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.4 Time series of state variables for R0 = 0.41390 with different initial condition. . . 56
4.5 Time series of state variables for R0 = 13.13692 with different initial condition. . . 56
4.6 Effects of ε on chain smokers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
4.7 (a) Parameters β1 having positive impact on the spread of smoking habit (b) Pa-
rameters β2 having positive impact on the spread of smoking habit . . . . . . . . . 57

5.1 Simulation showing the optimal solution for potential,occasional,chain smokers


through use of u1 &u2 ,when β1 = 2&β2 = 0.608. . . . . . . . . . . . . . . . . . . 68
5.2 (a) with and with out optimal control of temporarily quit smokers. (b) with and
with out optimal control of permanent quit smokers. . . . . . . . . . . . . . . . . . 68
5.3 Simulation showing the profile for the optimal control u1 &u2 . . . . . . . . . . . . 69

x
Abstract

In this thesis,we modified and analyzed a mathematical model that shows the dynamics of smoking
habit by considering contact of smokers with in susceptible population.The existence,positivity,and
boundednes of a solution of model are proved.The basic reproduction number is computed using
next generation matrix method.The analysis explains that smoking free equilibrium point is lo-
cally and globally asymptotically stable when R0 < 1,also the smoking present equilibrium point
is locally asymptotically stable when R0 > 1.Using sensitivity analysis we establish that R0 is
most sensitive to the contact rate and converting rate to control the smoking habit.The model
was extended to optimal control to verify the effectiveness of control strategies in order to reduce
the smoking habit from population.Characterization of the optimal control is established using
Pontryagin’s Maximum principle.Numerical simulation with and with out optimal control are per-
formed to verify the analytical solution by using MATLAB software.
Keywords: Smoking, Smoking generation number, stability analysis, Numerical simulation,
Optimal control.

xi
Chapter 1

Introduction

This chapter describes the overall introduction to the study. It mainly focuses on the background
of the study, statement of the problem, objectives and significance of the study.

1.1 Background of the study


Smoking may be a behavior during which a substance is heated or burned, and therefore the de-
rived smoke is breathed through the mouth or the nose and goes on to the lungs. The active
substances of the smoke are rapidly transmitted from the lungs to the bloodstream and thus to the
brain, where they operate[1]. The most commonly consumed substance is the dried leaves of the
tobacco plant, which are rolled into a little square of paper to make a little round cylinder called a
”cigarette”[2].Tobacco is consumed by nearly 1.3 billion smokers worldwide[1].
Smoking is one of the leading causes of health problems and continues to be one of the world’s
most vital health challenges. Consistent with the World Health Organization report on the world-
wide tobacco epidemic [16], tobacco use kills or disables many of us in our most efficient years,
which denies families their primary wage-earners, consumes family budgets, raises the cost of
health care and hinders economic development. Smoking or tobacco may be a known cause or
evidence of deaths from cancers of the mouth, larynx, lung, esophagus, bladder, pancreas, pelvis,
stomach, and cervix. Smoking is additionally a explanation for heart condition,strokes, peripheral
vascular diseases, chronic obstructive lung diseases, and other respiratory diseases, and low-birth
weight babies [17]
Cigarettes are a fancy mixture of highly addictive products: nicotine, polyaromatic hydrocar-
bons, phenols, and nitrosamines, which are submicron-sized solid particles; other components, like
carbon monoxide gas, compounds, and nitrogen oxides, are gases. When a cigarette is smoked,
there are approximately 600 ingredients that, when burned, create over 7, 000 chemicals, including
a minimum of 70 known carcinogens [12].

1
The habit of smoking was initially spread into Europe with the arrival of Columbus in the 16th
century [4]. However, after and before this habit, other species of strange nature have adversely
affected the human habitat and, therefore, the whole ecosystem. Nicotine was named after him as
a result of the current connection). Since the invention of cigarette manufacturing equipment in
the late nineteenth century, the assembly speed for smoke has increased from 200 units per minute
to 9000 units per minute.
More than 7 million people die each year from tobacco-related illnesses worldwide (more than
6 million from direct tobacco use and approximately 90, 000 nonsmokers exposed to secondhand
smoke) [10], and this number is expected to rise to more than 8 million by 2030 [11].
A number of nations have legislation restricting tobacco advertising, and regulating who can
purchase and use tobacco products, and where people can smoke [5].In Ethiopia, a signatory to the
United Nations Framework Convention on Tobacco Control [6], passed a proclamation banning the
advertisement of cigarettes or other tobacco products to the general public [7]. Within the Food,
Medicine, and Health Care Administration and Control Authority (FMHACA) guidelines, tobacco
use policy is boldly presented to regulate the harm from tobacco use [8].
The World Health Organization estimates that one billion people worldwide could die from
tobacco-related illnesses by the end of the 21st century if current rates of tobacco use continue
unabated[18].
A mathematical model is an idealization of a real-world phenomenon. Consistent with Ruther-
ford Aris, a mathematical model may be a set of mathematical equations that provide an adequate
description of a physical system. A ”physical system” is often broadly interpreted as any real-
world problem-natural or man-made, discrete or continuous, and may be deterministic, or random
in behavior.
A mathematical model involves the utilization of mathematics to explain,or predict behavior
or phenomena within the world. That is, modeling is one of the ways in which formulae are
deduced to unravel a true-life situation [13]. It is often particularly useful in investigating questions
or testing ideas within complex systems. The model is then analyzed, solved, or simulated on
a computer. The results are often interpreted in physical terms to assist understanding of the
underlying system or to point to parts of the system which may be targeted for change[14].
Why can we affect a model rather than a true world problem? A variety of mathematical models
are used for a variety of different reasons. How well any particular objective is achieved depends
on both the state of data in a few systems and the way the modeling is done[15]. It’s a crucial
tool in analyzing the spread and control of infectious diseases. The model formulation process
clarifies assumptions, variables, and parameters. Moreover,models provide conceptual results like
the essential reproduction numbers .

2
All human beings are born free and equal in dignity and rights. It is important to be careful,
especially since smoking can be difficult to quit once you have started, so it is a good idea to protect
yourself and your society from the harmful effects of smoking on your health.
In this study we want to proceed population growth rate is linear in closed system, probability
of contact rate of smokers with non smokers,probability of occasional smokers would become
permanent quit smokers, probability of temporary quit smokers would become permanent quit
smokers after that how to reduce those smokers and maximize the number of non smokers.

3
1.2 Statement of the problem
Smoking is one of the leading causes of health problems and continues to be one of the most
significant health challenges in the entire world, but smoking remains socially accepted. That
means smokers are smokes by choice only. Smoking may be a major public health problem in
the world and has the potential to be addictive because it contains a high proportion of nicotine.
However, a special danger in the sense of the smoking habit has begun to become widespread
among young people. Tobacco users who pass on prematurely deprive their families of earnings,
raise the value of fitness care, and hinder financial development. The immense public health peril
related to smoking impend to concentrate on the smoking dynamics within the community, focused
on investigating the techniques to prevent this harmful habit. In spite of these efforts, smoking
persists, causing socio-economic and health problems for a large population worldwide. We all
know that, nowadays, smoking is a serious burden on public health everywhere on the earth. Hence,
it seems to be quite important to undertake more research on the spread dynamics of smoking in
such areas, and a significant control strategy must be established.
In 2021, Omar et al. [1] studied mathematical modeling and optimal control analysis of giving
up smoking dynamics. Within the model, the population growth is linear,probability of contact of
potential smokers with occasional and chainsmokers,probabilty of temporarily quit smokers be-
come permanent quit smokers and the probability of occasional smokers becoming permanent quit
smokers weren’t considered. Thus, this work may be a modification of [1] by considering the
population growth is linear,probability of contact of potential smokers with occasional and chain
smokers,probability of temporarily quit smokers become permanent quit smokers and the probabil-
ity that occasional smokers will become permanent quit smokers after time.And also we considered
,how we minimize those smokers (i.e occasional smokers and chain smokers)by minimum cost by
using Pontryagin maximum principle (PMP) in order to prevention. The study is predicted to deal
with the subsequent basic questions:

• How to modify an existing mathematical model for the dynamics of smoking?

• How to analyze the modified mathematical model for the dynamics of smoking?

• How to scale (non-dimensionalize) the modified model ?

• How to show existence,positivity and boundedness of the solution of the model?

• To which parameters is the dynamics of smoking habit is more sensitive?

• Is the mathematical model we developed helpful to eradicate smoking habit?

4
• What are the biological interpretation of the solution of the mathematical model we devel-
oped to describe the dynamics of system?

• How to minimize the spread of smoking habit with optimum implementation cost?

• How to simulate a non-linear mathematical model using MATLAB software?

• How to show stability analysis of equilibrium states?

• How to determine the smoking generation number using the next generation matrix
method ?

1.3 Objective of the study


1.3.1 General objective
The general objective of this study is to analyze a modified mathematical model on the dynamics
of cigarette smoking with optimal control.

1.3.2 Specific objectives


The specific objectives of this study are:

• To formulate a modified mathematical model for smoking dynamics.

• To non-dimensionalize(Or scalling) the modified model of dynamic system .

• To show existence,positivity and boundedness of the solution of the model.

• To determine the smoking generation number for the model using the next generation matrix
method.

• To investigate stability analysis of the smoking free and persistent equilibrium point.

• To identify the sensitive parameter which has a great impact on the dynamics of the smoking
habit.

• To solve an optimal control problem value.

• To determine the relative importance parameter to the spread of smoking habit, and

• To simulate and analyses the model by using MATLAB software .

5
1.4 Significance of the study
This study prompted us to conclude that smoking is widespread in communities around the world.
And the way we planned to solve this problem, we thought it would help stop smoking in society.
The benefits we get from this study:

• Develop our understanding on the transmission of smoking cigarette dynamics.

• Reduce the number of smokers .

• To keep a generation from becoming addicted for smoking.

• It’ll be used as an input for further investigation into smoking intervention strategies.

• It’ll be used as a reference for related research within the field of epidemiology.

• It’ll give some information for brand new researchers to develop their own models.

• Develop an epidemiological modeling of smoking cigarette with optimal control of govern-


ment prohibition and education campaign and increase cigarette cost strategy.

6
1.5 Limitation of the Study
The following are some limitations of this study:

• In this thesis, gender, age, and education levels are not considered.

• Since we focused mainly on the analysis and simulation part of the modified model with
optimal control strategy we used secondary data from related review literature to estimating
the parameter values due to the scarcity of real data.

7
1.6 Organization of the thesis
The thesis is organized as follows: Chapter (2) presents review of the related literature. Chapter
(3) presents some epidemiological preliminaries and the methodology. The mathematical model
is formulated and described in Chapter (4). The qualitative analysis of the modified model by
examining the equilibrium points of smoking dynamics and its stability analysis is also studied in
this chapter. In Chapter (5) an optimal control problem is formulated and studied using Pontrya-
gin’s Maximum Principle. In this chapter, we also present numerical simulations for the optimality
systems of dynamics. Conclusions and recommendations of the study are presented in Chapter (6).

8
Chapter 2

Literature Review

Like many infectious diseases, mathematical models are often used to understand the spread of
smoking and predict the impact of smokers on the community, so as to assist in reducing the
number of smokers.
Mathematical biology is a wide field with many applications. In this field, researchers are
focusing on the description of different types of diseases with controls in the form of mathematical
models. In 1909, Brownlee [61] took the initiative for the development of mathematical biology.
He focused on the theory of chance, further in 1912, he presented basic laws for epidemic spreading
[62]. In 1927, the details of the epidemic study were discussed by Kermark and McKendrik [63].
Later, many researchers discussed different models of many other diseases.
In 2000, Castillo-Garsow et al.[19] for the first time proposed an easy mathematical model for
abandoning smoking. They consider a system with a completely constant population that is split
into three classes: potential smokers, that is, people that don’t smoke yet but might become smok-
ers within the future (P), smokers (S), and other people (former smokers) who have quit smoking
permanently (Q). In 2008 Sharami and Gumel developed mathematical models by introducing
mild and chain classes [49]. In their work, they presented the events and public health impacts
of smoking-related illnesses. Zaman [50] extended the work of Castillo-Garsow et al.[19] and de-
veloped a model taking into consideration the occasional smokers’ compartment within the given
smoking model and presented its qualitative behavior.
Hung et al.[20] established a mathematical model for smoking cessation with e-cigarettes and
presented a theoretical study for the effect of e-cigarettes, Straughan [21] considered a differential
equation model for the effect of smoking e-cigarettes may have on smokers of tobacco cigarettes or
on non-smokers. According to their discussion, there are only two studies on e-cigarette smoking
within the debate. With these motivations, during this study, they aimed to debate the mathematical
model developed by Straughan [21] in 2018 and to express the habit of e-cigarettes under peer
pressure.

9
Sintayehu Matintu [22] studied a basic non-constant population P OSQT QP model in 2017,
focusing on the effect of moderate and heavy smokers on potential smokers, as well as the effect of
heavy smokers on temporarily quitting smokers. Under his consideration, if an individual first be-
comes a smoker, it’s unlikely that she/he will quit for several years since tobacco contains nicotine,
which is shown to be an addictive drug.
Castillo-Garson et al.[19] and Sharomi et al. [49] have proposed two simple mathematical
models with standardized incidence rates and constant population size to research the dynamics
properties of smokers, respectively. Lahrouz et al. [51] have presented a four-dimensional ordinary
differential equations model with a bilinear incidence rate to review smoking behavior. Zeb et al.
[52] have studied a model of smoking cessation with a square-root incidence rate.
In 2014,Zainab et.al [53] introduced a brand new model by dividing smokers into two sub-
classes: occasional smokers and heavy smokers, and therefore the impact of those two subclasses
on the existence and stability of equilibrium points. They studied the effects of occasional smokers
on potential smokers.
Zaman [54] presented the optimal campaigns within the smoking dynamics. They considered
two possible control variables within the sort of education and treatment campaigns oriented to
decreasing the attitude towards smoking and first showed the existence of optimal control for the
control problem. However, in reality, the standard quit smokers are only temporary quit smokers.
A number of them may relapse since they have contact with smokers again, and the others may
become permanent quit smokers.
S. Al-Sheikh et al[31], proposed a smoking mathematical model. They derive and analyze a
model of smoking in which the population is divided into four classes: potential smokers, smokers,
temporary quitters, and permanent quitters. In this model, they studied the effects of smoking on
temporary quitters. Two equilibria of the model are found: one is the smoking-free equilibrium,
and the other corresponds to the presence of smoking. They examined the local and global stability
of both equilibria and supported the results by using numerical simulations.
Developing a mathematical model for tobacco smoking analysis, according to Medhin Asfaw
[32] The study was done in Ethiopia, in the case of Haremaya town, based on real data. The model
consists of five compartments; nonsmoker individuals (P), exposed (E), smokers (S), temporar-
ily quit smokers, and permanently quiet smokers. The study showed that the smoking habit in
Haremaya town is increasing. The researcher analyzed the stability of the model equilibria and
identified the main causes of smoking habits as peer pressure (chewing chat, watching the film),
smoking habits by parents around the home, considering smoking as a civilization, and even some
smokers do not know why they smoke because they started at an early age. It was determined that
the problem can be detected by lowering contact, reversion, and quitting rates.
Anwar Zeb et al.[59]consider a delayed smoking model that assumes that potential smokers

10
satisfy the logistic equation. they describe the dynamic behavior of the proposed model in terms
of delay differential equations (DDEs). they then give the conditions for the asymptotic stability
of the model at steady state. They also discuss Hopf bifurcation analysis for the model considered.
They use at the end a non-standard finite difference scheme (NSFD) for graphing results Using
MATLAB.
Abdullah Alzahrani [60] proposed that smokers are at increased risk of COVID-19 because
their fingers regularly touch their lips while smoking, increasing the likelihood of hand-to-mouth
transmission of the virus. A smoker, on the other hand, can suffer serious illnesses, especially
lung disease (or reduced lung capacity), which greatly increases her risk of COVID-19. For this
estimation, this work first formulates a mathematical model involving the reversal class. They use
various techniques to find the local and global stability of the presented model with respect to the
equilibrium point, the equilibrium point between free and positive smoking. Since the model con-
sists of nonlinear equations, they use non-standard finite difference schemes (NSFD), ODE45 and
RK4 methods to find numerical results. Finally, display the graph numerically using MATLAB.
Omar et al. [1] proposed a conceptual P LSQT Qp model to review smoking dynamics in 2021.
Routh–Hurwitz conditions are used to establish local stability of the equilibria. Global stability
was analyzed using the Lyapunov function. Furthermore, they introduced time-dependent controls
to the essential model and extended it to an optimal control model of smoking. An optimal control
problem was formulated and studied analytically using Pontryagin’s Maximum Principle.

11
Chapter 3

Methodology(Or Mathematical Modelling


Tools)

In this study, we used a non-linear deterministic model and therefore the model is formulated by
using system of ordinary differential equation. The stability analysis of the equilibrium points was
done by classifying their steady states as smoking free equilibrium point and persistent equilibrium
point . We will analyze local stability equilibrium point by using the Routh-Hurwitz stability
criteria and the global stability of equilibria by using Castillo-Chavez theorem .And also we used
the next generation matrix to work out the basic reproduction number. Computing and simulating
are done with the aid of mathematical software like MATLAB. Furthermore, the forward-backward
sweep method will be used to run the numerical simulations for the optimal control issue. Basic
epidemiological principles and dynamical system ideas are also crucial modeling and solution-
analysis tools. The following definitions and theorems are introduced because they are crucial for
the subject to progress.

3.1 System of ordinary differential equations


Let’s have a look at an autonomous n-dimensional system of the following type:
dx
= ẋ(t) = f (x(t))
dt (3.1.1)
x(t0 ) = x0 ,
where x0 , x ∈ D ⊂ Rn and f : Rn −→ Rn ; with f is continuous at x ∈ D ⊂ Rn .

Definition 3.1.1 ([34]). (Well-posedness)


If the following conditions are satisfied, an initial value problem (IVP) given in equation (3.1.1) is
considered to be well-posed:

12
i. Its solution exists,

ii. the solution is unique,

iii. Its solution continuously depends on the initial conditions.

Theorem 3.1.1 ([35]). (Picard’s theorem)


Consider the initial value problem raised in (3.1.1). if the function f is continuous and that all
∂fi
its partial derivatives ∂x j
, for i, j = 1, 2, 3, ..., n are continuous for x in some open connected set
n
D ⊂ R , then for x0 ∈ D the problem (3.1.1) has a solution x(t) on some time interval (−α, α) ,
α > 0 about t = 0, and the solution is unique.

Definition 3.1.2 ([36]). (Positivity of solution)


The solution of a given autonomous system, of (3.1.1) is said to be positive, if all trajectories x(t)
is positive for any t ≥ 0.

Definition 3.1.3 ([37]). (Boundedness of solution)


The positive solution given in (3.1.3) of an autonomous system, of (3.1.1) is said to be bounded, if
any solution, x(t, t0 , x0 ) (3.1.1) satisfies

kx(t, t0 , x0 )k ≤ K(kx0 k, t0 )

for all t ≥ t0 where, K : R+ × R+ → R+ is a constant that depends on t0 and x0 .

Definition 3.1.4. (Autonomous differential equation)


Is a system of ordinary differential equation which does not explicitly depend on the independent
variable.When the variable is time,they are also called time invariant systems which is the form of
the :

dx
= ẋ(t) = f (x(t))
dt
for certain function f

Definition 3.1.5. (ordinary differential equation)

An ordinary differential equation (ODE) is a differential equation involving one or several


derivatives of a dependent variable with respect to a single independent variable. That is a differ-
ential equation of order n is said to be ordinary differential equation if it is of the form:
dy d2 y dy 3 dy n
f (y, , 2 , 3 , · · · , n , t) = o
dt dt dt dt
for certain function f.

13
Definition 3.1.6. (Non-Dimensionalization Or Scaling )
It is a reasonable way of introducing dimensionless variables in to equation. Scaling an equation
means scaling all variables in the equation. To scale a variable u∗ we write the following form:

u∗ = U u where u∗ and U has the same dimension and u is dimension less.So from this we can
u∗
get dimensionless unit (i.e u = = 1)
U
Theorem 3.1.2 ([38]). Let us consider a linear ordinary differential equation which is given by
dy
ẏ(x) + p(x)y = + p(x)y = q(x). (3.1.2)
dx
We introduce integrating factor method,the integrating factor and its general solution, respectively,
are given by
R
p(x)dx
IF = e (3.1.3a)
1  R
p(α)dα

y(x) = R
p(x)dx
e q(x) + K , (3.1.3b)
e
where K is an arbitrary constant of integration.

3.1.1 Stability analysis of equilibrium points


Definition 3.1.7 ([39]). Given the autonomous system (3.1.1), a state x∗ is said to be an equilib-
rium point of the system if f (x∗ ) = 0.

Definition 3.1.8 ([40]). The solution x∗ is said to be stable if for every  > 0, there exists a
δ = δ() > 0 such that, | x∗ − x0 |< δ ⇒|x∗ − x(t)| < , t > t0 , t0 ∈ R, for every solution x(t) of
(3.1.1) with x(t0 ) = x0 .

Definition 3.1.9 ([40]). An equilibrium point x∗ is attracting if there is a δ > 0 such that
| x0 − x∗ |< δ ⇒ x(t) → x∗ as t → ∞, for every solution x(t) of (3.1.1) with x(0) = x0 . (All
trajectories that start near x∗ approach it as t → ∞).

Definition 3.1.10 ([40]). An equilibrium point x∗ is asymptotically stable if it is stable and attract-
ing. In other words, the solution x∗ is said to be asymptotically-stable if:

(i) it is stable, and

(ii) there exists a constant δ > 0 such that for any solution x(t) of (3.1.1) satisfying
| x∗ − x(0) |< δ, then lim | x(t) − x∗ |= 0
t→∞

14
Definition 3.1.11 ([40]). An equilibrium point x∗ is said to be globally asymptotically stable if it
is asymptotically stable for all initial condition x0 ∈ Rn .

Definition 3.1.12 ([40]). An equilibrium point x∗ of the model (3.1.1) is said to be locally asymp-
totically stable if it is locally stable and every trajectory that starts sufficiently close to x∗ tend
towards x∗ as t → ∞. A steady state x∗ which is not stable is said to be unstable.

Definition 3.1.13 ([40]). An equilibrium point of a given dynamical system is stable means all
solution curves of the equation attracts towards the equilibrium point, while an equilibrium point
is unstable means all solution curves of the dynamic system go away from the equilibrium point.

Definition 3.1.14 ([39]). An equilibrium point x∗ is globally stable if all trajectories converge to
x∗ i.e lim x(t) = x∗ .
t→∞

Definition 3.1.15 ([39]). A function V: Rn → R is positive definite if

i. V (x) > 0 for all x 6= x∗ ,

ii. V (x∗ ) = 0 where x∗ is an equilibrium point of (3.1.1).

Linearization Technique
Mathematically, the stability of equilibrium can be analyzed using the linearized system at the
equilibrium point.

Definition 3.1.16 ([41]). The Jacobian matrix associated to the system (3.1.1) at the equilibrium
point x∗ , which is denoted by Df (x∗ ), is given by the matrix
 ∂f1 (x) ∂f1 (x) ∂f1 (x) 
∂x1 ∂x2
... ∂xn
∂f2 (x) ∂f2 (x) ∂f2 (x)
∂fi (x∗ )
  
∗ ∂x1 ∂x2
... ∂xn

Df (x ) = =
 
∂xj

 : : 
∂fn (x) ∂fn (x) ∂fn (x)
∂x1 ∂x2
... ∂xn |x=x∗

where i, j = 1, 2, ..., n.

Proposition 3.1.1 ([39]). An equilibrium point x∗ of the dynamical


 system
 (3.1.1) is locally asymp-
∂fi (x∗ )
totically stable if all eigenvalues of the Jacobian Df (x∗ ) = ∂xj
evaluated at x∗ are negative.
The equilibrium x∗ is unstable if at least one of the eigenvalues of Df (x∗ ) is positive.

15
3.1.2 Routh-Hurwitz stability criterion
Routh Hurwitz criterion is an important criteria that gives necessary and sufficient condition of for
all of the roots of the characteristics polynomial (with real coefficients). Routh-Hurwitz Criteria is
used to determine asymptotic stability of an equilibrium point for non-linear system of differential
equations. Consider the characteristic equation of degree n given by

P (λ) = λn + a1 λn−1 + a2 λn−2 + ... + an−1 λ + an = 0,

where all the polynomial coefficients ai , for i = 1, 2, . . . , n are real constant. Define the n × n
Hurwitz matrix using the coefficients ai of the characteristic polynomial.

 
a1 1 0 0 ... 0
a3 a2 a1 1 ... 0
" #  
 
a1 1  
H1 = (a1 ), H2 = , ..., Hn =  a5 a4 a3 a2 ... 0 ,
0 a2 
 .. .. .. .. .. ..


 . . . . . . 
0 0 0 0 . . . an
where aj = 0 if j > n. All of the roots of the polynomial P (λ) are negative or have negative real
part if and only if the determinant of all Hurwitz matrix are positive.

det(Hj ) > 0, for j = 1, 2, ..., n.

For polynomials of degree n = 2, 3, 4 and 5 the Routh-Hurwitz criteria are summarized as:

n = 2 : a1 > 0 and a2 > 0


n = 3 : a1 > 0; a3 > 0 and a1 a2 > a3
n = 4 : a1 > 0; a3 > 0; a4 > 0 and a1 a2 a3 > a23 + a21 a4
n = 5 : ai > 0; for i = 1, 2, 3, 4, 5; a1 a2 a3 > a23 + a21 a4 and
(a1 a4 − a5 )(a1 a2 a3 − a23 − a21 a4 ) > a5 (a1 a2 − a3 )2 + a1 a25 .

Global stability
Theorem 3.1.3 ([42]). (Lyapunov Stability theorem)
Let x∗ be an equilibrium solution of (3.1.1)) and suppose that we can find a Lyapunov function i.e
a continuously differentiable, real valued function V(x) such that
(i) V (x) > 0 for all x 6= x∗ , and V (x∗ ) = 0. (V is positive definite),
(ii) dVdx(x) < 0 for all x 6= x∗ .
Then x∗ is globally asymptotically stable, i.e for all initial conditions, x(t) → x∗ as t → ∞.

16
Theorem 3.1.4 ([48]). (Castillo Chavez theorem)
Assume that the system (3.1.1) can be rewritten in the form
dM1
= F (M1 , M2 ) (3.1.4a)
dt
dM2
= G(M1 , M2 ) (3.1.4b)
dt
where the M1 ∈ Rm (represents the classes of non-smokers individuals) and M2 ∈ Rn−m (rep-
resents the classes of smokers individuals). Assume that G(M1 , 0) = 0 and let E0 = (M1∗ , 0)
be a steady state of (3.1.1) (the smoking free equilibrium point). If the following conditions are
satisfied:
dM1
a. For the system dt
= F (M1 , 0), the steady state M1∗ is globally asymptotically stable.

b. G(M1 , M2 ) = AM2 − Ĝ(M1 , M2 ), Ĝ(M1 , M2 ) ≥ 0 for (M1 , M2 ) ∈ Ω, where A is a Metzler


matrix (the off diagonal elements of A are non-negative) and Ω is our region where the our
model makes biological sense.
Then the steady state E0 = (M1∗ , 0) is a globally asymptotically stable for the system (3.1.4)
provided that the R0 < 1.

3.2 Optimal control theory


Optimal control theory is branch of a mathematical optimization deals with finding a control for a
dynamical system over period of time such that an objective function is to be optimized.
Lev Pontryagin and Richard Bellman’s work in the 1950s, which came after Edward J. Mc-
Shane’s contributions to the calculus of variations, is substantially responsible for the formulation
of the technique. An effective mathematical tool for making decisions concerning complicated bi-
ological circumstances is optimal control theory.To reduce the number of smokers and the expense
of executing the plan, for example, what proportion of the population should the number of smok-
ers minimize over time in a particular dynamic model. It also addresses the issue of determining
a control rule for a certain system that meets a particular optimality requirement. The process
of choosing a dynamic system’s control and state trajectories over time in order to minimize the
number of smokers and cost of implementing to prevent a performance index is known as optimal
control (OC)[43].
As a mathematical technique, optimal control theory can be used in a variety of ways to solve a
given issue optimally. Consider the controlled dynamical system to understand it.
x0 (t) = g(t, x(t), u(t)), t ∈ [t0 , T ]
(3.2.1)
x(t0 ) = x0 , x(T ) = xT .

17
In the dynamical system (3.2.1), x(t) refers to the state variable in a specified time t, x0 is the
initial condition of the state variables, xT is the final condition of the state variable, T is the
terminal intervention time and u(t) refers to time dependent control parameter. The controls affects
the dynamical system with the main purpose of minimizing or maximizing the cost functional(or
performance index).

3.2.1 The cost function


The purpose is to minimization or maximization with respect to the optimal issue and the initial
condition according to the cost function, which is an equation that describes the output such that:
Z T
Maximize/minimize J(t, x, u) = f (t, x(t), u(t))dt,
t0 (3.2.2)
0
subject to x (t) = g(t, x(t), u(t)),

where x ∈ X ⊂ Rn , with initial and terminal condition in (3.2.1). A control set is a set of points
characterized by u(t) ∈ U ⊂ Rm , m ∈ N. A control variable u(t) is said to be an admissible
control if it is piecewise continuous defined on some time interval t0 ≤ t ≤ T with range in the
control region U, u(t) ∈ U, ∀t ∈ [t0 , T ] [44].

3.2.2 Pontryagin’s Maximum Principle (PMP)


Thus according to Pontryagin’s Maximum Principle, an optimal trajectory would satisfy a neces-
sary condition.This result is considered as one of the most important results of Mathematics in the
20th century. Finding a piecewise continuous control u(t) and the corresponding state x(t) that
optimizes a cost functional(or performance index) J[x(t), u(t)] constitutes an optimum control
problem. The Pontryagin’s Maximum Principle, a first-order requirement for locating the optimal
solution, is a pillar of the majority of mathematical models that employ optimal control theory. For
simplicity, it is given below.

Theorem 3.2.1 ([44]). (Pontryagin’s Maximum Principle (PMP))


Let u(t) be a time optimal control and X(t) be the corresponding response of the system. Then

18
there exists a function λ(t) : [t0 , T ] → Rn , such that:

∂H(x, u, λ)
x0 (t) = , x(t0 ) = x0 (State Equations) (3.2.3)
∂λ
∂H(x, u, λ)
λ0 (t) = − (Co-state Equations) (3.2.4)
∂x
λ(T ) = 0 (Transversality condition) (3.2.5)
∂H
H(x∗ , u∗ , λ∗ ) = max H(x∗ , u, λ∗ ) or = 0. (3.2.6)
u∈U ∂u
Where H(x, u, λ) = f (t, x, u) + λ(t)g(t, x, u) is called the Hamiltonian of the optimal con-
trol problem. Equation (3.2.6) is given in two forms because, when the Hamiltonian is differ-
entiable with respect to u, the condition ∂H
∂u
= 0 can often be used to replace H(x∗ , u∗ , λ∗ ) =
maxu∈U H(x∗ , u, λ∗ ).

3.2.3 Numerical computation of optimal control problems


The solutions to many optimal control problems cannot be found by analytical means. Solving
optimal control problems in the form of (3.2.2) numerically involves finding piecewise continuous
function ui (t) that optimize the objective functional. To do this, one can use total-enumeration
methods or linear programming approaches, but it is that any solution to the issue must fulfill
the state and co-state equations as well as the optimality requirements. The control variable u(t)
may usually be explicitly expressed by manipulating the optimality conditions. Once this is done,
the control variable can be inserted into the state equations (3.2.3) and (3.2.4) to create a two-
point boundary value problem. Numerical techniques for addressing boundary value problems
and ordinary differential equations can then be used to solve the succeeding two-point boundary
value problems. The Forward-Backward Sweep technique is the numerical approach used in this
research to solve the associated optimum control problems. A basic description of this strategy
was provided by Lenhart and T.[44], and it is given here for convenience.

Step 1. Make the first-pass prediction of u over the interval. Save the first assumption as u.

Step 2. Using the initial condition x(t0 ) = x0 and the stored values for u, solve x forward in time
according to it differential equation in the optimality system.

Step 3. Using the transversality condition λ(T ) = 0 and the stored values for x and u, solve λ
backward in time according to its differential equation in the optimality system.

Step 4. Update the control u by entering the new x and λ values into the characterization.

19
Step 5. Verify convergence: if the variables are sufficiently close to the corresponding in the pre-
vious iteration, then output the current values as solutions, else return to Step 2.

3.3 Basic concepts in epidemiological modeling


Definition 3.3.1 ([45]). Epidemiology is the study of the occurrence and consequences of health-
related conditions or occurrences in certain groups, as well as the application of this study to the
prevention and management of health issues.

Smoking free equilibrium point (SFE)


We say that the population is smoking free if a tobacco smoking habit can disappear completely
from the population. In epidemiological modeling, a smoking free equilibrium (SFE) is a steady
state in which the coordinates in the smoker compartments are zero. If the smoking free equilib-
rium is stable, then it is expected that the smoking habit in population will be free over time.

Smoking present equilibrium point (SPE)


Unlike the smoking free equilibrium, a smoking present equilibrium point (SPE) is a steady state
in which at least one of its coordinates in the smoker compartment is nonzero. It is a steady state
solution where the smoking habit remain in the population and continue until take prevention.

3.3.1 Smoking Generation Number/ Basic Reproduction Number


The smoking generation number is one of the most useful threshold parameter or invariant that
plays critical role in the control and eradicate habit of smoking from population.

Definition 3.3.2 ([46]). The Smoking Generation Number R0 is defined as the average number of
new smokers that are produced when one smoker individual is introduced into a group of potential
individuals. Basically,it is used to predict the global dynamics of the smoking habit whether the
smoking is controllable or not.R0 is computed using the method of next-generation matrix.

If R0 > 1, then each single smoker individual can affect more than one person (in average).
Hence the total number of smokers will continuously increase and the smokers will become en-
demic (it will remain in the population forever). However if R0 < 1, then each single smokers
individual can affect zero or at most one person (in average). Over time the number of newly
smokers will decrease and the population can become smoker free. Therefore it is always expected
that the smokers free equilibrium is stable when R0 < 1 and unstable when R0 > 1.

20
The SFE& SPE exchange their stability at R0 = 1 which implies that the smoking habit remains
in the population at a constant rate.
The basic reproduction number R0 is computed using the method of next-generation matrix.
Assume there are n smokers compartment and m non-smokers compartment. Let x ∈ Rn and y ∈
Rm be sub populations in each compartment. Further denote by Fi the rate of secondary smokers
increase in the ith compartment(or rate of appreance of new smokers in the ith compartment),and
Vi the rate of transfer of individuals in and out of the ith compartment. The compartment model
can be written as:
dxi
= Fi − Vi (x, y), i = 1, 2, ......n
dt
dyi
= Gj (x, y), j = 1, 2, ......m
dt
Denoting F and V the matrices,
∂Fi
F =( (0, y ∗ ))1≤i,j≤n
∂xj
∂Vi
V =( (0, y ∗ ))1≤i,j≤n
∂xj
then F V −1 is called the next generation matrix. The basic reproduction number is the spectral
radius (dominant eigenvalue) of the matrix F V −1 that is R0 = ρ(F V −1 ).

3.3.2 Sensitivity Analysis


Sensitivity analysis is used to determine the relative importance of model parameters to smoking
habit progression. We will perform the analysis by calculating the sensitivity indices of the smok-
ing generation number, R0 because it determines whether or not the smoking habit will spread in
the population. Sensitivity analysis is commonly used to determine the relevant changes in a state
variable when a parameter changes.
We offer a technique for doing sensitivity analysis known as normalized forward sensitivity
index of a variable, which is defined as the ratio of the relative change in the variable to the relative
change in the parameter. The sensitivity may also be defined using partial derivatives when the
variable is a differentiable function of the parameter[47].

Definition 3.3.3 ([47]). The relative sensitivity coefficient (or normalized forward sensitivity index)
of a quantity Ψ with respect to a parameter p is defined as:
∂Ψ p
Ψp = × . (3.3.1)
∂p Ψ

21
Chapter 4

Mathematical Model Formulation and


Analysis

4.1 The Existing Mathematical Model


The mathematical model used to study the dynamics of smoking that was proposed by Omar et
al. [1]. In this model the entire population is divided into five sub-populations: potential smokers
(denoted by P) are those who are not smokers but might become smokers within future. Occasional
smokers(denoted by L) are those who use smoke cigarette less frequently but do not grow to
addicted stage . Chain smokers (denoted by S) are those who are addicted by smoking and spent
most of their time on it.Temporarily quit smokers (denoted QT )who were smokers but quit smoking
for a while.Permanent quit smokers(denoted QP ) who were smokers but completely quit smoking.
People will join the potential smoker’s P,at a constant recruitment rate of Λ.Some of people will
leave from compartment at a constant natural death rate a well as constant death due to smoking by
µ ,δ,respectively from all compartment. Individuals means non-smokers start using smoke tobacco
by the peer pressure effective with contact rate β between potential & occasional smokers .σ is the
converting rate at which occasional smokers become chain smokers.people will join chain smoker’s
class,S with the rate of σL& αQT ,where α is ratio of those who temporarily quit smokers start to
smoke again.People will enter in to temporarily quitter’s compartment ,QT with a rate γ(1 − η)S&
leave with the rate αQT ,where ,(1 − η) is the fraction of chain smoker’s who temporarily quit
smoking (at rate γ),where the remaining fraction (i.e η)of heavy smokers who permanently quit
smoking (at a rate γ). The permanent quitter’s,Qp are increased when chain smokers,S become to
stop smoking with rate γηS . The dynamics of smoking habit is governed by the following system

22
QT

હQT
(𝜇 + δ)QT
𝚲 2𝛃𝑃𝐿/(𝑃 + 𝐿) 𝝈𝐿
P L S
(𝜇 + δ)𝑃 (𝜇+ δ )L
QP

Figure 4.1: Flow diagram of the existing model.

of non-linear ordinary differential equations:


 
dP PL
= Λ − 2β − (µ + δ)P, (4.1.1)
dt P +L
 
dL PL
= 2β − (µ + δ)L, (4.1.2)
dt P +L
dS
= σL − (µ + δ + γ)S + αQT , (4.1.3)
dt
dQT
= γ(1 − η)S − (µ + δ + α)QT , (4.1.4)
dt
dQP
= γηS − (µ + δ)QP , (4.1.5)
dt
with initial conditions P (0) = P0 > 0, L(0) = L0 ≥ 0, S(0) = S0 ≥ 0, QT (0) = QT 0 ≥ 0,
QP (0) = QP 0 ≥ 0.

23
4.2 The Modified Mathematical Model
In this section, we modify the existing model Omar et al. [1]to study the dynamics of smoking habit
spread in a population. Our model is an extension of the existing P LSQT QP model for smoking
dynamics of cigarette. In the present model we introduced total natural birth rate ΦN recruited in
to potential smokers in closed system .When susceptible people contact with smokers,will affected
by fraction of (1 − u1 ) due to the without proper use of smokers in population of rate β1 &β2
.The remaining fraction u1 of smoking is remain in smokers with out affect others.People will join
chain smokers class S,with rate of σ(1 − u2 )L,while the remaining fraction u2 ,with rate σ join to
permanently quit, QP smokers.After time people will smoke again with fraction of 1 − ε with rate
α,and also the remaining fraction ε will join permanently quit smokers with α rate.

4.2.1 Mathematical Model Formulation


We formulate a deterministic compartmental model for the dynamics of spreading smoking cigarette.
The model divides the total population in to five sub-classes according to their smoking status. The
following assumptions are crucial in the formulation of the model:

1. The population under study is varying with time.

2. The recruitment rate Φ is different from the natural death rate µ so that the total population
is not constant.

3. we assumed that individuals who have permanently quit smokers do not become potential
smokers again.

4. Legal age could be considered to start smoking, in the model.

5. Sex,and race do not affect the probability of being smokers.

6. The natural mortality rate µ are assumed to be the same for all the compartments.

7. All parameters in the model being non-negative.

8. A non-smoker is not considered as a smoker just because his or her interaction with smokers.

9. Temporarily quite smokers could be allowed to fall back into chain smokers’ class.

10. Individuals in the smokers class have fully developed smoking habit and can affect other
people.

24
11. The smoking habit spread in a closed environment ,but only immigration happen in environ-
ment.

12. We assume that Natural birth rate Φ is greater than death due to smoking cigarette δ

Table 4.1: The state variables and their descriptions.

Variables Description of the state variables


P(t) potential individuals those are non-smokers
L(t) Individuals those smoke less frequently
S(t) Individuals those are addicted by smoking
QT (t) Individuals who are quit smoking for some time
QP (t) Individuals who are quit completely smoking

The flow chart of the modified mathematical model is illustrated in Figure (4.2).
𝛂(1 − ε)QT

𝜇 QT
QT

(1 − u1)(઺1𝑃𝐿 + ઺2𝑃𝑆)/𝑁
𝚽N
P L 𝝈(1- u2)𝐿
S
𝜇𝑃
𝜇L
𝜇QP
QP

𝝈 u2L

Figure 4.2: The flow chart for the modified mathematical model of smoking dynamics.

25
Based on our assumptions and the flow chart (4.2), the modified mathematical model for the
dynamics of smoking cigarette is given by the following deterministic system of non-linear differ-
ential equations:
 
dP β1 P L + β2 P S


 = ΦN − (1 − u1 ) − µP,



 dt N
  

 dL β1 P L + β2 P S

 = (1 − u1 ) − (µ + σ)L,
 dt N



dS
= σ(1 − u2 )L − (µ + δ + γ)S + α(1 − ε)QT ,


 dt
dQT


= γ(1 − η)S − (µ + α)QT ,





 dt
 dQP = σu L + γηS + αεQ − µQ ,



2 T P
dt
with non-negative initial conditions P (0) = P0 > 0, L(0) = L0 ≥ 0, S(0) = S0 ≥ 0, QT (0) =
QT 0 ≥ 0, QP (0) = QP 0 ≥ 0, where
dN (t)
dt
= dPdt(t) + dL(t)
dt
+ dS(t)
dt
+ dQdtT (t) + dQdt
P (t)
= ΦN (t) − µN (t) − δS(t)

26
Table 4.2: Parameters of the modified mathematical model and their descriptions.

Parameter Description of the parameter


Λ Recruitment rate of potential smokers
γ rate of removal from smokers to temporarily quit & permanent quit smokers
µ Natural death rate
β1 The rate of contact of potential smokers with occasional smokers
δ Induced death rate due to smoking
β2 The rate of contact of potential smokers with chain smokers
α Rate of return back from quit smoker to smokers& permanent quit
Φ Constant recruitment rate
N Total population
ΦN Total natural birth rate
µ Natural death rate
σ The rate of conversion from occasional smokers to smokers class& permanent quit
η The probability of become temporary quit smokers & permanently quit smokers
ε The probability of conversion from temporarily quit smokers & become to permanently quit
1 − u1 The fraction of impact of smokers on non-smokers
u1 The remaining fraction of impact of smoking on themselves
1 − u2 Fraction of occasional smokers become chain smokers
u2 The remaining fraction of occasional smokers become quit smokers

27
4.3 Scaling of Modified Model Equation
To simplify the analysis of the model given in above we modified model equations ,we work with
fractional quantities instead of actual populations by scaling the population of each class by the
total species population, namely P (t), L(t), S(t), QT (t), QP (t) by
P (t) L(t) S(t) QT (t) QP (t)
p(t) = , l(t) = ,s(t) = , qT (t) = ,qP (t) = with,
N (t) N (t) N (t) N (t) N (t)
N1 (t) = p(t) + l(t) + s(t) + qT (t) + qP (t) = 1 and let t = τ , with this representation we do have
a new non-linear system of differential equation which is equivalent to original one .

Thus,we can write a system for the fraction variables (p, l, s, qT , qP ) by direct differentiation
.For instance
dp P 0N − P N 0

0
 = p = ,
N2



 dt
dl L0 N − LN 0


0
= l = ,


 2


 dt N
ds S 0 N − SN 0

= s0 = ,

 dt N2
dqT Q0T N − QT N 0


 0

 = q T = ,


 dt N2
0 0
 dqP = q 0 = QP N − QP N



P
dt N2
After computing,the state variables p,l,s,qT & qP satisfy the following system of non-linear differ-
ential equations:
dp
= Φ − (1 − u1 )(β1 pl + β2 ps) − (Φ − δs)p, (4.3.1a)
dt
dl
= (1 − u1 )(β1 pl + β2 ps) − (Φ + σ − δs)l, (4.3.1b)
dt
ds
= σ(1 − u2 )l − (Φ + δ + γ − δs)s + α(1 − ε)qT , (4.3.1c)
dt
dqT
= γ(1 − η)s − (Φ + α − δs)qT , (4.3.1d)
dt
dqP
= σu2 l + γηs + αεqT − (Φ − δs)qP , (4.3.1e)
dt
with non-negative initial conditions p(0) = p0 > 0, l(0) = l0 ≥ 0, (0) = s0 ≥ 0, qT (0) = qT 0 ≥ 0,
qP (0) = qP 0 ≥ 0

28
4.4 Qualitative Analysis of the Modified Model
In this section, we present some basic qualitative properties of the scaled modified model. These
analysis seek to show that the modified model is epidemiologically appropriate in the sense the
model and its predictions make sense. These analysis include finding the set inside which the model
can be sufficiently studied (i.e., the invariant region); local and global stability of equilibrium points
of the model (4.3.1).

4.4.1 Well-posedness
Since all the functions on the right hand side of the system (4.3.1) are continuously differentiable.
Thus, the existence and uniqueness of the solutions is established by the Picard’s theorem (3.1.1) .
Now, we show the positivity and boundedness of solutions in the following .

Theorem 4.4.1. (Existence and Uniqueness of the Solution)



Let Q = (p, l, s, qT , qP ) ∈ R5+ denotes the region defined

|t − t0 | 6 a, ||x − x0 || 6 b, x = (x1 , x2 , ...xn ), x0 = x01 , x02 , ...x0n (4.4.1)

and suppose that f(t,x) satisfies the Lipschitz condition

||f (t, x1 ) − f (t, x2 )|| 6 L||x1 − x2 || (4.4.2)

when the pairs (t, x1 )&(t, x2 ) belongs to Q where L is a positive constant.Then,there exists a
constant ζ > 0 such that there exists a unique continuous vector solution x(t) of condition 4.4.1 in
the interval |t − t0 | < ζ .It is important to note that condition 4.4.1 is satisfied by the requirement
∂fi
that ,i, j = 1, 2, ...n is continous and bounded in Q.
∂xj
∂fi
Proof Let Q denote the region above.It serves to show that ,i, j = 1, 2, 3, 4, 5 are continous
∂xj
and bounded in Q. At present,from 4.4.1,Q = (p, l, s, qT , qP ) where

|p − p0 | 6 b1 , |l − l0 | 6 b2 , |s − s0 | 6 b3 , |qT − qT 0 | 6 b4 , |qP − qP 0 | 6 b5 (4.4.3)

29
Let
dp
f1 (p, l, s, qT , qP ) = = Φ − (1 − u1 )(β1 pl + β2 ps) − (Φ − δs)p,
dt
dl
f2 (p, l, s, qT , qP ) = = (1 − u1 )(β1 pl + β2 ps) − (Φ + σ − δs)l,
dt
ds
f3 (p, l, s, qT , qP ) = = σ(1 − u2 )l − (Φ + δ + γ − δs)s + α(1 − ε)qT ,
dt
dqT
f4 (p, l, s, qT , qP ) = = γ(1 − η)s − (Φ + α − δs)qT ,
dt
dqP
f5 (p, l, s, qT , qP ) = = σu2 l + γηs + αεqT − (Φ − δs)qP
dt
Consider the partial derivatives with respect to each state variable
∂f1 ∂f1
= δs − Φ − (1 − u1 )(β1 l + β2 s), = −β1 (1 − u1 )p,
∂p ∂l
∂f1
= δp − β2 (1 − u1 )p,
∂s
∂f1 ∂f1
= 0, = 0,
∂qT ∂qP
∂f2 ∂f2 ∂f2 ∂f2
= (1 − u1 )(β1 l + β2 s), = β1 (1 − u1 )p + δs − (Φ + σ), = β2 (1 − u1 )p + δl, = 0,
∂p ∂l ∂s ∂qT
∂f2 ∂f3 ∂f3 ∂f3
= 0, = 0, = σ(1 − u2 ), = 2δs − (Φ + δ + γ),
∂qP ∂p ∂l ∂s
∂f3 ∂f3 ∂f4 ∂f4 ∂f4
= α(1 − ε), = 0, = 0, = 0, = γ(1 − η) + δqT ,
∂qT ∂qP ∂p ∂l ∂s
∂f4 ∂f4 ∂f5 ∂f5 ∂f5 ∂f5 ∂f5
= δs − (Φ + α), = 0, = 0, = σu2 , = γη + δqP , = αε, = δs − Φ
∂qT ∂qP ∂p ∂l ∂s ∂qT ∂qP
(4.4.4)

30
Now,by substituting 4.4.3 in to 4.4.4,we can get the following
∂f1
| | = |δs − Φ − (1 − u1 )(β1 l + β2 s)| 6 (|δs| + | − Φ − (1 − u1 )(β1 l + β2 s)|)
∂p
6 (1 − u1 )(β1 + β2 )(b2 + b3 + l0 + s0 ) + Φ + δ(s0 + b3 ) < ∞,
∂f1 ∂f1
| | = | − β1 (1 − u1 )p| = β1 (1 − u1 )p 6 β1 (1 − u1 )(b1 + p0 ) < ∞, | | = 0 < ∞,
∂l ∂qT
∂f1 ∂f2
| | = 0 < ∞, | | = |(1 − u1 )(β1 l + β2 s)| 6 (1 − u1 )(β1 + β2 )(l0 + b2 + s0 + b3 ) < ∞,
∂qP ∂p
∂f2
| | = |β1 (1 − u1 )p + δs − (Φ + σ)|
∂l
6 β1 (1 − u1 )p + δs + Φ + σ 6 β1 (1 − u1 )(b1 + p0 ) + δ(b3 + s0 ) + Φ + σ < ∞,
∂f2
| | = |β2 (1 − u1 )p + δl| 6 β2 (1 − u1 )(b1 + p0 ) + δ(l0 + b2 ) < ∞,
∂s
∂f2
| | = 0 < ∞,
∂qT
∂f2
| |=0<∞
∂qP
And we continue in the same manner for the f3 , f4 &f5 .Thus,the system of equation 4.3.1 are
differentiable with respect to all state variables & all the partial derivatives of the system equa-
tion4.3.1 are continous and bounded.
Therefore,by the existence & uniqueness theorem the modified mathematical model of the equation
system 4.3.1 has unique solution.

31
Theorem 4.4.2. (Positivity)
If p(0) > 0, l(0) ≥ 0, s(0) ≥ 0, qT (0) ≥ 0, qp (0) ≥ 0, then the solution (p(t), l(t), s(t), qT (t), qP (t))
of the dynamical system (4.3.1) is non-negative for all time t ≥ 0.

Proof. To show the positivity of the solution from scaled modified mathematical model system of
equation (4.3.1),now let us consider the first equation is :

dp
= Φ − (1 − u1 )(β1 pl + β2 ps) − (Φ − δs)p, (4.4.5)
dt
> −(1 − u1 )(β1 pl + β2 ps) + δs)p since Φ is positive
= −p((1 − u1 )(β1 l + β2 s) + δs),
dp
> − ((1 − u1 )(β1 l + β2 s) + δs) dt, Integrate both sides,
p
Z Z
dp
⇒ > − ((1 − u1 )(β1 l + β2 s) + δs) dt,
p
Z
ln p = − ((1 − u1 )(β1 l + β2 s) + δs) dt + c1 ,
R
elnp = e− (1−u1 )(β1 l+β2 s)+δs))dt+c1
,
R
p = e− ((1−u1 )(β1 l+β2 s)+δs)dt+c1
, p(0) = ec1 ,
T heref ore,
R
=⇒ p(t) = p(0)e− ((1−u1 )(β1 l+β2 s)+δs)dt
>0

32
Proof. To show the positivity of the solution from modified mathematical model system of equa-
tion (4.3.1),from second equation is :

dl
= (1 − u1 )(β1 pl + β2 ps) − (Φ + σ − δs)l,
dt
(4.4.6)
=(1 − u1 )β1 pl + (1 − u1 )β2 ps − (Φ + σ − δ)l,
=(1 − u1 )β1 pl + (1 − u1 )β2 ps − (Φ + σ − δs)l
= ((1 − u1 )β1 p − (Φ + σ − δs)) l + (1 − u1 )β2 ps
dl
⇒ + (Φ + σ − δs − β1 (1 − u1 )p)l = β2 (1 − u1 )ps
dt
dl
+ M (t)l =β2 (1 − u1 )ps where, M (t) = Φ + σ − δs − β1 (1 − u1 )p
dt R
W e use integrating f actor IF =e M (t)dt both side , we have
dl R M (t)dt R R
e + M (t)le M (t)dt =β2 (1 − u1 )pse M (t)dt
dt
d R R
⇒ (le M (t)dt ) =β2 (1 − u1 )pse M (t)dt Integrate both side,we get
dt Z 
R R 
M (t)dt M (t)dt
l(t)e = β2 (1 − u1 )pse dt + c2 ,

Divide both side f or integrating f actor


R
Z  R  
− M (t)dt M (t)dt
⇒ l(t) =e β2 (1 − u1 )pse dt + l(0)

since l(0) = c2
T heref ore,
=⇒ l(t) >0(Since, any exponential f unction is non − negative.)

33
Proof. To show the positivity of the solution from modified mathematical model system of equa-
tion (4.3.1),from third equation is :

ds
= σ(1 − u2 )l − (Φ + δ + γ − δs)s + α(1 − ε)qT , (4.4.7)
dt
> −(Φ + δ + γ − δs)s + α(1 − ε)qT , since l(t) is non − negative, so
ds
+ (Φ + δ + γ − δs)s > α(1 − ε)qT
dt
ds
+ Z(t)s > α(1 − ε)qT , where Z(t) = (Φ + δ + γ − δs)
dt R
By using the integrating f actor IF = e Z(t)dt
ds R Z(t)dt R R
e + Z(t)se Z(t)dt > α(1 − ε)qT e Z(t)dt
dt 
d seZ(t)dt
⇒ > α(1 − ε)qT eZ(t)dt dt
dt
Integrate both, weget
R
Z
⇒ se Z(t)dt > α(1 − ε)qT eZ(t)dt dt + c3


Divide both side f or integrating f actor, we get


Z 
−Z(t)dt Z(t)dt

s(t) > e α(1 − ε)qT e dt + s(0) > 0

34
Proof. To show the positivity of the solution from modified mathematical model system of equa-
tion (4.3.1),from fourth equation is :

dqT
= γ(1 − η)s − (Φ + α − δs)qT , (4.4.8)
dt
> −(Φ + α − δs)qT , since s(t) is positive solution
dqT
⇒ > −(Φ + α − δs) Integrate both
qT
Z Z
dqT
⇒ > − (Φ + α − δs)dt
qT
Z
lnqT > − (Φ + α − δs) dt
R
elnqT > e− (Φ+α−δs)dt
R
qT (t) > e− (Φ+α−δs)dt+c4
R
qT >qT (0)e− (Φ+α−δs)dt
where c4 = qT (0)
T here qT (t) >0

Proof. To show the positivity of the solution from modified mathematical model system of equa-
tion (4.3.1),from last equation is :

dqP
= σu2 l + γηs + αεqT − (Φ − δs)qP , (4.4.9)
dt
> −(Φ − δs)qP , since f rom above l(t) , s(t) &qT are non − negative so
dqP
> −(Φ − δs)
qP
Integrate both sides
Z Z
dqP
> − (Φ − δs)dt
qP
Z
ln qP > −(Φ − δs)dt + c5
R
qP > e− (Φ−δs)dt+c5
, qP (0) = ec5
R
=⇒ qP (t) = qP (0)e− (Φ−δs)dt
>0

There for by above proof we can conclude that the solution of the modified mathematical model
p(t), l(t), s(t), qT (t)&qP (t)of the equation (4.3.1)are positive for all t > 0

35
Theorem 4.4.3. (Boundedness)
There exists a positively invariant region Ω in which the solution (p(t), l(t), s(t), qT (t), qP (t)) of
the dynamical system (4.3.1) is bounded.

Proof. The positivity has already been established by Theorem (4.4.2). For this model the total
population of scaled model is N1 (t) = p(t) + l(t) + s(t) + qT (t) + qP (t) = 1.
The rate of change of total population of scaled become :
dN1 dp dl ds dqT dqP
= + + + + = 0,
dt dt dt dt dt dt
dN1
= Φ + δsN1 + δs2 − ΦN1 − δs = 0,
dt
Φ + δsN1 + δs2 − ΦN1 − δs > Φ − ΦN1 − δs

⇒ Φ − ΦN1 − δs 6 0
Φ − ΦN1 − δN1 > 0 since s 6 N1
Φ > ΦN1 + δN1 = (Φ + δ)N1
Φ
> N1
(Φ + δ)
Φ
⇒ N1 6
(Φ + δ)
Φ
There for, p(t) + l(t) + s(t) + qT (t) + qP (t) 6 (Φ+δ) .This implies that total population after scaled
remain constant. Thus, the feasible solution set of the system (4.3.1) remain in the region Ω:
 
5 Φ
Ω = (p(t), l(t), s(t), qT (t), qP (t)) ∈ R+ : 0 6 N1 6 .
(Φ + δ)

4.4.2 Equilibrium Point


The equilibrium point of the system is solution which is obtained by setting the left side of the
system equation (4.3.1) is equal to zero:

Φ − (1 − u1 )(β1 pl + β2 ps) − (Φ − δs)p = 0


(1 − u1 )(β1 pl + β2 ps) − (Φ + σ − δs)l = 0
σ(1 − u2 )l − (Φ + δ + γ − δs)s + α(1 − ε)qT = 0
γ(1 − η)s − (Φ + α − δs)qT = 0
σu2 l + γηs + αεqT − (Φ − δs)qP = 0.

36
There are two steady states for the system (4.3.1): the smoking free- equilibrium E0 and smoking
persist equilibrium E1∗ . The smoking-free equilibrium point of our model is obtained by setting
the smoking state variables l = 0, s = 0 qT = 0and qP = 0. Thus, the smoking free equilibrium
point is given by
E0 = p0 , l0 , s0 , qT0 , qP0 = (1, 0, 0, 0, 0) .


The existence of the smoking present equilibrium point E ∗ will be presented after.

4.4.3 Basic Reproduction Number


The basic reproduction number, which is denoted by R0 , and defined as the average number of
secondary smokers produced by a single smokers individual in a completely susceptible popu-
lation(potential smokers). Using the next generation matrix method [55], the basic reproduction
number R0 can be calculated from the relation R0 = ρ(F V −1 ). Let F be the vector for the new
smokers and V be the vector for the transfer of individuals into and out of the smokers compart-
ments (i.eVi = Vi− − Vi+ ). Let x = (l, s) is the smoker’s class , where it is following :
Smokers compartment class
dl


 = (1 − u1 )(β1 pl + β2 ps) − (Φ + σ − δs)l,
dt
 ds = σ(1 − u2 )l − (Φ + δ + γ − δs)s + α(1 − ε)qT

dt
And also non-smokers compartment class
dp


 = Φ − (1 − u1 )(β1 pl + β2 ps) − (Φ − δs)p,


 dt
dqT

= γ(1 − η)s − (Φ + α − δs)qT ,

 dt
 dqP = σu2 l + γηs + αεqT − (Φ − δs)qP



dt
,then to find reproduction number of system become the following:
 
" # " # " #
F1 (1 − u1 )(β1 pl + β2 ps) V1
Fi = = and Vi = = (Φ + σ − δs)l .
 
F2 0 V2
−σ(1 − u2 )l + (k1 − δs)s − αk2 qT
where k1 = (Φ + δ + γ), k2 = 1 − ε.

The Jacobian matrix to F and V at smoking free equilibrium points are calculated by
  " #
∂Fi (E0 ) (1 − u1 )β1 (1 − u1 )β2
F = = ,
∂xj 0 0

37
  " #
∂Vi (E0 ) (Φ + σ) 0
V = = ,
∂xj −σ(1 − u2 ) (Φ + δ + γ)
Now ,we find the inverse of martices V ,V − 1
adj(V )
V −1 = where
|V | " #
(Φ + δ + γ) 0
adj(V ) =
σ(1 − u2 ) (Φ + σ)
and |V | = (Φ + σ)(Φ + δ + γ)
After some algebraic computations, the inverse of the matrix V is given by
" #
1
−1 (Φ+σ)
0
V = σ(1−u2 ) 1
.
(Φ+σ)(Φ+δ+γ) (Φ+δ+γ)
−1
The next-generation matrix G = F V is given by the following
" #
(1−u1 )β1 (1−u1 )β2 σ(1−u2 ) (1−u1 )β2
+
F V −1 = Φ+σ (Φ+σ)(Φ+δ+γ) (Φ+δ+γ)
,
0 0
The basic reproduction number R0 is the dominant eigenvalue of F V − 1 or the spectral radius
of F V − 1,that is R0 = ρ(F V − 1) . We find the eigenvalues of F V −1 by solving the characteristic
equation |F V −1 − λI| = 0 as λ1 = (1−uΦ+σ
1 )β1
+ (1−u1 )β2 σ(1−u2 )
(Φ+σ)(Φ+δ+γ)
, &λ2 = 0
Hence from those value the dominant one is

(1 − u1 )β1 (1 − u1 )β2 σ(1 − u2 )


λ1 = +
Φ+σ (Φ + σ)(Φ + δ + γ)
We can write the basic reproduction number of smoking dynamics as follows:
 
β1 β2 σ(1 − u2 )
R0 = (1 − u1 ) + . (4.4.10)
Φ + σ (Φ + σ)(Φ + δ + γ)

4.4.4 Stability Analysis of the Smoking Free Equilibrium

4.4.5 Stability of Smoking Free Equilibrium Point


In this subsection, we establish the local and global stability of the smoking free equilibrium (SFE)
of the system (4.3.1).

4.4.6 Local Stability of Smoking Free Equilibrium


According to linear stability analysis,an equilibrium state is stable if the eigenvalues of the Jaco-
bian at the equilibrium point have negative real part.

38
Theorem 4.4.4. The equilibrium solution E0 of the non-linear system of equation (4.3.1) is locally
asymptotically stable,if R0 6 1 at a3 > 0.

Proof. To determine the stability of smoking free equilibrium point,first construct the Jacobian
matrix of the modified system of equation (4.3.1) .
Jacobian matrix is given by
J(p, l, s, qT , qP ) =
 
−Φ − (1 − u1 )(β1 l + β2 s) + δs −(1 − u1 )β1 p δp − (1 − u1 )β2 p 0 0
(1 − u1 )(β1 l + β2 s) (1 − u1 )β1 p − (Φ + σ) + δs (1 − u1 )β2 p + δl 0 0
 
 
 

 0 σ(1 − u2 ) 2δs − (Φ + γ + δ) α(1 − ε) 0 .

0 0 γ(1 − η) + δqT δs − (Φ + α) 0
 
 
0 σu2 γη + δqP αε δs − Φ

Now,the Jacobian matrix at the smoking free equilibrium point E0 is given by


 
−Φ −(1 − u1 )β1 δ − (1 − u1 )β2 0 0
0 (1 − u1 )β1 − (Φ + σ) (1 − u1 )β2 0 0
 
 
 
J(1, 0, 0, 0, 0) = 
 0 σ(1 − u2 ) −(Φ + δ + γ) α(1 − ε) 0 .

0 0 γ(1 − η) −(Φ + α) 0
 
 
0 σu2 γη αε −Φ

The characteristic equation of the Jacobian matrix is defined by


| J(E0 ) − λI |= 0.
Here I is 5 by 5 identity matrix & λ is the eigenvalue.

−Φ − λ −(1 − u1 )β1 δ − (1 − u1 )β2 0 0


0 [(1 − u1 )β1 − (Φ + σ)] − λ (1 − u1 )β2 0 0
| J(E0 )−λI |= 0 σ(1 − u2 ) −(Φ + δ + γ) − λ α(1 − ε) 0 .
0 0 γ(1 − η) −(Φ + α) − λ 0
0 σu2 γη αε −Φ − λ

Thus the eigenvalues λi , i = 1, 2, 3, 4, 5 are obtained from characteristic equation of J(E0 ):

(Φ + λ)2 (λ3 + λ2 [3Φ + α + δ + γ + σ − β1 (1 − u1 )]


+ λ[(Φ + α)(Φ + δ + γ) + (Φ + σ)(Φ + δ + γ) + (Φ + σ)(Φ + α)
− β2 (1 − u1 )σ(1 − u2 ) − αγ(1 − η)(1 − ε)
− β1 (1 − u1 )(Φ + δ + γ) − β1 (1 − u1 )(Φ + α)] + β1 (1 − u1 )αγ(1 − η)(1 − ε)
+ (Φ + α)(Φ + δ + γ)(Φ + σ) − αγ(Φ + σ)(1 − η)(1 − ε) − σβ2 (1 − u1 )(Φ + α)(1 − u2 )
− β1 (1 − u1 )(Φ + α)(Φ + δ + γ) = 0

39
From this equation,we obtain the values for the λ to be λ1,2 = −Φ which are negative eigenvalues
and the remaining eigenvalues λ3 , λ4 &λ5 are the roots of cubic polynomial:

p(λ) = λ3 + a1 λ2 + a2 λ + a3 = 0 (4.4.11)

Where
a1 = 3Φ + α + δ + γ + σ − β1 (1 − u1 ) ,
a2 = (Φ + α)(Φ + δ + γ) + (Φ + σ)(Φ + δ + γ) + (Φ + σ)(Φ + α) − β2 (1 − u1 )σ(1 − u2 )
− αγ(1 − η)(1 − ε) − β1 (1 − u1 )(Φ + δ + γ) − β1 (1 − u1 )(Φ + α)),
a3 = β1 (1 − u1 )αγ(1 − η)(1 − ε) + (Φ + α)(Φ + δ + γ)(Φ + σ) − αγ(Φ + σ)(1 − η)(1 − ε) −
β2 (1 − u1 )(Φ + α)σ(1 − u2 ) − β1 (1 − u1 )(Φ + α)(Φ + δ + γ)

For the characteristic polynomial in equation 4.4.11,where n = 3,from Hurwitz criteria are

a1 > 0, a2 , a3 > 0

H1 = a1 , det(H1 ) = a1 > 0
!
a1 1
H2 = , det(H2 ) = a1 a2 > 0,
0 a2
 
a1 1 0
H(3 ) =  a3 a2 a1  , det(H3 ) = a1 a2 a3 − a23 > 0 ⇒ a1 a2 > a3
 

0 0 a3

Using the Routh-Hurwitz criteria,it can be seen that all the eigenvalues of the characteristic
equation 4.4.11 have negative real part if

a1 > 0, a2 > 0, a3 > 0&a1 a2 − a3 > 0

Now

a1 = 3Φ + α + δ + γ + σ − β1 (1 − u1 ) > 0, if 3Φ + α + δ + γ + σ > β1 (1 − u1 )
a2 = [(Φ + α)(Φ + δ + γ) + (Φ + σ)(Φ + δ + γ) + (Φ + σ)(Φ + α)
− β2 (1 − u1 )σ(1 − u2 ) − αγ(1 − η)(1 − ε) − β1 (1 − u1 )(Φ + δ + γ) − β1 (1 − u1 )(Φ + α)] > 0, if
(Φ + α)(Φ + δ + γ) + (Φ + σ)(Φ + δ + γ) + (Φ + σ)(Φ + α) >
[β2 (1 − u1 )σ(1 − u2 ) + αγ(1 − η)(1 − ε) + β1 (1 − u1 )(Φ + α) + β1 (1 − u1 )(Φ + δ + γ)]

40
And also,

a1 a2 − a3 > 0

This is the same as a1 a2 > a3


From Houth-Hurwitz criteria it follows that

a3 > 0

⇒β1 (1 − u1 )αγ(1 − η)(1 − ε) − β1 (1 − u1 )(Φ + α)(Φ + δ + γ) − β2 (1 − u1 )σ(1 − u2 )(Φ + α)


> αγ(1 − η)(1 − ε)(Φ + σ) − (Φ + δ + γ)(Φ + σ)
Take out common factor from both side

(1 − u1 )[β1 (αγ(1 − η)(1 − ε) − (Φ + α)(Φ + δ + γ)) − β2 σ(1 − u2 )(Φ + α)]



> [Φ + σ][αγ(1 − η)(1 − ε) − (Φ + α)(Φ + δ + γ)]

Multiply both side by −1

(1 − u1 )[β1 ((Φ + α)(Φ + δ + γ − αγ(1 − η)(1 − ε))) + β2 σ(1 − u2 )(Φ + α)]



< (Φ + σ)(Φ + α)(Φ + δ + γ) − αγ(1 − η)(1 − ε)(Φ + σ)

Divide both side for ((Φ + σ)(Φ + α)(Φ + δ + γ) − αγ(1 − η)(1 − ε)(Φ + σ))
 
β1 β2 σ(1 − u2 )(Φ + α)
⇒ (1 − u1 ) + <1
(Φ + σ) ((Φ + σ)(Φ + α)(Φ + δ + γ) − αγ(1 − η)(1 − ε)(Φ + σ))

By using simple calculation,we get


" #
β1 β2 σ(1 − u2 )
(1 − u1 ) + <1
(Φ + σ) (Φ + σ)(Φ + δ + γ) − αγ(1−η)(1−ε)(Φ+σ))
(Φ+α)
 
β1 β2 σ(1 − u2 )
⇒ (1 − u1 ) + 61
(Φ + σ) (Φ + σ)(Φ + δ + γ)
⇒ R0 6 1
This implies that a3 > 0 for R0 6 1
There for,the smoking free equilibrium of the system is locally asymptotically stable for R0 6 1.

41
4.4.7 Global Stability of Smoking Free Equilibrium
Theorem 4.4.5. For R0 < 1, then smoking free equilibrium E0 of the system (4.3.1) is globally
asymptotically stable for very small δ in the feasible domain .

Proof. Let us rewrite our model system (4.3.1) as


dM1
= F (M1 , M2 ),
dt
dM2
= G(M1 , M2 ), G(M1 , 0) = 0.
dt

Where M1 = (p, qT , qP ) ∈ R3+ represents the class of non-smokers individuals and M2 =


(l, s) ∈ R2+ represents the class of smokers individuals. The smoking free equilibrium point of the
model is denoted by E0 = (M1∗ , 0), where M1∗ = (1, 0, 0) . Since the smoking free equilibrium
point is locally asymptotically stable (see theorem (4.4.4) ), to prove global stability, we will apply
the Castillo-Chavez theorem (3.1.4). From system (4.3.1), we have
 
Φ − (1 − u1 )(β1 pl + β2 ps) − (Φ − δs)p
dM1
= F (M1 , M2 ) =  γ(1 − η)s − (Φ + α − δs)qT ,
 
dt
σu2 l + γηs + αεqT − (Φ − δs)qP
" #
dM2 (1 − u1 )(β1 pl + β2 ps) − (Φ + σ − δs)l
= G(M1 , M2 ) = .
dt σ(1 − u2 )l − (Φ + δ + γ − δs)s + α(1 − ε)qT

dM1
I. To show M1∗ is globally asymptotically stable for the system = F (M1 , 0), let us con-
dt
sider the reduced system
 
(1 − p)Φ
dM1
= F (M1 , 0) =  −(Φ + α)qT  . (4.4.12)
 
dt
αεqT − ΦqP

We can rewrite the system (4.4.12) as:


dp
= Φ − Φp, (4.4.13a)
dt
dqT
= −(Φ + α)qT (4.4.13b)
dt
dqP
= αεqT − ΦqP . (4.4.13c)
dt

42
The system (4.4.13) is non-homogeneous linear system of ordinary differential equations.
By applying Theorem (3.1.4) for the system (4.4.13), we obtain solutions

p(t) = 1 − p(0)e−Φt .
qT (t) = qT (0)e−(Φ+α)t
Z 
1 Φt

qP (t) = −Φt αεqT e dt + qp (0)
e
Taking the limit as t goes to ∞, we obtain

(p(t), qT (t), qP (t)) → (1, 0, 0) = M1∗ .


dM1
Therefore, M1∗ is globally asymptotically stable for the system = F (M1 , 0).
dt
II. We will show that G(M 1, M2 ) = AM2 − Ĝ(M1 , M2 ), Ĝ(M1 , M2 ) ≥ 0 for (M1 , M2 ) ∈ Ω
where A = ∂M ∂G
2
(M1∗ , 0) is a Metzler matrix (the off diagonal elements of A are non-negative)
and Ω is the region where the model makes biological sense. Consider a matrix
" #
∂G ∗ (1 − u1 )β1 p0 − (Φ + σ − δs0 ) (1 − u1 )β2 p0 + δl0
A= (M , 0) = .
∂M2 1 σ(1 − u2 ) 2δs0 − (Φ + δ + γ
" #
(1 − u1 )β1 p0 − (Φ + σ) (1 − u1 )β2 p0
⇒A= .
σ(1 − u2 ) −(Φ + δ + γ)
Hence, A is a Metzer matrix (off diagonal elements are non-negative). Here,

Ĝ(M1 , M2 ) = AM2 − G(M1 , M2 ).


" #
(1 − u1 )β1 p0 l − (Φ + σ)l + (1 − u1 )β2 p0 s
AM2 =
σ(1 − u2 )l − (Φ + δ + γ)s
After some simplification, we obtain
" #
(p0 − p) ((1 − u1 )(β1 l + β2 s)) − δls
Ĝ(M1 , M2 ) = ,
−δs2 − α(1 − ε)qT
" #
(p0 − p) ((1 − u1 )(β1 l + β2 s)) + δls
≥ ,
δs2 + α(1 − ε)qT
" #
(p0 − p) ((1 − u1 )(β1 l + β2 s))
≥ ≥ 0,
α(1 − ε)qT

Since 0 < p ≤ p0 and qT ≥ 0,it is true that Ĝ(M1 , M2 ) ≥ 0,the condition (ii) holds true.
There for by Castillo-Chavez theorem (3.1.4), the smoking free equilibrium point E0 of the system
(4.3.1) is globally asymptotically stable for R0 < 1.

43
4.4.8 Smoking Present Equilibrium Point(SPE)
A smoking present equilibrium point is a steady state solution where the smokers persists in the
population. The smokers will persists in the population and the persistent equilibrium point of the
modified mathematical model is given by E ∗ = (p∗ , l∗ , s∗ , qT∗ , qP∗ ). It can be obtained by setting
each equation of the system (4.3.1) equal to zero:

Φ − (1 − u1 )(β1 p∗ l∗ + β2 p∗ s∗ ) − (Φ − δs∗ )p∗ = 0 (4.4.14a)


(1 − u1 )(β1 p∗ l∗ + β2 p∗ s∗ ) − (Φ + σ − δs∗ )l∗ = 0 (4.4.14b)
∗ ∗ ∗
σ(1 − u2 )l − (Φ + δ + γ − δs )s + α(1 − ε)qT∗ =0 (4.4.14c)
∗ ∗
γ(1 − η)s − (Φ + α − δs )qT∗ =0 (4.4.14d)
σu2 l∗ + γηs∗ + αεqT∗ − (Φ − δs∗ )qP∗ = 0 (4.4.14e)

Under this endemic equilibrium point ,we assumed that s∗ is positive solution of the system.
From equation (4.4.14d), we obtain
γ(1 − η)s∗
qT∗ = . (4.4.15)
Φ + α − δs∗
From equation (4.4.14c),we obtain

∗ (Φ + δ + γ − δs∗ )s∗ α(1 − ε)qT∗


l = − (4.4.16)
σ(1 − u2 ) σ(1 − u2 )
From 4.4.15&4.4.16,we can get
(Φ + δ + γ − δs∗ )(Φ + α − δs∗ )s∗ − αγ(1 − ε)(1 − η)s∗
l∗ = . (4.4.17)
σ(1 − u2 )(Φ + α − δs∗ )
From the 4.4.14e,we solve for qP∗
σu2 l∗ γηs∗ αεqT∗
⇒ qP∗ = + + (4.4.18)
(Φ − δs∗ ) (Φ − δs∗ ) (Φ − δs∗ )
Substituting equation 4.4.15 and equation 4.4.17 in to equation 4.4.18 we can get
σu2 [(Φ + δ + γ − δs∗ )(Φ + α − δs∗ )s∗ − αγ(1 − ε)(1 − η)s∗ ]
qP∗=
σ(1 − u2 )(Φ − δs∗ )(Φ + α − δs∗ )
σ(1 − u2 )αγε(1 − η)s∗ γηs∗ (Φ + α − δs∗ )σ(1 − u2 )
+ + (4.4.19)
σ(1 − u2 )(Φ − δs∗ )(Φ + α − δs∗ ) σ(1 − u2 )(Φ − δs∗ )(Φ + α − δs∗ )
From the equation 4.4.14a,we solve for the p∗
Φ
⇒ p∗ = (4.4.20)
(1 − u1 )(β1 l∗ + β2 s∗ ) + Φ − δs∗

44
Substituting equation (4.4.16) in place of l∗ in to (4.4.20), we get

Φσ(1 − u2 )(Φ + α − δs∗ )


p∗ = (4.4.21)
[1 − u1 ][β1 ((Φ + δ + γ − δs∗ )(Φ + α − δs∗ )s∗ − αγ(1 − ε)(1 − η)s∗ )] + A

Where,A = σ(1 − u2 )(Φ + α − δs∗ ) ((1 − u1 )β2 + (Φ − δs∗ ))


Then the components p∗ , l∗ , qT∗ & qP∗ are given by p∗ (s∗ ), l∗ (s∗ ), qT∗ (s∗ )&qP∗ (s∗ ) ,where s∗ is posi-
tive solution of scaled model of system (4.3.1) ,such that

Φσ(1 − u2 )(Φ + α − δs∗ )





 p = ∗ )(Φ + α − δs∗ )s∗ − αγ(1 − ε)(1 − η)s∗ )] + A
,



 [1 − u 1 ][β 1 ((Φ + δ + γ − δs
(Φ + δ + γ − δs∗ )(Φ + α − δs∗ )s∗ − αγ(1 − ε)(1 − η)s∗



l = ,


∗)




 σ(1 − u2 )(Φ + α − δs
γ(1 − η)s∗

qT∗ = ∗
,

 Φ + α − δs
σu2 [(Φ + δ + γ − δs∗ )(Φ + α − δs∗ )s∗ − αγ(1 − ε)(1 − η)s∗ ]





 q P =
σ(1 − u2 )(Φ − δs∗ )(Φ + α − δs∗ )




σ(1 − u2 )αγε(1 − η)s∗ γηs∗ (Φ + α − δs∗ )σ(1 − u2 )



 +
 +
σ(1 − u2 )(Φ − δs∗ )(Φ + α − δs∗ ) σ(1 − u2 )(Φ − δs∗ )(Φ + α − δs∗ )

4.4.9 Local Stability of Smoking Present Equilibrium Point


In this section we discuss the local stability of the smoking present equilibrium point by stating in
the form of a theorem.

Theorem 4.4.6. If R0 > 1 ,then the smoking present equilibrium point E ∗ is locally asymptotically
stable in Ω.

The smoking present equilibrium E ∗ (p∗ , l∗ , s∗ , qT∗ , qP∗ ) is expressed interms of R0 .It is noted from
these equations that the system has no positive smoking present equilibrium point if R0 < 1.This
is because p∗ , l∗ , s∗ , qT∗ , qP∗ will assume negative values which is not biological realistic.Thus a
positive smoking present equilibrium point is achieved only when R0 > 1.
That is for existence of smoking present equilibrium, E ∗ = (p∗ , l∗ , s∗ , qT∗ , qP∗ ) its coordinates should
satisfy p∗ > 0 ,l∗ > 0,s∗ > 0,qT∗ > 0&qP∗ > 0.

Proof. The stability of the smoking present equilibrium point is then determined based on the signs
of the eigenvalues of the Jacobian matrix which is computed at the smoking present equilibrium,E ∗

45
Now consider system (4.3.1).The Jacobian matrix of the system at E ∗ is given by :
 
−(1 − u1 )(β1 l∗ + β2 s∗ ) − w −(1 − u1 )β1 p∗ −(1 − u1 )β2 p∗ + x 0 0

 ∗
(1 − u1 )(β1 l + β2 s )∗ ∗
(1 − u1 )β1 p − y ∗
(1 − u1 )β2 p + z 0 0



 
J(E ) =  0 σ(1 − u2 ) −(Φ + δ + γ − 2δs ) ∗ α(1 − ε) 0 
 

0 0 γ(1 − η) + r ∗
−(Φ + α − δs ) 0

 
0 σu2 γη + u αε −w

Where

w = (Φ − δs∗ ),
x = δp∗
y = (Φ + σ − δs∗ ),
z = δl∗ ,
r = δqT∗ ,
u = δqP∗
 
−A −B −C 0 0
D −E F 0 0
 
 

 
=⇒ J(E ) = 
 0 σ(1 − u2 ) −G α(1 − ε) 0 .

0 0 H −J 0
 
 
0 σu2 I αε −w
Here

A = [(1 − u1 )(β1 l∗ + β2 s∗ ) + w],


B = (1 − u1 )β1 p∗ ,
C = [(1 − u1 )(β1 l∗ + β2 s∗ )p∗ − x],
D = (1 − u1 )(β1 l∗ + β2 s∗ ),
E = y − (1 − u1 )β1 p∗ ,
F = (1 − u1 )β2 p∗ + z,
G = (Φ + δ + γ − 2δs∗ ),
H = γ(1 − η) + r,
I = γη + u,
J = (Φ + α − δs∗ )

Where,

46
p∗ , l∗ , s∗ , qT∗ &qP∗ are taken from the smoking present equilibrium point of the modified model.
The characteristic polynomial at the smoking present equilibrium point E ∗ is defined by

| J(E ∗ ) − λI |= 0,

where I is 5 by 5 identity matrix & λ is eigenvalue of jacobian matrix.

−(A + λ) −B −C 0 0
D −(E + λ) F 0 0
| J(E ∗ ) − λI |= 0 σ(1 − u2 ) −(G + λ) α(1 − ε) 0 = 0.
0 0 H −(J + λ) 0
0 σu2 I αε −(w + λ)

47
Now by using some row& column operation,we have following expression

−(A + λ) −B −C 0
D −(E + λ) F 0
−(w + λ) =0
0 σ(1 − u2 ) −(G + λ) α(1 − ε)
0 0 H −(J + λ)

−(E + λ) F 0
(w + λ)(A + λ) σ(1 − u2 ) −(G + λ) α(1 − ε)
0 H −(J + λ)

−B −C 0
+(w + λ)D σ(1 − u2 ) −(G + λ) α(1 − ε) =0
0 H −(J + λ)
After some simplification,we can get as following

(w + λ)[(A + λ)(−λ3 − λ2 J − λ2 G − λ2 E − λGJ − λEJ − λEG + λHα + λF σ(1 − u2 ) + λHαε


− EHαε − EGJ + F Gσ(1 − u2 )] + (w + λ)D[−λ2 B − λBG − λBJ − λCσ(1 − u2 ) − BHαε
− CJσ(1 − u2 ) − BGJ + BHα] = 0

Take out common factor (w + λ)

⇒(w + λ)[−λ4 − λ3 J − λ3 G − λ3 E − λ3 A
− λ2 AJ − λ2 AG − λ2 AE − λ2 BD − λ2 GJ − λ2 EJ − λ2 EG + λ2 Hα + λ2 F σ(1 − u2 )
+ λ2 Hαε − λEGJ − λEHαε − λAGJ − λAEJ − λAEG − λBDG − λCDσ(1 − u2 )
+ λAHαε + λAHα + λAF σ(1 − u2 ) + λEHα + λF G(1 − u2 )
− BDGJ − BDHαε − CDσ(1 − u2 ) − AEHαε − AEGJ + AF σ(1 − u2 ) + BDHα + AEHα] = 0

From this λ1 = −w = −(Φ − δs∗ ) which is negative real part .Rearranging and multiplying both
side by(-1),we get

λ4 + (A + E + G + J)λ3 + (AJ + AG + AE + BD + GJ + EJ + EG − Hα − F σ(1 − u2 ) − Hαε)λ2


+ (EGJ + EHαε + AGJ + AEJ + AEG + BDJ + CDσ(1 − u2 ) − BDG − AHαε − AHα
− AF σ(1 − u2 ) − EHα − F Gσ(1 − u2 ))λ
+ BDGJ + BDHαε + CDσ(1 − u2 ) + AEHαε + AEGJ − AEHα − BDHα − AF σ(1 − u2 ) = 0

48
By Simplifying this equation we have the polynomial degree 4 of the form:

p(λ) = λ4 + a1 λ3 + a2 λ2 + a3 λ + a4 , (4.4.22)
Where

a1 = A + E + G + J,
a2 = AJ + AG + AE + BD + GJ + EJ + EG − Hα − F σ(1 − u2 ) − Hαε,
a3 = EGJ + EHαε + AGJ + AEJ + AEG + BDJ + CDσ(1 − u2 ) − BDG − AHαε − AHα
− AF σ(1 − u2 ) − EHα − F Gσ(1 − u2 ),
a4 = BDGJ + BDHαε + CDσ(1 − u2 ) + AEHαε + AEGJ − AEHα − BDHα − AF σ(1 − u2 )

Since solving the characteristic polynomial equation (4.4.22) for eigenvalues is boring we will
use the Routh-Hurwitz criterion to determine whether all roots have negative real parts and estab-
lish the stability of the system without solving the characteristic equation itself.
The roots of the characteristic equation have negative real parts if and only if all the principal diago-
nal minors of the matrix are positive or for the characterstic polynomial in equation (4.4.22),where
degree 4 (i.en = 4),from Routh-Hurwitz criteria are a1 > 0,a3 > 0,a4 > 0, and a1 a2 a3 > a21 a4 +a23
given by

H1 = a1 , det(H1 ) = a1 > 0,
!
a1 1
H2 =
0 a2
det(H2 ) = a1 a2 > 0,
 
a1 1 0
H3 =  a3 a2 a1 
 

0 0 a3
det(H3 ) = a1 a2 a3 − a23 > 0 ⇒ a1 a2 > a3
 
a1 1 0 0
 a a a 1 
 3 2 1
H4 = 


 0 a4 a3 a2 
0 0 0 a4
det(H4 ) = −a4 (a21 a4 − a1 a2 a3 + a23 ) > 0 ⇔ a1 a2 a3 > a21 a4 + a23 ,

49
Now for polynomial of degree 4 (i.en = 4) of the characterstics equation (4.4.22),the Routh-
Hurwitz criteria satisfies the following condition:

a1 > 0
⇒ A + E + G + J > 0,
a3 > 0

⇒ EGJ + EHαε + AGJ + AEJ + AEG + BDJ + CDσ(1 − u2 ) − BDG − AHαε − AHα
−AF σ(1 − u2 ) − EHα − F Gσ(1 − u2 ) > 0
⇒ EGJ + EHαε + AGJ + AEJ + AEG + BDJ + CDσ(1 − u2 ) > BDG + AHαε + AHα
+AF σ(1 − u2 ) + EHα + F Gσ(1 − u2 ),

a4 > 0

⇒ BDGJ + BDHαε + CDσ(1 − u2 ) + AEHαε + AEGJ − AEHα − BDHα − AF σ(1 − u2 ) > 0


⇒ BDGJ + BDHαε + CDσ(1 − u2 ) + AEHαε + AEGJ > AEHα + BDHα + AF σ(1 − u2 )
Thus,
a1 a2 a3 − a23 − a21 a4 > 0
This the same as
a1 a2 a3 > a23 + a21 a4

[A + E + G + J][AJ + AG + AE + BD + GJ + EJ + EG
−Hα − F σ(1 − u2 ) − Hαε]
[EGJ + EHαε + AGJ + AEJ + AEG + BDJ
+CDσ(1 − u2 ) − BDG − AHαε − AHα
−AF σ(1 − u2 ) − EHα − F Gσ(1 − u2 )]
> [A + E + G + J]2 [BDGJ + BDHαε + CDσ(1 − u2 )
+AEHαε + AEGJ − AEHα − BDHα − AF σ(1 − u2 )]
+[EGJ + EHαε + AGJ + AEJ + AEG
+BDJ + CDσ(1 − u2 ) − BDG − AHαε − AHα
−AF σ(1 − u2 ) − EHα − F Gσ(1 − u2 )]2
There fore, by Routh-Hurwitz’s stability criterion the smoking present equilibrium E ∗ is locally
asymptotically stable for R0 > 1.

50
4.4.10 Sensitivity Analysis(Or Sensitivity Of the Basic Reproduction
Number)
Any mathematical epidemiology study’s objective is to cover a smoking habit dynamics in order to
control it by focusing on a few sensitive parameters. Sensitivity analysis depending on the model
parameters can be used to achieve this.
Sensitivity analysis demonstrates how the behavior of the model responds to changes in parameter
values, making it a helpful tool for both model construction and model evaluation[56]. It is used
to discover parameters that have a high impact on the threshold R0 and should be targeted by
intervention strategies. The threshold parameter R0 which determines stability is a function of
the parameters Φ, β1 , β2 , u1 , u2 , σ, γ, δ. In order to study the effect of this parameters on R0 we
performed a sensitivity analysis on R0 with respect to this parameters. First, We recall that the
basic reproduction number R0 is given by
 
β1 β2 σ(1 − u2 )
R0 = (1 − u1 ) + .
Φ + σ (Φ + σ)(Φ + δ + γ)

The magnitude of basic reproduction number is depend on the parameters that are associated
with smoke epidemiology. If the result is negative, then the relationship between the parameters
and R0 is inversely proportional. On the other hand, a positive sensitivity index means an increase
in the value of a parameter.

We use [47] to check this by computing the partial derivative of basic reproduction number with

51
respect to each parameter.

∂Ψ β1 (Φ + δ + γ)β1

 ΨβR10 = × = > 0,
∂β1 R0 (Φ + δ + γ)β1 + β2 σ(1 − u2 )





 ∂Ψ β2 β2 σ(1 − u2 )
ΨR


 β2 =
0
× = > 0,
∂β2 R0 (Φ + δ + γ)β1 + β2 σ(1 − u2 )





 ∂Ψ u1 u1
ΨR0


 u1 = × =− < 0,
∂u1 R0 1 − u1





 ∂Ψ u2 β2 σu2
 ΨR0 u2 = × =− < 0,


 ∂u2 R0 (Φ + δ + γ)β1 + β2 σ(1 − u2 )
∂Ψ γ β2 σ(1 − u2 )γ
ΨR0
γ = × =− < 0,


∂γ R0 (Φ + δ + γ) ((Φ + δ + γ)β1 + β2 σ(1 − u2 ))






 ∂Ψ δ β2 σ(1 − u2 )δ
ΨR0
δ = × =− < 0,


∂δ R0 (Φ + δ + γ) ((Φ + δ + γ)β1 + β2 σ(1 − u2 ))






 ∂Ψ σ ((Φ + δ + γ)β1 − β2 (1 − u2 )Φ)σ
ΨR0
σ = × =− < 0,


∂σ R0 (Φ + σ) ((Φ + δ + γ)β1 + β2 σ(1 − u2 ))





∂Ψ Φ [(Φ + δ + γ)2 β1 + β2 σ(1 − u2 )(2Φ + δ + γ + σ)] Φ


 ΨR0 = × = − < 0.


Φ
∂Φ R0 (Φ + δ + γ)(Φ + σ) ((Φ + δ + γ)β1 + β2 σ(1 − u2 ))

Table 4.3: The Sensitivity index value of modified model

Parameter Value
β1 0.5351
β2 0.4649
σ -0.4287
γ -0.2807
Φ -0.2775
δ -0.0130
u1 -0.2107
u2 -0.0382

52
From the table 4.3, we can see that R0 is most sensitive to β1 ,β2 &σ.But from those parameters
σ is inversely correlated with R0 that means ,an increase in σ would decrease smoking habit from
population,and vice-versal.And also,an increase in parameters β1 & β2 , have positively correlated
with R0 ,that means the smoking habit will increase with those parameters,and vice-versal.In the
study of sensitivity analysis,it is not ethically acceptable increasing human natural birth rate Φ to
control smoking habit and hence do not considered.

53
4.5 Numerical Simulations and Discussion
In this section,we work with numerical simulation to support our analytical solution.The numerical
simulations are carried out with help of the Ode45 MATLAB soft ware.Using the parameter values
from review literature’s given in table 4.4 and initial conditions of scaled population size p(0) =
0.53,l(0) = 0.29,s(0) = 0.09,qT = 0.04&qP = 0.02 in the scaled model equations 4.3.1and
simulation result given in figure 4.3-4.7.

Table 4.4: The parameter values of the modified model.

Parameter Value Source


Φ 0.025 Assumption
β1 2 Assumption
β2 0.608 Assumption
α 0.048 [1]
γ 0.041 [1]
δ 0.0019 [1]
η 0.06 [1]
ε 0.085 Assumption
σ 0.021 [1]
u1 0.174 Assumption
u2 0.076 Assumption

In figure(4.3a) with R0 = 0.4139,the basic reproduction number R0 < 1,which show that all
the all solutions curve goes to the smoking free equilibrium point.As consequence the smoking
habit is expect to eradicate from population .
In figure 4.4 with R0 < 1 and different initial conditions for the state variables of scaled popu-
lation,all the time series of state variables goes to their component of smoking free equilibrium
point.These indicate that the smoking free equilibrium point is locally and globally asymptotically
stable .This implies that the whatever the smoking habit lives in population in the long run,the
smoking habit will eradicate /cease from the population.
In figure (4.3b) with R0 = 13.13692,we see that the reproduction number R0 > 1,all the solu-
tions curves goes away from the smoking free equilibrium point.These indicate that the smoking
free equilibrium point is unstable for the value of R0 > 1,and the solutions will go to the smoking
present equilibrium point.As a result ,the smoking habit spread in population.In figure(4.5) with
R0 > 1 and different initial condition of scaled population,all time series of scaled population of
state variables goes to their components of smoking present equilibrium point.These indicate that

54
(a) (b)

Figure 4.3: (a) Time sries of state variables for R0 = 0.41390. (b) Time sries of state variables for
R0 = 13.13692.

the smoking present equililibrium point is locally asyptotically stable for the the value of R0 > 1.
In figure (4.6) shows that effects of proportion of temporarily quit smokers .we see that when
effects of temporarily quit smokers increases,the chain smokers individuals will decrease because
of population become to stop smoking cigarette and vice-versal.Figure (4.6)shows that when con-
tact rate of smokers increase ,the smoking habit will spread in population (or when smokers smoke
without properly in Susceptible population,then the smoking habit will spread continuously.)

55
Figure 4.4: Time series of state variables for R0 = 0.41390 with different initial condition.

Figure 4.5: Time series of state variables for R0 = 13.13692 with different initial condition.

56
Figure 4.6: Effects of ε on chain smokers

(a) (b)

Figure 4.7: (a) Parameters β1 having positive impact on the spread of smoking habit (b) Parameters
β2 having positive impact on the spread of smoking habit .

57
Chapter 5

Extension of the Modified Model into an


Optimal Control

.
In this chapter ,we introduced to plsqT qP scaled dynamic model 4.3.1 a control function u1 (t)
that represents government prohibition of smoking in public area while the control function u2 (t)
represents symbolizes the education campaign and the increase of cigarette cost to reduce the
spread smoking habit in society.The role of first control is to reduce the contact between non-
smokers and smokers while the second control is to prevent occasional smokers to become chain
smokers .We will use optimal control theory to minimize the number of people who smoke and to
increase the number of individuals who permanently quit smokers with the controls strategies over
a fixed treatment period of time T .Optimal control theory uses Pontryagin’s Maximum principle to
reduce cost by finding optimal strategies for control parameters (i.e u1 &u2 ) of the system.The nu-
merical simulations were performed in MATLAB using fourth-order forward Runge-Kutta method
to solve the state system and fourth order backward Runge-Kutta method to solve adjoint sys-
tem(because of the transversality condition).

5.1 Characterization of the Optimal Control


In this section,we study the optimal control of government prohibition and education campaign
and increase of cigarette cost .For this purpose we define our control set U to be

U = {(u1 (t), u2 (t)) : 0 6 ui (t) ≤ 1, 0 ≤ t ≤ T, i = 1, 2} (5.1.1)

Where u1 (t)& u2 (t) are Lebesgue measurable quantities bounded above by 1,which depends
on the amount of resource available for the implementation of the control strategies.The primary

58
objective of this section is to minimize the cost functional
Z T 
1 2 2
J[u1 (t), u2 (t)] = A1 l(t) + A2 s(t) + (B1 u1 (t) + B2 u2 (t)) dt (5.1.2)
0 2
constants A1 , A2 , B1 &B2 are positive.

The weight constants B1 &B2 are cost balancing coefficients associated with the government
prohibition control u1 and Education campaign control u2 respectively,and also balances the units
of integrand .In the objective functional,the terms A1 l(t) and A2 s(t) are the cost associated with
occasional smokers & chain smokers respectively.The expression 21 Bi u2i , i = 1, 2 represents cost
which is associated with control ui .The form is quadratic because we assume that costs are non-
linear its nature.
The functional J corresponds the total cost due to reduce smoking habit in community and its
control strategies . And also,the integrand function
L(l, s, u1 , u2 , t) = A1 l(t) + A2 s(t) + 21 (B1 u21 (t) + B2 u22 (t)), measures the current cost at time
t.Finally the parameter T denotes the duration of time in years of awareness campaign and treat-
ment progress .Here,we want to find optimal values u1 (t) and u2 (t) that minimizes the objective
functional subject to the state system.that is
Z T  
1 2 2
min(u1 (t),u2 (t))∈U A1 l(t) + A2 s(t) + (B1 u1 (t) + B2 u2 (t)) dt (5.1.3)
0 2
subject to
dp


 = Φ − (1 − u1 )(β1 pl + β2 ps) − (Φ − δs)p,


 dt
dl


= (1 − u1 )(β1 pl + β2 ps) − (Φ + σ − δs)l,





 dt
ds

g(x, u, t) = = σ(1 − u2 )l − (Φ + δ + γ − δs)s + α(1 − ε)qT , (5.1.4)

 dt
dq

T

= γ(1 − η)s − (Φ + α − δs)qT ,





 dt
dq

 P

= σu2 l + γηs + αεqT − (Φ − δs)qP

dt
with non-negative initial conditions

p(0) = p0 > 0, l(0) = l0 ≥ 0, s(0) = s0 ≥ 0, qT (0) = qT 0 ≥ 0, qP (0) = qP 0 ≥ 0. (5.1.5)

Our objective is to find the optimal values for the control parameters u∗1 (t), u∗2 (t) such that:

J[u∗1 (t), u∗2 (t)] = min(u1 ,u2 )∈U J[u1 (t), u2 (t)] (5.1.6)

59
5.1.1 Existence of Optimal Control Solution
The existence of the optimal control solution can be shown by conditions are followed by Flem-
ing and Rishel’s theorem[57].We have already justified the boundedness of the solution of the
dynamics of smoking model (4.3.1) and all the state variables involved in the scaled model are
continuously differentiatable .This results can be used to prove the existence of optimal control
solution.

Theorem 5.1.1. Given objective functional J(u1 (t), u2 (t)) (5.1.3) with admissible control set
U ,subject to the system (5.1.4) wih initial condition (5.1.5),then there exist an optimal control
solution u∗ = (u∗1 , u∗2 ) in U such that

J[u∗1 , u∗2 ] = min(u1 ,u2 )∈U J[u1 (t), u2 (t)] (5.1.7)

Proof. To prove the existence of optimal control solution,we need to verify the following condi-
tions.

(a) The set of solutions to system (5.1.4) with initial condition (5.1.5) and control parameters in
(5.1.1) are non-empty

(b) The system U is convex and closed

(c) The right hand side of system (5.1.4) is bounded above by sum of bounded control and state
and can be written as a linear function of the control variables with coefficients depend on
time and state variables.

(d) The integrand function L(l, s, u1 , u2 , t) is convex on U & L(l, s, u1 , u2 .t) > k(u),where k(u)
is continuous & ||u||− 1k(u) → ∞ when ||u|| → ∞.Here u = u1 (t), u2 (t)

In theorem (4.4.3),we have already justified the boundedness of the solution of the state system
Φ
(4.3.1).Since our solution for the model is bounded by 0 6 N1 (t) 6 (Φ+δ) for all t ≥ 0.This im-
plies that the solutions of the state system are continuous and bounded for each admissible control
functions in U .Moreover,the right hand side of the model equations (5.1.4) satisfies the Lipschitz
condition with respect to state variables.Hence the state system (5.1.4) with intial condition (5.1.5)
has unique solution corresponding to each admissible control function (u1 (t), u2 (t)) ∈ U .Thus
condition (a) is achieved.

To prove (b),let us consider the control set U = {u ∈ R2 : ||u||∞ ≤ 1}.Let ρ ∈ [0, 1] and
ν1 , ν2 ∈ U such that ||ν1 ||∞ ≤ 1 & ||ν2 ||∞ ≤ 1,then

60
||ρν1 + (1 − ρ)ν2 ||∞ ≤ ρ||ν1 ||∞ + (1 − ρ)||ν2 ||∞ ≤ 1
This implies that set U is convex and closed.

To prove (c),let u = (u1 (t), u2 (t)) ∈ U, X = (p, l, s, qT , qP ) & the right hand side of the state
system of equation (5.1.4) is written in the form of matrix is
 
Φ − (1 − u1 )(β1 pl + β2 ps) − (Φ − δs)p,
(1 − u1 )(β1 pl + β2 ps) − (Φ + σ − δs)l,
 
 
 
h(t, X, u) = 
 σ(1 − u2 )l − (Φ + δ + γ − δs)s + α(1 − ε)qT , 
 (5.1.8)
γ(1 − η)s − (Φ + α − δs)qT ,
 
 
σu2 l + γηs + αεqT − (Φ − δs)qP
Then from equation (5.1.8) we can split in to the form of h(t, X, u) = r(t, X) + k(t, X)uT , where
 
Φ − (β1 pl + β2 ps) − (Φ − δs)p,
(β1 pl + β2 ps) − (Φ + σ − δs)l,
 
 
 
r(t, X) =  σl − (Φ + δ + γ − δs)s + α(1 − ε)q T , 

γ(1 − η)s − (α + µ)qT ,
 
 
γηs + αεqT + (δs − Φ)qP

&
 
(β1 pl + β2 ps) 0
−(β1 pl + β2 ps) 0 
 

 
k(t, X) = 
 0 −σl 

0 0 
 

0 −σl
Now by using properties of a norm of matrix ,we get the following
||h(t, X, u)|| = ||r(t, X) + k(t, X)uT || ≤ ||r(t, X)|| + ||k(t, X)||||u||
This implies that (c) is proved
To prove (d)
Let consider integrand of the objective functional from equation 5.1.2 is defined by
1
L(l, s, u1 , u2 , t) = A1 l(t) + A2 s(t) + (B1 u21 (t) + B2 u22 (t))
2
Assume,n(t, u) = 21 (B1 u21 (t) + B2 u22 (t)).To verify the integrand L of the cost functional is
convex,if suffices to show that n(t, u) = 12 (B1 u21 (t) + B2 u22 (t)) is convex with respect to the
control control parameters u1 (t)&u2 (t).Thus,n(t, u) is a finite linear combination with positive
coefficients of the functions n1 (u1 ) = 21 u21 (t), n2 (u1 ) = 21 u22 (t).

61
We verify the function n : U → R+ defined by
n(u) = 21 ||u||2 is convex. To verify this,take r, s ∈ U .Then for any ρ ∈ [0, 1],

n(ρr + (1 − ρ)s) − (ρn(r) + (1 − ρ)n(s))


1
⇒ = [||ρr + (1 − ρ)s||2 − ρ||r||2 − (1 − ρ)||s||2 ]
2
1 2
6 [ρ ||r||2 + 2ρ(1 − ρ)||r|||s|| + (1 − ρ)2 ||s||2 − ρ||r||2 − (1 − ρ)||s||2 ]
2
1
= [ρ(1 − ρ)||r||2 + 2ρ(1 − ρ)||r||||s|| + ρ(ρ − 1)||s||2 ]
2
1
= ρ(ρ − 1)[||r||2 − 2||r|||s|| + ||s||2 ]
2
1
= ρ(ρ − 1)(||r|| − ||s||)2 6 0, because ρ ∈ [0, 1]
2
⇒n(ρr + (1 − ρ)s) 6 ρn(r) + (1 − ρ)n(s).

There for,the integrand of the objective function L(l, s, u1 , u2 , t) is convex .In addition to
1 1
L(l, s, u1 , u2 , t) = A1 l(t) + A2 s(t) + (B1 u21 (t) + B2 u22 (t)) > (B1 u21 (t) + B2 u22 (t))
2 2
Define continuous function k(u) = ω||u||2 ,when ω = min( B21 , B22 ) &u = (u1 , u2 ),then

1
L(l, s, u1 , u2 , t) > (B1 u21 (t) + B2 u22 (t)) > ω||u||2
2
⇒ L(l, s, u1 , u2 , t) > k(u), s0
||u||− 1k(u) = ||u||− 1ω||u||2 = ω||u||
⇒ ||u||− 1k(u) = ω||u|| → ∞, when ||u|| → ∞

The condition of (d) is proved


Hence,all conditions shows that there exists an optimal control solution u∗ = (u∗1 , u∗2 ) that min-
imizes the cost functional J(u1 (t), u2 (t)) over U .There fore,there exist optimal control solution.

5.1.2 The Hamiltonian and Optimality System


We used pontryagin’s Maximum principle (3.2.6) to drive the necessary conditions that an optimal
control must satisfy .This principle converts the objective functional (5.1.2) subject to the state
system (5.1.4) in to a problem of minimizing point-wise a Hamiltonian,with respect to u1 (t), u2 (t)
as:

62
1 1
H(x(t), λ(t), u(t)) =A1 l + A2 s + B1 u21 + B2 u22
2 2
+ λ1 [Φ − (1 − u1 )(β1 pl + β2 ps) − (Φ − δs)p]
+ λ2 [(1 − u1 )(β1 pl + β2 ps) − (Φ + σ − δs)l] (5.1.9)
+ λ3 [σ(1 − u2 )l − (Φ + δ + γ − δs)s + α(1 − ε)qT ]
+ λ4 [γ(1 − η)s − (Φ + α − δs)qT ]
+ λ5 [σu2 l + γηs + αεqT − (Φ − δs)qP ]

where λi , i = 1, 2, 3, 4, 5,represents the adjoint variables associated with the state variables of
scaled model of p, l, s, qT , &qP to determined suitably by applying Pontragin’s Maximum principle
(3.2.6)

Theorem 5.1.2. For an optimal control set u1 , u2 that minimizes J over U ,there are adjoint vari-
ables λ1 , ...λ5 such that
dλ1
= [(1 − u1 )(β1 l + β2 s) + Φ − δs]λ1 − (1 − u1 )(β1 l + β2 s)λ2
dt
dλ2
= −A1 + (1 − u1 )β1 pλ1 − [(1 − u1 )β1 p + δs − Φ − σ]λ2 − σ(1 − u2 )λ3 − σu2 λ5
dt
dλ3
= −A2 + [(1 − u1 )β2 p − δp]λ1 − [(1 − u1 )β2 p + δl]λ2 + (Φ + δ + γ − 2δs)λ3
dt
− [γ(1 − η) + δqT ]λ4 − [γη + δqP ]λ5
dλ4
= −α(1 − ε)λ3 + (Φ + α − δs)λ4 − αελ5
dt
dλ5
= (Φ − δs)λ5
dt
(5.1.10)

With transiversality conditions

λi (T ) = 0, i = 1, 2, 3, 4, 5 (5.1.11)

Moreover,we obtain the control set u∗1 , u∗2 characterized by


  
∗ (β1 pl + β2 ps)(λ2 − λ1 )
u1 (t) = min 1, max 0, ,
B1
   (5.1.12)
∗ σl(λ3 − λ5 )
u2 (t) = min 1, max 0,
B2

63
Proof. The form of the adjoint equations and transiversality conditions are standard results from
Pontryagini’s Maximum principle (3.2.6) .we differentiate Hamilitonian(H)(5.1.9) with respect to
the state variable p, l, s, qT , qP ,respectively and then the adjoint system can be written as
dλ1 ∂H
=− = [(1 − u1 )(β1 l + β2 s) + Φ − δs]λ1 − (1 − u1 )(β1 l + β2 s)λ2
dt ∂p
dλ2 ∂H
=− = −A1 + (1 − u1 )β1 pλ1 − [(1 − u1 )β1 p + δs − Φ − σ]λ2 − σ(1 − u2 )λ3 − σu2 λ5
dt ∂l
dλ3 ∂H
=− = −A2 + [(1 − u1 )β2 p − δp]λ1 − [(1 − u1 )β2 p + δl]λ2 + (Φ + δ + γ − 2δs)λ3
dt ∂s
− [γ(1 − η) + δqT ]λ4 − [γη + δqP ]λ5
dλ4 ∂H
=− = −α(1 − ε)λ3 + (Φ + α − δs)λ4 − αελ5
dt ∂qT
dλ5 ∂H
=− = (Φ − δs)λ5
dt ∂qP
(5.1.13)

with transiversality conditions


λi (T ) = 0, i = 1, 2, 3, 4, 5
Similarly by following the approach of Pontryagin et al[58] ,the characterization of optimal
controls u∗1 (t), u∗2 (t) ,that is the optimality equations are obtained based on the condition ∂u
∂H
1
∂H
& ∂u2

which gives
∂H
= B1 u1 + (β1 pl + β2 ps)λ1 − (β1 pl + β2 ps)λ2
∂u1
(5.1.14)
∂H
= B2 u2 − σlλ3 + σlλ5
∂u2
Using the optimality condition it follows that
∂H
= 0, at u1 = u∗1
∂u1
(5.1.15)
∂H
= 0, at u2 = u∗2
∂u2
From equation 5.1.15,we have
(β1 pl + β2 ps)(λ2 − λ1 )
B1 u∗1 + (β1 pl + β2 ps)λ1 − (β1 pl + β2 ps)λ2 = 0 ⇒ u∗1 = (5.1.16)
B1
&
σl(λ3 − λ5 )
σlλ5 + B2 u2 − λ3 σl = 0 ⇒ u∗2 = (5.1.17)
B2

64
From boundedness u∗ in U and minimality condition we get;

∂H
0, if > 0;


∂u




(β1 pl + β2 ps)(λ2 − λ1 ) ∂H

u∗1 = , if = 0;

 B1 ∂u


 ∂H
 1,
 if <0
∂u

∂H
0, if > 0;


∂u




σl(λ3 − λ5 ) ∂H

u∗2 = , if = 0;

 B2 ∂u


 ∂H
 1,
 if <0
∂u
Concluding the optimal control u∗ can be summarized as the following
  
∗ (β1 pl + β2 ps)(λ2 − λ1 )
u1 (t) = min 1, max 0, ,
B1
  
∗ σl(λ3 − λ5 )
u2 (t) = min 1, max 0,
B2

This completes the proof


The optimality system is formed the state system 5.1.4 and the adjoint system variable system
5.1.13 by incorporating the characterized control set and initial and transiversal condition.Then we
have the following optimality system:

65
dp


 = Φ − (1 − u∗1 )(β1 pl + β2 ps) − (Φ − δs)p,


 dt
dl


= (1 − u∗1 )(β1 pl + β2 ps) − (Φ + σ − δs)l,





 dt
ds


= σ(1 − u∗2 )l − (Φ + δ + γ − δs)s + α(1 − ε)qT ,





 dt
dqT


= γ(1 − η)s − (Φ + α − δs)qT ,





 dt
dqP


= σu∗2 l + γηs + αεqT − (Φ − δs)qP ,





 dt
dλ1


= [(1 − u∗1 )(β1 l + β2 s) + Φ − δs]λ1 − (1 − u∗1 )(β1 l + β2 s)λ2





 dt
dλ2

= −A1 + (1 − u∗1 )β1 pλ1 − [(1 − u∗1 )β1 p + δs − Φ − σ]λ2 − σ(1 − u∗2 )λ3 − σu∗2 λ5

 dt
dλ3


== −A2 + [(1 − u∗1 )β2 p − δp]λ1 − [(1 − u∗1 )β2 p + δl]λ2 + (Φ + δ + γ − 2δs)λ3





 dt

− [γ(1 − η) + δqT ]λ4 − [γη + δqP ]λ5






 dλ4
= −α(1 − ε)λ3 + (Φ + α − δs)λ4 − αελ5


dt





 dλ5
= (Φ − δs)λ5 ,


dt



   
(β1 pl + β2 ps)(λ2 − λ1 )



u1 (t) = min 1, max 0, ,






 B 1
  
σl(λ3 − λ5 )


 ∗
 u2 (t) = min 1, max 0,

B2

λi (T ) = 0, i = 1, 2, 3, 4, 5, (p(0), l(0), s(0), qT (0), qP (0)) = (p0 , l0 , s0 , qT 0 , qP 0 )

66
5.1.3 Numerical Simulations of Optimal Control Problem
Now we perform some numerical solutions for the scaled model 4.3.1 and the result of optimality
system consisting of state equation 5.1.4and the adjoint system 5.1.13 with the characteristic equa-
tion 5.1.15and by using parameter values given in table (4.4) we simulate it.And also,an iterative
scheme is used to find the optimal solution of the optimalty system.Since the state system 5.1.15
has initial conditions and the adjoint system equation 5.1.13 have final conditions ,we verify the
state system by applying a forward fourth-order Runge -Kutta method & show adjoint system ap-
plying a back ward fourth-order Runge -Kutta method.The process is repeated until the values in
the current iteration are close enough to the previous iteration values [44]
We have used A1 = 0.001,A2 = 10,B1 = 10.001,B2 = 10& entire time T = 365 days for
simulation of dynamics of smoking model with optimal control and also,for initial condition of
scaled model p(0) = 0.53,l(0) = 0.29,s(0) = 0.09,qT = 0.04&qP = 0.02.
In figure (5.3)the profile of optimal controls u1 and u2 shown.From these we observe that to
control the spread of smoking habit in population in entire 365 days,the use of government pro-
hibition in public area and education campaign to control occasional smokers to become chain
smokers ,the strategies must be hold effectively.By applying control and varying some parameters
value,the dynamics of state variables of potential smokers,occasional smokers and chain smokers
are shown in figure 5.1.
In figure 5.1 we detect that without control the time series show decline of potential smokers
that means susceptible population start smoking cigarette,but with control susceptible population
is going to stop smoking cigarette, without control those occasional smokers is slowly going to
decrease,but with control those smokers become fastly going to stop smoking and population be-
come quit smokers, and also Without control chain smokers smokes highly and then decrease for
some day after reach risks and smoking habit is constant until take prevention but with control
chain smokers going to stop smoking fastly.
From both figure (5.2) without control smokers are stop smoking slowly,but with control the
smokers quit fastly.

67
Figure 5.1: Simulation showing the optimal solution for potential,occasional,chain smokers
through use of u1 &u2 ,when β1 = 2&β2 = 0.608.

(a) (b)

Figure 5.2: (a) with and with out optimal control of temporarily quit smokers. (b) with and with
out optimal control of permanent quit smokers.

68
Figure 5.3: Simulation showing the profile for the optimal control u1 &u2 .

69
Chapter 6

Conclusion and Recommendation

6.1 Conclusion
In this thesis the scaled modified mathematical model on the dynamics of smoking cigarette were
formulated .We showed that there exists a domain where the model is epidemiological and mathe-
matically well-posed.
The basic reproduction number,R0 ,smoking free and smoking present equilibrium points are
computed.Based on the value of basic reproduction ,R0 ,the stability of smoking free and smoking
present equilibrium points are introduced.If R0 < 1,then smoking free equilibrium point E0 ,is
locally and globally asymptotically stable (i.e the smoking habit eradicated from the community
over a specific period of time )and if R0 > 1,then E0 is unstable .Also we have showed that smok-
ing present equilibrium E1∗ ,exists and locally asymptotically stable for all R0 > 1(i.e smokers
/smoking habit become rooted in population).The analysis and the numerical simulations without
control shows that the effective contact rate of smokers increase,the smoking habit spread in popu-
lation.Further more,as the rate at which occasional,chain smokers individual become permanently
quit smokers as well as the rate at which the temporarily quit smokers individuals increases,the
chain smokers decrease.Other wise,the smoking habit spreads rapidly in the society.
The basic model is then extended to include two time dependent control variables:government
prohibition in public place u1 and education campaign u2 .The study established and proved
the existence and characterization of an optimal control by using Pontryagin Maximum princi-
ple.Numerical simulations of optimal was carried out using MATLAB software.The out come
shows that the two strategies u1 and u2 are effective to reduce the spread of smoking habit do not
long live in community.

70
6.2 Recommendation
Based on the results of this thesis, we recommend that:

1. Efforts should be made to minimize the the interaction of non-smokers and smokers ;since
the spread of smoking habit mostly depends on the contact rate β1 &β2 .

2. Workshops,seminars,ceremony and training should be done carried on to create awareness


of the community on the transmission of the smoking habit and control.

3. We all should do our part to bring out those who entered in to smoking habit with care.

4. We kindly advice that an interested body can do further by considering real data on age and
sex structure of smokers since due to there is difference of smoking habit on population.

71
Bibliography

[1] Jha,P.(2019).Smoking cessation and e-cigarettes in China and India.bmj,367.

[2] W ikipedia(2019), P assiveSmoking, Available : https : en.wikipedia.org/wiki/ Passive-


smoking

[3] World Health Organization.(2011).WHO report on the global tobacco epidemic,2011:warning


about the dangers of tobacco.Geneva:World Health Organization.

[4] Smith,E.A.,& Malone,R.E.(2009).“Everywhere the soldier will be”:wartime tobacco

[5] Defar,A.,Getachew,T.,Teklie,H.,Bekele,A.,Gonfa,G.,Gelibo,T.,...& Teferra,S.(2017).Tobacco


use and its predictors among Ethiopian adults:A further analysis of Ethiopian NCD STEPS
survey-2015.Ethiopian Journal of Health Development,31(1),331-339.

[6] WHO FCTC secretary.frame work convention on tobacco control,Nov 2017;Available from
http://www.int/fctc/en/

[7] Lavers,T.(2012).‘Land grab’as development strategy? The political economy of agricultural


investment in Ethiopia.Journal of Peasant Studies,39(1),105-132.

[8] Available from http://www.fmhaca.gov.et/ standards directives guidelines.html

[9] Central Statistical Agency and ICF International. Ethiopia Demographic and Health Survey
2016; pp. 1–452, 2012.

[10] WHO report on the global tobacco Epidemic, 2017: Monitoring tobacco use and prevention
policies. Accessible at:http://apps.who.int/iris/bitstream/10665/255874/1/9789241512824-
eng.pdf?ua = 1&ua = 1.

[11] WHO report on the global tobacco Epidemic,2011: Warn-


ing about the dangers of tobacco.Accessible at: http :
//apps.who.int/iris/bitstream/10665/44616/1/9789240687813e ng.pdf .

72
[12] . Reinskje,T.,Thomas,S.,Ewa,F.,Jan, V.B.,Piet,W.,& Antoon, O.(2011).Hazardous com-
pounds in tobacco smoke.Int.J.Environ.Res.Public Health,8(2),613-628.

[13] Tripathi,J.P.,& Abbas,S.(2016).Global dynamics of autonomous and nonautonomous SI epi-


demic models with nonlinear incidence rate and feedback controls. Nonlinear Dynamics,
86(1), 337-351.

[14] Chitnis,N.,Schpira,A.,Smith,D.,Hay,S.I.,Smith,T.,& Steketee,R.(2010).Mathematical mod-


elling to support malaria control and elimination.

[15] Basu, A.,Mohanty,S.,& Sharma,R.(2017).Tuning of FOPID controller for meliorating the


performance of the heating furnace using conventional tuning and optimization technique.
International journal of electronics engineering research,9(1),69-85.

[16] World,H: Organization report on the global tobacco epidemic.http :


//whqlibdoc.who.int/publications/2009/9789241563918e ngf ull.pdf (2009)

[17] .Lahrouz,A.,Omari, L.,Kiouach,D.,& Belmaâti,A.(2011).Deterministic and stochastic stabil-


ity of a mathematical model of smoking. Statistics & Probability Letters,81(8),1276-1284.

[18] Centers for Disease Control and Prevention. Smoking and Tobacco Use: Youth and
Tobacco Use:Current Estimates.CDC Fact Sheet. September 24,2009. Accessed at http :
//www.cdc.gov/tobacco/datas tatistics/f acts heets/youthd ata/tobaccou se/index.htmonJanuary15,

[19] Castillo-Garsow,C.,Jordan-Salivia,G.,& Rodriguez-Herrera,A.(1997).Mathematical models


for the dynamics of tobacco use, recovery and relapse.

[20] Jung,J.H.,Park, A.,& Jung,I.H.(2018).Qualitative and sensitivity analysis of the effect of elec-
tronic cigarettes on smoking cessation. Computational and mathematical methods in medicine,
2018.

[21] Straughan,B.(2019).E-cigarette smoking with peer pressure. Mathematical Methods in the


Applied Sciences, 42(6), 2098-2108.

[22] Matintu,S.A.(2017).Smoking as Epedemic: Modeling and Simulation study. American J.


Appl. Math, 5, 31-38.

[23] Awan, A.U.Sharif, A.,Hussain,T.,& Ozair, M. (2017). Smoking model with cravings to
smoke. Advanced Studies in Biology, 9(1), 31-41.

[24] Matintu,S.A.(2017).Smoking as Epedemic: Modeling and Simulation study. American J.


Appl. Math, 5, 31-38.

73
[25] Jung,J.H.,Park, A.,& Jung,I.H.(2018). Qualitative and sensitivity analysis of the effect of
electronic cigarettes on smoking cessation. Computational and mathematical methods in
medicine, 2018.

[26] Straughan,B.(2019).E-cigarette smoking with peer pressure. Mathematical Methods in the


Applied Sciences, 42(6), 2098-2108.

[27] Pulecio-Montoya,A.M.,López-Montenegro,L.E.,& Benavides,L.M.(2019).Analysis of a


mathematical model of smoking.Contemp.Eng.Sci.,12(3),117-129.

[28] Lahrouz,A.,Omari,L.,Kiouach,D.& Belmaâti, A. (2011). Deterministic and stochastic stabil-


ity of a mathematical model of smoking. Statistics & Probability Letters, 81(8), 1276-1284.

[29] Zaman,G.(2011). Qualitative behavior of giving up smoking models. Bulletin of the


Malaysian Mathematical Sciences Society.Second Series,34(2),403-415.

[30] G.Zaman, G.(2011).Optimal campaign in the smoking dynamics. Computational and Mathe-
matical Methods in Medicine, 2011.

[31] Alkhudhari,Z.,Al-Sheikh,S.,& Al-Tuwairqi,S. (2014).The effect of occasional smokers on


the dynamics of a smoking model. In International Mathematical Forum (Vol. 9, No. 25, pp.
1207-1222).

[32] Adhana,M. A.,& Mekonnen,T.T.(2019).A mathematical model analysis of smoking tobacco


in the case of Haremaya Town; Ethiopia. Int.J. Res.Stud.Sci.Eng.Technol, 6(2), 14-24.

[33] Khyar,O., Danane,J.,& Allali,K.(2021). Mathematical Analysis and Optimal Control of Giv-
ing up the Smoking Model. International Journal of Differential Equations, 2021.

[34] Zachmanoglou, E.C. nd Thoe,D.W.(1986). Introduction to partial differential equations with


applications. Courier Corporation

[35] Miranker, W. (1962). Existence, uniqueness and stability of solutions of systems of nonlinear
difference-differential equations. Journal of Mathematics and Mechanics, 11(1):101–107

[36] Valcher,M.E.(2002).Positive systems in the behavioral approach: main issues and recent re-
sults. Electronic proceedings of MTNS,2002.

[37] Raffoul,Y.N.(2003).Boundedness in nonlinear differential equations. Nonlinear Studies,


10(4):343–350

74
[38] Korn,G .A.and Korn, T.M.(2000).Mathematical handbook for scientists and engineers: defi-
nitions, theorems, and formulas for reference and review. Courier Corporation

[39] Perko, L.(2013).Differential equations and dynamical systems, volume 7.Springer Science &
Business Media

[40] Melesse, D.Y.(2010).Mathematical analysis of an seirs model with multiple latent and

[41] Tilahun, G. T. (2018). Mathematical Model for Co-Infection of Pneumonia and Typhoid
Fever Disease with Optimal Control. PhD thesis, JKUAT.

[42] Wiggins,S.,and Golubitsky,M.(1990).Introduction to applied nonlinear dynamical systems


and chaos,volume 2. Springer.

[43] Rodrigues,H.S.(2014).Optimal control and numerical optimization applied to epidemiologi-


cal models.arXiv preprint arXiv:1401.7390.

[44] Lenhart, S.and Workman,J.T.(2007).optimal control applied to biological models. CRC press.

[45] Nelson, K.and Williams, C.(2013).Infectious disease epidemiology: theory and practice.
third. nelson ke, williamns cm, editors.

[46] Castillo-Chavez,C.,Blower,S.,Van den Driessche, P., Kirschner,D.,and Yakubu,A.-A. (2002).


Mathematical approaches for emerging and reemerging infectious diseases: models, methods,
and theory, volume 126. Springer Science & Business Media.

[47] Silva,C.J.,& Torres,D.F.(2017).A SICA compartmental model in epidemiology with applica-


tion to HIV/AIDS in Cape Verde. Ecological complexity,30,70-75.

[48] Castillo-Chavez,C.,Blower,S.,Van den Driessche,P.,Kirschner,D.,& Yakubu,


A.A.(Eds.).(2002).Mathematical approaches for emerging and reemerging infectious
diseases: models, methods, and theory (Vol. 126). Springer Science & Business Media.

[49] Sharomi,O.,& Gumel,A.B.(2008).Curtailing smoking dynamics: a mathematical modeling


approach. Applied Mathematics and Computation,195(2),475-499.

[50] Zaman,G.(2011). Qualitative behavior of giving up smoking models. Bulletin of the


Malaysian Mathematical Sciences Society. Second Series, 34(2), 403-415.

[51] A. Lahrouz, L.Omari,D.Kiouach,A.Belmaati, Deterministic and stochastic stability of a


mathematical model of smoking,Stat.Probab.Lett.81(2011)1276–1284

75
[52] Zeb,A.,Zaman, G., & Momani, S. (2013). Square-root dynamics of a giving up smoking
model. Applied Mathematical Modelling,37(7),5326-5334.

[53] Alkhudhari,Z.,Al-Sheikh,S.,& Al-Tuwairqi,S. (2014). The effect of occasional smokers on


the dynamics of a smoking model. In International Mathematical Forum (Vol. 9, No. 25, pp.
1207-1222).

[54] Zaman,G.(2011).Optimal campaign in the smoking dynamics. Computational and Mathemat-


ical Methods in Medicine, 2011.

[55] Diekmann,O.,Heesterbeek,J.,and Roberts,M.G.(2010). The construction of next genera-


tion matrices for compartmental epidemic models. Journal of the Royal Society Interface,
7(47):873– 885.

[56] Martcheva, M. (2015).An introduction to mathematical epidemiology, volume 61. Springer.

[57] Fleming, W.H.and Rishel,R.W.(2012). Deterministic and stochastic optimal control, volume
1. Springer Science & Business Media

[58] Pontryagin, L.S.(2018).Mathematical theory of optimal processes.Routledge

[59] Zeb,A.,Bano,A.,Alzahrani,E.,& Zaman,G.(2018).Dynamical analysis of cigarette smoking


model with a saturated incidence rate.AIP Advances,8(4),045317.

[60] Zeb,A.,& Alzahrani,A.(2021).Non-standard finite difference scheme and analysis of smoking


model with reversion class.Results in Physics,21,103785.

[61] Brownlee,J.(1909).Certain considerations on the causation and course of epidemics. Proceed-


ings of the Royal Society of Medicine,2(EpidemS tateM ed),243-258.

[62] Brownlee,J.(1912).XIV.The Mathematical Theory of Random Migration and Epidemic Dis-


tribution. Proceedings of the Royal Society of Edinburgh,31,262-289.

[63] Kermack,W.O.,& McKendrick,A.G.(1991).Contributions to the mathematical theory of


epidemics–I.1927.Bulletin of mathematical biology,53(1-2),3-55.

76
Appendices
MATLAB codes for Simulation of smoking dynamics with out optimal control model:

% M-File for five compartment


function dy = ModelS(t,y,phi,alpha,u_1,u_2,delta,beta1,beta2,gamma,...
varepsilon,sigma,etha,Y0)
dy=zeros(size(y));
p=y(1);
l=y(2);
s=y(3);
q_T=y(4);
q_P=y(5);
dy(1)=phi-((1-u_1)*(beta1*y(1).*y(2)+beta2*y(1).*y(3))),...
-(phi-delta*y(3))*y(1);
dy(2)=((1-u_1)*(beta1*y(1).*y(2)+beta2*y(1).*y(3))),..
-(phi+sigma-delta*y(3))*y(2);
dy(3)=sigma*(1-u_2)*y(2)-(phi+delta+gamma-delta*y(3))*y(3),...
+alpha*(1-varepsilon)*y(4);
dy(4)=gamma*(1-etha)*y(3)-(alpha+phi-delta*y(3))*y(4);
dy(5)=sigma*u_2*y(2)+gamma*etha*y(3)+alpha*varepsilon*y(4),...
-(phi-delta*y(3))*y(5);
dy=[dy(1),dy(2),dy(3),dy(4),dy(5)]’;
end

%%%% Script file for R_0>1


%Matlab code for dynamics of smoking cigarette where R_0>1
phi=0.025;alpha=0.048; u_1=0.174; u_2 =0.076;
delta=0.0019; beta1=2;beta2=0.608; gamma =0.041;varepsilon=0.085;...
sigma=0.21; etha=0.06;
N=1401;
tspan=0:0.01:365;
Y0=[0.56 0.29 0.09 0.04 0.02]’;
[t,y]=ode45(@ModelS,tspan,Y0,[],phi,alpha,u_1,u_2,delta,beta1,beta2,...
gamma,varepsilon,sigma,etha);

77
C = [’b ’;’r ’;’g ’;’k ’;’m ’];
Y0=[0.53 0.29 0.09 0.04 0.02];
[t y]=ode45(@ModelS,tspan,Y0,[],phi,alpha,u_1,u_2,delta,beta1,beta2,....
gamma,varepsilon,sigma,etha);
R_0=((1-u_1)*((phi+delta+gamma)*beta1+(beta2*sigma)*(1-u_2)))....
/((phi+delta+gamma)*(phi+sigma))
fprintf(’Value of parameter R0 is %.5f’,((1-u_1)*((phi+delta+gamma)*beta1
/((phi+delta+gamma)*(phi+sigma)))
for i =1:5
plot(t,y(:,i),C(i,:),’LineWidth’,2)
title(’All population with R_0>1 of system (4.3)’)
xlabel(’time(days)’)
ylabel(’all populaton size’)
legend(’potential smokers’,’occasional smokers ’,’chain smokers’,....
’temp quit smokers’,’perm quit smokers’);
hold on
end

%Matlab code for dynamics of smoking cigarette where R_0<1


phi=0.025;alpha=0.048; u_1=0.0174; u_2 =0.076;
delta=0.0019; beta1=0.102;beta2=0.0108; gamma =0.0041,...
;varepsilon=0.085; sigma=0.921; etha=0.06;
N=1401;
tspan=0:0.01:365;
%Y0= [0.56 0.29 0.09 0.04 0.02]’;
[t,y]=ode45(@ModelS,tspan,Y0,[],phi,alpha,u_1,u_2,delta,...
beta1,beta2,gamma,varepsilon,sigma,etha);
C = [’b ’;’r ’;’g ’;’k ’;’m ’];
Y0=[0.53 0.29 0.09 0.04 0.02]’;
[t y]=ode45(@ModelS,tspan,Y0,[],phi,alpha,u_1,u_2,delta,...
beta1,beta2,gamma,varepsilon,sigma,etha);
R_0=((1-u_1)*((phi+delta+gamma)*beta1+(beta2*sigma)*(1-u_2)))...
/((phi+delta+gamma)*(phi+sigma))
fprintf(’Value of parameter R0 is %.5f’,((1-u_1)*((phi+delta+gamma)*beta1
/((phi+delta+gamma)*(phi+sigma)))

78
for i =1:5
plot(t,y(:,i),C(i,:),’LineWidth’,2)
title(’All population with R_0<1 of system (4.3)’)
xlabel(’time(days)’)
ylabel(’all populaton size’)
legend(’potential smokers’,’occasional smokers ’,’chain smokers’,...
’temp quit smokers’,’perm quit smokers’);
hold on
end

function dS = Effeps(t,S)
%%%% M-File for smoking class with varying rate at which smoker can be
global varepsilon1 varepsilon2 varepsilon3 varepsilon4 varepsilon5 u_2,..
delta sigma phi gamma alpha
dS= zeros(size(S));
l=.50; q_t=.50;
dS(1)=(sigma*(1-u_2))*l-((phi+delta+gamma-delta*S(1))*S(1))...
+alpha*(1-varepsilon1)*q_t;
dS(2)=(sigma*(1-u_2))*l-((phi+delta+gamma-delta*S(2))*S(2))...
+alpha*(1-varepsilon2)*q_t;
dS(3)=(sigma*(1-u_2))*l-((phi+delta+gamma-delta*S(3))*S(3))...
+alpha*(1-varepsilon3)*q_t;
dS(4)=(sigma*(1-u_2))*l-((phi+delta+gamma-delta*S(4))*S(4))...
+alpha*(1-varepsilon4)*q_t;
dS(5)=(sigma*(1-u_2))*l-((phi+delta+gamma-delta*S(5))*S(5))...
+alpha*(1-varepsilon5)*q_t;

end

79
%%%% Parameter value of portion of varepsilon
global varepsilon1 varepsilon2 varepsilon3 varepsilon4 varepsilon5 phi..
u_2 delta gamma alpha sigma
phi=0.025;u_2=0.076;delta=0.0019;gamma=0.041;sigma=0.21;alpha=0.048;
varepsilon1=0.085;varepsilon2=0.5;varepsilon3=0.8 ;varepsilon4=0.85;...
varepsilon5=0.89;
tspan=0:0.001:365;
[t,S]=ode45(@Effeps,tspan,[.50 .50 .50 .50 .50]);
plot(t,S(:,1),’g’,t,S(:,2),’k’,t,S(:,3),’r’,t,S(:,4),’b’,t,S(:,5),...
’y’,’linewidth’,1.5)
legend(’varepsilon1=0.085’,’varepsilon2=0.5’,’varepsilon3=0.8’,...
’varepsilon4=0.85’,
’varepsilon5=0.89’)
xlabel(’Time (days)’);
ylabel(’chain Smokers’);
title(’ effect of varepsilon on chain smokers ’)
grid minor

%%%%% MATLAB code used for simulation of smoking dynamics with control
function [Y L u] = smokopt(P,A1,A2,B1,B2,T,Y0)
phi=P(1);beta1=P(2);beta2=P(3);sigma=P(6);...
alpha=P(7);varepsilon=P(8);gamma=P(9);delta=P(10);etha=P(11);Lamda=P(12);
u_1=P(4);u_2=P(5);
f=@(t,y,u) [phi-(1-u(1))*(beta1*y(1)*y(2)+beta2*y(1)*y(3))...
-(phi-delta*y(3))*y(1);(1-u(1))*(beta1*y(1)*y(2)+beta2*y(1)*y(3))...
-(phi+sigma-delta*y(3))*y(2);...
sigma*(1-u(2))*y(2)-(phi+delta+gamma-delta*y(3))*y(3)...
+alpha*(1-varepsilon)*y(4);gamma*(1-etha)*y(3)..
-(phi+alpha-delta*y(3))*y(4);...
sigma*u(2)*y(2)+gamma*etha*y(3)+alpha*varepsilon*y(4)...
-(phi-delta*y(3))*y(5)];
g=@(t,y,l,u) [(1-u(1))*((beta1*y(2)+beta2*y(3))...
+phi-delta*y(3))*l(1)-((1-u(1))*(beta1*y(2)+beta2*y(3)))*l(2);...
-A1+((1-u(1))*beta1*y(1))*l(1)-((1-u(1))*beta1*y(1)...
+delta*y(3)-phi-sigma)*l(2)-(sigma*(1-u(2)))*l(3)-sigma*u(2)*l(5);...

80
-A2+((1-u(1))*beta2*y(1)-delta*y(1))*l(1)...
-((1-u(1))*beta2*y(1)+delta*y(2))*l(2)+(phi+delta+gamma-2*delta*y(3))
-alpha*(1-varepsilon)*l(3)+(phi+alpha-delta*y(3))*l(4)...
-alpha*varepsilon*l(5);(phi-delta*y(3))*l(5)];
test=-1;
delta1=0.001;t=0:0.1:T;M=length(t)-1;h=0.1;h_2=h/2;
Y=zeros(5,M+1);Y(:,1)=Y0;L=zeros(5,M+1);u=zeros(2,M+1);k=100;m=0;
while (test<0 )
oldu=u;oldY=Y;oldL=L;
for i=1:M
k_1=f(t(i),Y(:,i),u(:,i));
k_2=f((t(i)+h_2),(Y(:,i)+h_2*k_1),(u(:,i)+u(:,i+1))/2);
k_3=f((t(i)+h_2),(Y(:,i)+h*k_2),(u(:,i)+u(:,i+1))/2);
k_4=f((t(i)+h),(Y(:,i)+k_3*h),u(:,i+1));
Y(:,i+1)=Y(:,i)+(1/6)*(k_1+2*k_2+2*k_3+k_4)*h;
end
for j=1:M
i=M+2-j;
m_1=g(t(i),Y(:,i),L(:,i),u(:,i));
m_2=g((t(i)-h_2),(Y(:,i)+Y(:,i-1))/2,(L(:,i)-h_2*m_1),(u(:,i)+u(:,i-1))/2
m_3=g(t(i)-h_2,(Y(:,i)+Y(:,i-1))/2,(L(:,i)-h_2*m_2),(u(:,i)+u(:,i-1))/2);
m_4=g((t(i)-h),(Y(:,i-1)),(L(:,i)-h*m_3),u(:,i-1));
L(:,i-1)=L(:,i)-(1/6)*(m_1+2*m_2+2*m_3+m_4)*h;
end
u1=[((beta1.*Y(1,:).*Y(2,:)+beta2.*Y(1,:).*Y(3,:)).*(L(2,:)-L(1,:)))/B1;
(sigma.*Y(2,:).*(L(3,:)-L(5,:)))/B2];
for i=1:M
u1(:,i)=min([1,1]’,max([0,0]’,u1(:,i)));
end
u=0.5*u1+0.5*oldu;
temp1=delta1*norm(u(1,:),1)-norm(oldu(1,:)-u(1,:),1);
temp2=delta1*norm(u(2,:),1)-norm(oldu(2,:)-u(2,:),1);
temp3=delta1*norm(Y(1,:),1)-norm(oldY(1,:)-Y(1,:),1);
temp4=delta1*norm(Y(2,:),1)-norm(oldY(2,:)-Y(2,:),1);
temp5=delta1*norm(Y(3,:),1)-norm(oldY(3,:)-Y(3,:),1);
temp6=delta1*norm(Y(4,:),1)-norm(oldY(4,:)-Y(4,:),1);

81
temp7=delta*norm(Y(5,:),1)-norm(oldY(5,:)-Y(5,:),1);
temp8=delta1*norm(L(1,:),1)-norm(oldL(1,:)-L(1,:),1);
temp9=delta1*norm(L(2,:),1)-norm(oldL(2,:)-L(2,:),1);
temp10=delta1*norm(L(3,:),1)-norm(oldL(3,:)-L(3,:),1);
temp11=delta1*norm(L(4,:),1)-norm(oldL(4,:)-L(4,:),1);
temp12=delta1*norm(L(5,:),1)-norm(oldL(5,:)-L(5,:),1);
test=max([temp1,temp2,temp3,temp4,temp5,temp6,temp7,temp8,...
temp9,temp10,temp11,temp12]);
%m=m+1;
end

%%%% Script-File for Optimal control


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
phi=0.025;alpha=0.048; delta=0.0019; beta1=2;beta2=0.608;gamma =0.041;..
varepsilon=0.085; sigma=0.21; etha=0.06;A1=.001;A2=10;B1=10.001;B2=10.0;.
T=365;t=0:0.1:T;Lamda=0.0009;u_1=0.174; u_2 =0.076;
M=length(t)-1;
P=[phi beta1 u_1 u_2 beta2 sigma alpha varepsilon gamma delta etha Lamda]
Y0=[0.56 0.29 0.09 0.04 0.02];
[Y L U]= smokopt(P,A1,A2,B1,B2,T,Y0);
[t,y] =ode45(@Kbez,t,Y0,P);
subplot(3,2,1)
plot(t,Y(1,:),t,y(:,1),’linewidth’,1.5)
xlabel(’Time(Days)’)
ylabel(’potentialsmoker(P)’)
legend(’P with optimal control’ , ’P with out optimal control’)
hold on
subplot(3,2,2)
plot(t,Y(2,:),t,y(:,2),’linewidth’,1.5)
xlabel(’Time(Days)’)
ylabel(’Occasionalsmokers(L)’)
legend(’L with optimal control’ , ’L with out optimal control’)
hold on
subplot(3,2,3)
plot(t,Y(3,:),t,y(:,3),’linewidth’,1.5)

82
xlabel(’Time(Days)’)
ylabel(’chainsmoker(S)’)
legend(’S with optimal control’ , ’S with out optimal control’)
hold on
subplot(3,2,4)
plot(t,Y(4,:),t,y(:,4),’linewidth’,1.5)
xlabel(’Time(Days)’)
ylabel(’temporarly quit smokers(Q_{T})’)
legend(’Q_{T} with optimal control’ , ’Q_{T} with out optimal control’)
hold on
subplot(3,2,5)
plot(t,Y(5,:),t,y(:,5),’linewidth’,1.5)
xlabel(’Time(Days)’)
ylabel(’permanentquit smokers(Q_{P})’)
legend(’Q_{P} with optimal control’ , ’Q_{P} with out optimal control’)
hold on
% plot(t,U,’Linewidth’,2.5)
% xlabel(’Time(days)’)
% ylabel(’control profile’)
% legend(’government prohibition u_1’,’education campaign u_2’)

83

You might also like