Dynamic analysis of a SIV Filippov system with media coverage and protective measures

Shifan Luo; Dongshu Wang; Wenxiu Li; Shifan Luo; Dongshu Wang; Wenxiu Li

doi:10.3934/math.2022745

AIMS Mathematics

2022, Volume 7, Issue 7: 13469-13492. doi: 10.3934/math.2022745

Previous Article Next Article

Research article

Dynamic analysis of a SIV Filippov system with media coverage and protective measures

1.
School of Mathematical Sciences, Huaqiao University, Quanzhou, Fujian 362021, China
2.
College of science, Zhejiang Sci-Tech University, Hangzhou, Zhejiang 310018, China

Received: 28 February 2022 Revised: 27 April 2022 Accepted: 10 May 2022 Published: 20 May 2022
MSC : 34D05, 34D23, 92B05

This study aims to analyze a class of SIV systems considering the transmission rate influenced by media coverage and protective measures, in which the transmission rate is represented by a piecewise-smooth function. Firstly, for the SIV Filippov system, we take the dynamic behaviors of two subsystems into consideration, and obtain the basic reproduction number and the equilibria of the subsystems respectively. Secondly, based on the Filippov convex method, we calculate the sliding domain and the sliding mode equation, and further analyze the global dynamic behaviors of the system, through which we verify that there is no closed orbit in the system. Furthermore, we prove the global asymptotical stability of the disease-free equilibrium, two real equilibria, and the pseudo-equilibrium under certain conditions. The results demonstrate that the threshold value, the protective measures, and the media coverage could affect the number of infected individuals and the final scale of the disease. To prevent the spread of the disease, it is necessary to select an appropriate threshold and take applicable protective measures combined with media coverage. Lastly, we verify the validity of the results by numerical simulations.

Keywords:

Citation: Shifan Luo, Dongshu Wang, Wenxiu Li. Dynamic analysis of a SIV Filippov system with media coverage and protective measures[J]. AIMS Mathematics, 2022, 7(7): 13469-13492. doi: 10.3934/math.2022745

Related Papers:

[1]	Nina Huo, Bing Li, Yongkun Li . Global exponential stability and existence of almost periodic solutions in distribution for Clifford-valued stochastic high-order Hopfield neural networks with time-varying delays. AIMS Mathematics, 2022, 7(3): 3653-3679. doi: 10.3934/math.2022202
[2]	Jin Gao, Lihua Dai . Weighted pseudo almost periodic solutions of octonion-valued neural networks with mixed time-varying delays and leakage delays. AIMS Mathematics, 2023, 8(6): 14867-14893. doi: 10.3934/math.2023760
[3]	Jin Gao, Lihua Dai . Anti-periodic synchronization of quaternion-valued high-order Hopfield neural networks with delays. AIMS Mathematics, 2022, 7(8): 14051-14075. doi: 10.3934/math.2022775
[4]	Yanshou Dong, Junfang Zhao, Xu Miao, Ming Kang . Piecewise pseudo almost periodic solutions of interval general BAM neural networks with mixed time-varying delays and impulsive perturbations. AIMS Mathematics, 2023, 8(9): 21828-21855. doi: 10.3934/math.20231113
[5]	Călin-Adrian Popa . Synchronization of Clifford-valued neural networks with leakage, time-varying, and infinite distributed delays on time scales. AIMS Mathematics, 2024, 9(7): 18796-18823. doi: 10.3934/math.2024915
[6]	Yongkun Li, Xiaoli Huang, Xiaohui Wang . Weyl almost periodic solutions for quaternion-valued shunting inhibitory cellular neural networks with time-varying delays. AIMS Mathematics, 2022, 7(4): 4861-4886. doi: 10.3934/math.2022271
[7]	Qian Cao, Xiaojin Guo . Anti-periodic dynamics on high-order inertial Hopfield neural networks involving time-varying delays. AIMS Mathematics, 2020, 5(6): 5402-5421. doi: 10.3934/math.2020347
[8]	Yuwei Cao, Bing Li . Existence and global exponential stability of compact almost automorphic solutions for Clifford-valued high-order Hopfield neutral neural networks with $D$ operator. AIMS Mathematics, 2022, 7(4): 6182-6203. doi: 10.3934/math.2022344
[9]	Li Zhu, Er-yong Cong, Xian Zhang . Global exponential stability conditions for quaternion-valued neural networks with leakage, transmission and distribution delays. AIMS Mathematics, 2023, 8(8): 19018-19038. doi: 10.3934/math.2023970
[10]	R. Sriraman, P. Vignesh, V. C. Amritha, G. Rachakit, Prasanalakshmi Balaji . Direct quaternion method-based stability criteria for quaternion-valued Takagi-Sugeno fuzzy BAM delayed neural networks using quaternion-valued Wirtinger-based integral inequality. AIMS Mathematics, 2023, 8(5): 10486-10512. doi: 10.3934/math.2023532

Abstract

1. Introduction

Nowadays, all hospitals around the world have confronted a huge surge of COVID-19 patients that require emergency health-care and services. This matter, however, has caused an enormous pressure on the available resources in many hospitals, especially with limited resources ^[1]. Meanwhile, worldwide knowledge and expertise have rapidly developed new and innovative ideas to reduce the spread of COVID-19 and treat infected patients. Acute Respiratory Distress Syndrome (ARDS) is one of the likely symptoms of coronavirus patients that can possibly cause death ^[2]. This intractable symptom can be overcome using one of the significant medical devices called an Artificial Ventilator (AV). In general, such a device is put forward to help individuals to regain healthy breathing when normal breathing is not possible. The principle of its operation is based on introducing oxygen into the lungs and withdrawing extra carbon dioxide from the body. This would enhance the gas exchange rate and hence provide comfortable breathing of sufferers. In view of technological advancement and continual modernization in introducing some technical procedures associated with control theory for medical devices, an AV system is one of these typical devices (see, e.g., ^{[3,4,5,6,7,8]}). Hence, for complementing this work, this paper proposes several optimal proportional-integral-derivative fractional-order controllers (or simply $PI^\rho D^\mu$ -controllers), where $0 < \rho < 1$ and $0 < \mu < 1$ represent the orders of the integral and the differential parts of the controller, respectively. This is for the purpose of developing dynamic stabilization capabilities of the Volume-Controlled Artificial Ventilation (VCAV) system, see Figure 1.

Figure 1. Modern high class artificial ventilator needed by most COVID-19 patients in ICU.

DownLoad: Full-Size Img PowerPoint

It is known that the optimized $PI^\rho D^\mu$ -controller has better performance than the traditional PID controller. As is known to all, it can offer more degrees of freedom by adding two parameters to the construction of the traditional one, as was reported in numerous references. For instance, it was demonstrated in ^[9] that one can obtain better simulation results with the use of the $PI^\rho D^\mu$ -controller than with the integer-order PID controllers when we deal with some time-delay systems. In addition, the same result was presented in ^[10], when the authors designed a multivariable decoupling $PI^\rho D^\mu$ -controller for variable air volume systems.

In this work, all proposed controllers have been optimally designed by carrying out two optimization techniques, namely, Particle Swarm Optimization (PSO) ^[11] and Bacteria Foraging Optimization (BFO) ^[12], for the purpose of fulfilling the high performance of the VCAV system. In particular, the key role of these two techniques is minimizing the objective function's value in light of specific constraints related to time or frequency domain, such as the system's rise time, overshoot and settling time. These constraints have a major task in quantifying the robustness of the controlled system. It is entirely normal that, once the two algorithms generate successfully the five optimal parameters ( $\kappa_p$ , $\kappa_i$ , $\kappa_d$ , $\rho$ and $\mu$ ) of the $PI^\rho D^\mu$ -controller, the focus would be oriented towards the needs of approximating the so-called fractional-order Laplacian operator $s^\rho$ and/or $s^\mu$ (or simply $s^{\pm\gamma}$ , where $\gamma = \{\rho, \ \mu\}$ and $0 < \gamma < 1$ ). Such approximation would be in the form of a finite integer-order rational transfer function due to the arbitrariness of $s^\gamma$ in its original form ^{[13,14,15,16,17,18]}. Actually, this new analytical function of $s^\gamma$ , which permits one to design and analyze the system without a need to address some hard time-domain compositions, can be typically derived by applying multiple approximations, like Oustaloup's approximation ^[19,20], the Continued Fraction Expansion (CFE) ^[19,21] approximation and more recently the $1^{st}$ - and $2^{nd}$ -order El-Khazali approximations ^[22,23]. In this work, only these four approximations will be used in order to provide the operators $s^{\pm\gamma}$ with their corresponding transfer functions. Thus, we intend to tune eight $PI^\rho D^\mu$ -BFO/PSO controllers by generating them through implementing the two aforesaid optimization algorithms. Numerical simulations of these eight controllers provide sufficient information to adopt the best performance of the controlled AV system, such as settling time, minimal overshoot and least rise time.

The rest of the paper is arranged as follows: An overview of the fractional-order linear time-invariant system is presented in the next section, followed by presenting some basic concepts of designing $PI^\rho D^\mu$ -controllers in Section 2. Four approximations of finite-order rational transfer functions of the fractional-order integro-differential Laplacian operators are outlined in Section 3, while the final section includes the main results and all numerical outcomes of the proposed design methods.

2. Fractional-order linear time-invariant systems

Fundamentally, the elementary principles of fractional calculus are typically utilized as a tool to transmit a class of control systems called the integer-order linear-time invariant (LTI) system into its fractional-order version, which is called the fractional-order LTI system (or simply FoLTI system) ^[24,25]. In light of some manifestations presented in ^[24,26], it has been confirmed that the FoLTI systems exceed the other integer-order counterparts by virtue of their flexibility in taking into account more additional parameters. In general, a FoLTI system can be expressed by the following fractional-order differential equation ^[24,27]:

$\begin{equation} p_{n}D^{\delta_{n}}y(t)+ p_{n-1}D^{\delta_{n-1}}y(t)+...+p_{1}D^{\delta_{1}}y(t)+ p_{0}D^{\delta_{0}}y(t) = q_{m}D^{\nu_{m}}u(t)+q_{m-1}D^{\nu_{m-1}}u(t)+ ...+ q_{1}D^{\nu_{1}}u(t)+q_{0}D^{\nu_{0}}u(t), \end{equation}$

(2.1)

where $y(t)$ and $u(t)$ are two variables over the time $t$ that indicate the control output and input of the system, respectively, and where $D^{\{\delta_i, \nu_k\}}$ represents the Caputo fractional differential operator of orders $\delta_i, i = 1, 2, 3, ..., n,$ and $\nu_k, k = 1, 2, 3, ..., m$ , such that $n, m\in \mathbb{N}$ .

Modeling system (2.1) can be accomplished by replacing $s^\pm\gamma$ by finite-order rational transfer functions, where $0 < \gamma < 1$ . This would permit one to design and analyze the controlled system with finite-order controllers. The numerical method that generates such an approximate transfer function determines the bounds on the order of the controller. In a similar context, the frequency response of the LTI system can be represented by a transfer function which is the ratio of the Laplace transform of the system's output to the Laplace transform of its input for zero initial conditions ^[27], i.e.,

$\begin{equation} T(s) = \frac{Y(s)}{U(s)} = \frac{q_{m}s^{\nu_{m}}+q_{m-1}s^{\nu_{m-1}}+ ...+q_{1}s^{\nu_{1}}+q_{0}s^{\nu_{0}}}{p_{n}s^{\delta_{n}}+p_{n-1}s^{\delta_{n-1}}+ ...+p_{1}s^{\delta_{1}}+p_{0}s^{\delta_{0}}}, \end{equation}$

(2.2)

where $Y(s) = \mathscr{L}\{y(t)\}, \ U(s) = \mathscr{L}\{u(t)\}$ are the Laplace transforms of $y(t)$ and $u(t)$ , respectively.

3. Design of $PI^\rho D^\mu$ controllers

The principal structure of the $PI^\rho D^\mu$ -controller was proposed by Podlubny et al. ^[28]. It was shown that the $PI^\rho D^\mu$ -controller outperforms the traditional PID-controller. From this vantage point, many real-life technical applications have been improved by applying this controller. The structure of this controller is based upon appending two further parameters ( $\rho$ and $\mu$ ) to the primary parameters $(\kappa_p, \kappa_i, \kappa_d)$ of the traditional PID-controller. Those two additional parameters would offer additional degrees of freedom in the controller algorithms. Nevertheless, the $PI^\rho D^\mu$ -controller is definitely inferred from the following fractional-order integro-differential equation ^[24,28]:

$\begin{equation} u(t) = \kappa_pe(t)+\kappa_iI^{\rho}e(t)+\kappa_dD^{\mu}e(t), \end{equation}$

(3.1)

where $e(t)$ is the error signal, $D^{\mu}$ is the Caputo operator of order $\mu$ , and $I^{\rho}$ is the Riemann-Liouville operator of order $\rho$ . The transfer function of the $PI^\rho D^\mu$ -controller is given by

$\begin{equation} C(s) = \frac{U(s)}{E(s)} = \kappa_p+\kappa_is^{-\rho}+\kappa_d s^{\mu}, \end{equation}$

(3.2)

where $E(s) = \mathscr{L}\{e(t)\}$ .

The subsequent mission concentrates on using the $PI^\rho D^\mu$ -controller with one of the most significant industrial systems during this coronavirus time: the AV system. The key purpose of implementing such a controller is to further improve the process control of the VCAV system by enhancing its dynamic performance. This demands implementing a robust optimization algorithm to enhance the system's step response by means of optimally designing the five parameters of the $PI^\rho D^\mu$ -controller. For this purpose, the BFO and PSO algorithms will be applied to determine the optimum values of these parameters through different approximations of $s^{\pm\gamma}$ that are encountered in (3.2), where $\gamma = \{\rho, \ \mu\}$ , $0 < \gamma < 1$ . It is necessary to establish the main objective function of the optimization algorithms. Notice that there are multiple standard objective functions that could be engaged for tuning the best parameters for the $PI^\rho D^\mu$ -controller, like the Integral Absolute Error (IAE), Integral Time Square Error (ITSE), Integral Time-Absolute Error (ITAE) and Integral Square Error (ISE). For instance, minimizing any objective function value is the key goal of the chosen optimization algorithm with the aim of accomplishing the best values of the $PI^\rho D^\mu$ -controller. The motivation of using the PSO and BFO algorithms comes back to their simple concepts, easy implementations, powerful abilities to control parameters and computational efficiencies when compared with other mathematical algorithms and other heuristic optimization techniques ^[29]. Nevertheless, to get a full overview about the BFO and PSO algorithms, one may consult ^{[27,30,31,32,33,34]}. The design procedure of the $PI^\rho D^\mu$ -controller through the BFO and the PSO algorithms is described by the block diagram shown in Figure 2.

Figure 2. Block diagram of BFO/PSO running to tune the

$PI^\rho D^\mu$ -controller.

DownLoad: Full-Size Img PowerPoint

To obtain the best parameters of the $PI^\rho D^\mu$ -controller, an objective function is established based on four terms to be minimized: the rise time, steady-state error, peak overshoot and the settling time of the controlled system. It takes the following form ^[32,35]:

$\begin{equation} J = (1-e^{-\beta})(M_p+e_{ss})+e^{-\beta}(T_s-T_r), \end{equation}$

(3.3)

where $e_{ss}$ is the steady state error, $T_s$ is the settling time, $T_r$ is the rise time, $M_p$ is the peak overshoot and $\beta$ is a scaling factor. It is relevant to note that although the scaling factor $\beta$ is typically chosen by a designer, it can definitely identify the roles of the four aforementioned items in the basic objective function ^[32,35]. In this work, this factor has been selected to be $0.5$ for the same reasons as reported in ^[32].

4. Rational approximations of fractional-order Laplacian operators

This part briefly introduces four approximations formulated as finite-order rational transfer functions of the fractional-order integro-differential Laplacian operators, $s^{\pm\gamma}$ , where $0 < \gamma < 1$ . The need to employ such approximations promptly emerges after determining the best values of the $PI^\rho D^\mu$ -controller by applying some optimization algorithms. Those four approximations permit one to design the target system without the need to treat some hard time-domain compositions ^[13,14].

4.1. El-Khazali Approximations

El-Khazali proposed two practical approximations of $1^{st}$ - and $2^{nd}$ -orders, respectively. The $1^{st}$ -order approximation is represented by a rational transfer function to replace $s^{\gamma}$ , where $0 < \gamma < 1$ , and given by ^[22]:

$\begin{equation} s^{\gamma} = \frac{N(s, \gamma)}{D(s, \gamma)}\cong\frac{\frac{s\tau}{\omega_{cn}}+1}{\frac{s}{\omega_{cn}}+\tau}, \end{equation}$

(4.1)

where $\omega_{cn}$ is a corner frequency, and

$\begin{equation} \tau = \tan\frac{\gamma\pi}{2}+\sec\frac{\gamma\pi}{2}. \end{equation}$

(4.2)

It is worth mentioning that the fractional-order integral Laplacian operator $s^{-\gamma}$ could be simply found from the reciprocal of (4.1) ^[22]. MATLAB Code C1, to calculate (4.1), is given in Appendix A for completeness.

In ^[23], a $2^{nd}$ -order approximation was proposed for setting up a finite-order rational transfer function of the operators $s^{\gamma}$ , where $0 < \gamma < 1$ . The reciprocal of such approximation yields $s^{-\gamma}$ . It provides an exact phase response at its center frequency and can be represented by the following modular structure ^[23]:

$\begin{equation} \Big ( \frac{s}{\omega_g}\Big)^\gamma = \prod\limits_{i = 1}^{n}F_i(s/\omega_i) = \prod\limits_{i = 1}^{n}\frac{N_i\Big(\frac{s}{(\omega_i/\omega_g)}\Big)}{D_i\Big(\frac{s}{(\omega_i/\omega_g)}\Big)}, \end{equation}$

(4.3)

where $\omega_i, i = 1, 2, ..., n$ , is the center frequency of each biquadratic module and $\omega_g = \sqrt[n]{\prod_{i = 1}^{n}\omega_i}$ is their geometric mean. If the leading center frequency $\omega_1$ of the beginning section is determined, then to detect the constant phase element, the next center frequencies of every section could be evaluated from the following recursive equation ^[19]:

$\begin{equation} \omega_i = \omega_x^{2(i-1)}\omega_1, \ \ i = 2, 3, ..., n, \end{equation}$

(4.4)

where $\omega_x$ is the maximum real solution of the following equation:

$\begin{equation} a_0a_2\eta\lambda^4+a_1(a_2-a_0)\lambda^3+(a_1^2-a_2^2-a_0^2 )\eta\lambda^2+a_1(a_2-a_0)\lambda+a_0a_2\eta = 0, \end{equation}$

(4.5)

where $\eta = \tan\frac{\gamma\pi}{4}$ . Every biquadratic component in (4.3) can be determined by

$\begin{equation} \Big( \frac{s}{\omega_g}\Big)^\gamma = F_i\Big(\frac{s}{\omega_i}\Big) = \frac{N_i(\frac{s}{\omega_i})}{D_i(\frac{s}{\omega_i})} \cong\frac{a_0(\frac{s}{\omega_i})^2+a_1(\frac{s}{\omega_i})+a_2}{a_2(\frac{s}{\omega_i})^2+a_1(\frac{s}{\omega_i})+a_0}, \end{equation}$

(4.6)

where $i = 1, 2, 3, ...$ , and

$\begin{equation} \begin{cases} a_0 & = \gamma^\gamma+2\gamma+1 \\ a_2 & = \gamma^\gamma-2\gamma+1 \\ a_1 & = (a_2-a_0) tan\Big(\frac{(2+\gamma)\pi}{4}\Big) = -6\gamma \ tan\Big(\frac{(2+\gamma)\pi}{4}\Big). \end{cases} \end{equation}$

(4.7)

Note that (4.6) is the exclusive approximation that leads to $F_i(\frac{s}{\omega_i}) = (\frac{s}{\omega_i})$ as $\gamma\rightarrow1$ . In other words,

$\begin{equation} F_i\Big(\frac{s}{\omega_i}\Big) = \frac{a_0(\frac{s}{\omega_i})^2+a_0(\frac{s}{\omega_i})}{a_0(\frac{s}{\omega_i})+a_0} = \Big(\frac{s}{\omega_i}\Big). \end{equation}$

(4.8)

Similarly, Code C2 in Appendix A simulates the approximation given by (4.6).

4.2. Oustaloup's approximation

This approximation is the most widespread one among several approximations that are employed to generate approximate finite-order rational transfer functions for the operators $s^{\pm\gamma}$ , where $0 < \gamma < 1$ . The bandwidth could be defined to offer a proper adaptation to such operators by predetermining the frequency band. Typically, in order to obtain a finite-order approximation of $s^{\gamma}$ over the frequency range $(\omega_b, \omega_h)$ , the following rational transfer function is implemented ^[36]:

$\begin{equation} s^\gamma\cong K\prod\limits_{k = -N}^{N}\frac{s+\omega_{k}^{'}}{s+\omega_{k}}, \end{equation}$

(4.9)

where the gain, zeros and poles can be calculated from the following states:

$\begin{align} &\omega_{k} = \omega_{b} \Big (\frac{\omega_{h}}{\omega_{b}}\Big)^{\frac{k+N+0.5(1+\gamma)}{2N+1}} \end{align}$

(4.10a)

$\begin{align} &\omega_{k}^{'} = \omega_{b} \Big (\frac{\omega_{h}}{\omega_{b}}\Big)^{\frac{k+N+0.5(1-\gamma)}{2N+1}} \end{align}$

(4.10b)

$\begin{align} &K = \Big (\frac{\omega_{h}}{\omega_{b}}\Big)^\frac{-\gamma}{2}\prod\limits_{k = -N}^{N}\frac{\omega_{k}}{\omega_{k}^{'}}. \end{align}$

(4.10c)

It should be pointed out that the following equation allows one to compute the unity-gain geometric frequency $\omega_u$ ^[36]:

$\begin{equation} \omega_u = \sqrt{\omega_{b}\cdot\omega_{h}}. \end{equation}$

(4.11)

Having the form of the transfer function given in (4.9) in mind, one could observe that such a function will be always of odd order ( $n = 2N+1$ ). Anyhow, Code C3, given in Appendix A, illustrates suitable MATLAB code for calculating a finite-order rational transfer function corresponding to $s^{\pm\gamma}$ using Oustaloup's approximation, where $0 < \gamma < 1$ .

4.3. The CFE approximation

This scheme is regarded as the principal mathematical method for offering approximate rational transfer functions for $s^{\pm\gamma}$ , where $0 < \gamma < 1$ . This approximation was set as follows ^[21]:

$\begin{equation} (1+z)^\gamma = \frac{1}{1-\frac{\gamma z}{1+\frac{(1+\gamma)z}{2+\frac{(1-\gamma)z}{3-\frac{(2+\gamma)z}{2+\frac{(2-\gamma)z}{5+\frac{\ldots+(n+\gamma)z}{2+\frac{(n-\gamma)z}{2n+1+...}}}}}}}}, \end{equation}$

(4.12)

where $0 < \gamma < 1$ , and $n\in\mathbb{N}$ . For the sake of simplification, (4.12) can be rewritten in the following form ^[21]:

$\begin{equation} (1+z)^\gamma = \frac{1}{1-}\frac{\gamma z}{1+}\frac{(1+\gamma)z}{2+}\frac{(1-\gamma)z}{3+}\frac{(2+\gamma)z}{2+}\frac{(2-\gamma)z}{5+}\frac{(n+\gamma)z}{2+}\frac{(n-\gamma)z}{2n+1+}. \end{equation}$

(4.13)

In order to find a proper finite-order approximation for $s^{\gamma}$ using (4.12), one may replace $z = (s^\gamma-1)$ in (4.12) or (4.13). This replacement step permits the $n^{th}$ -order approximation of such operator to be found, about the center frequency $\omega_0 = 1\ rad/sec$ , in the following form ^[26]:

$\begin{equation} s^\gamma\cong\frac{\gamma_0s^n+\gamma_1s^{n-1}+\ldots+\gamma_{n-1}s+\gamma_n}{\gamma_ns^n+\gamma_{n-1}s^{n-1}+\ldots+\gamma_{1}s+\gamma_0}, \end{equation}$

(4.14)

where $\gamma_i\in(0, \ 1)$ , which can be found in ^[37] for $i = 0, 1, ..., 5$ . Two more observations should be made at the end of this part. The first one is associated with the operator $s^{-\gamma}$ , which could be easily achieved by inverting upside down the formula given in (4.14). The second one is related to computing the finite-order rational transfer function corresponding to $s^{\pm\gamma}$ using the CFE approximation, which can be illustrated by tracking the MATLAB Code C4 given in Appendix A.

5. Optimum design for VCAV System

It is common knowledge in the biomedical engineering field that the $1^{st}$ -order lumped parameter model is the most streamlined model for the mechanics of the breathing system. In this model, the pressure in the upper airways is typically assumed to equal zero throughout spontaneous ventilation. The diaphragm, if it is active, may add subatmospheric pressure ^[38,39]. In general, there are commonly two actuators within many breathing systems of ventilators employed in ICU or anesthesia. These actuators are responsible for exhaling and inhaling processes. Since the AV system is just one of those breathing systems, its major components consist of these two actuators. In many circumstances, the exhaling actuator is actuated by a Positive End Expiratory Pressure (PEEP) valve, whereas the inhalator actuator, on the other hand, can be actuated by a piston drive mechanism. Also, the inhalator actuator system is linked with piston driven and lung mechanisms ^[38,39]. The overall system could be modeled by a $3^{rd}$ -order differential equation by taking into account the linear friction and the electrical time constant of the motor (see ^[38,39]). Through such system, the two input variables $T_l$ and $V_a$ , which respectively indicate the load torque and the applied motor voltage, are fed back to the subsystem of the lung mechanics. In particular, in view of possible estimations for all required parameters of the VCAV system, an open-loop transfer function of this system has been recently reported in ^[38] as follows:

$\begin{equation} G(s) = \frac{Y(s)}{V_i(s)} = \frac{14.471}{s^3+76.43s^2+109.76s+0.129}, \end{equation}$

(5.1)

where $Y(s) = \mathscr{L}\{y(t)\}$ and $V_i(s) = \mathscr{L}\{v_i(t)\}$ are respectively the Laplace transforms of the output variable, $y(t)$ , which indicates the piston position, and the input variable of the system, $v_i(t)$ , which indicates the voltage source of the motor.

The key target of this work is to enhance the dynamic behavior of the VCAV system by considering the transfer function given in (5.1). However, to show the efficiency of all feedback tuning schemes of $PI^\rho D^\mu$ -controllers, the objective function given in (3.3) is minimized by running the BFO and PSO algorithms, which take into account the four approximations of $s^{\pm\gamma}$ in Section 3. The maximum value of iterations and the population size in both algorithms have been taken as 100 and 20, respectively. The optimum parameters of the $PI^\rho D^\mu$ -controllers for $\beta = 0.5$ and $\beta = 0.9$ , which are used to generate the corresponding approximation schemes, are given in Tables 1 and 2, respectively.

Table 1. Tuning results of the

$PI^\rho D^\mu$ -controllers using the PSO and the BFO algorithms for

$\beta = 0.5$ .

Algorithm	Type of approximation	$\kappa_p$	$\kappa_i$	$\kappa_d$	$\rho$	$\mu$
PSO	$1^{st}$ -order El-Khazali	31.4119	0.54000	25.4162	0.31300	0.90600
	$2^{nd}$ -order El-Khazali	41.0000	15.4404	45.0000	0.09000	0.91100
	Oustaloup	59.0000	0.76000	61.0000	0.82100	0.92503
	CFE	0.16000	45.5900	51.0000	0.17700	0.86700
BFO	$1^{st}$ -order El-Khazali	2.71690	2.50680	7.71930	0.91510	0.57810
	$2^{nd}$ -order El-Khazali	20.8839	12.9392	21.7497	0.86760	0.90610
	Oustaloup	25.4065	18.5733	32.2031	0.54090	0.15850
	CFE	2.56650	10.0268	15.2471	0.18520	0.33260

| Show Table

DownLoad: CSV

Table 2. Tuning results of the

$PI^\rho D^\mu$ -controllers using the PSO and the BFO algorithms for

$\beta = 0.9$ .

Algorithm	Type of approximation	$\kappa_p$	$\kappa_i$	$\kappa_d$	$\rho$	$\mu$
PSO	$1^{st}$ -order El-Khazali	31.6262	0.5400	25.5702	0.9030	0.9060
	$2^{nd}$ -order El-Khazali	41.000	0.3200	45.000	0.7990	0.8557
	Oustaloup	59.000	24.5393	61.0000	0.1650	0.9760
	CFE	41.5691	0.3100	51.000	0.1770	0.8670
BFO	$1^{st}$ -order El-Khazali	4.5230	4.3728	6.7855	0.8750	0.2870
	$2^{nd}$ -order El-Khazali	13.7804	2.0452	16.5894	0.2662	0.9468
	Oustaloup	17.5025	11.2505	21.6484	0.8129	0.2091
	CFE	2.2440	2.1768	8.1292	0.9151	0.5781

| Show Table

DownLoad: CSV

Remark 1. The scaling factor $\beta$ is chosen to define a trade-off between the overshoot and the steady-state error ( $M_p+e_{ss}$ ) and the time difference between the settling and the rise times ( $T_s-T_r$ ). For example, for $\beta > 0.7$ more emphasis will be given to ( $M_p+e_{ss}$ ) than ( $T_s-T_r$ ). The impact of $\beta$ is highlighted by several numerical results of the proposed system model for $\beta = 0.5$ and $\beta = 0.9$ , reported below for completeness.

Clearly, and show the performances of the closed-loop controlled system using both the PSO and BFO optimization algorithms by utilizing the four different approximations of the fractional-order operator repeated many times, coupled with assuming $\beta = 0.5$ . In addition, we also plot in and the performances of the closed-loop controlled system using both algorithms by assuming $\beta = 0.9$ .

Figure 3. Step responses of the controlled system using the PSO algorithm for

$\beta = 0.5$ .

DownLoad: Full-Size Img PowerPoint

Figure 4. Step responses of the controlled system using the BFO algorithm for

$\beta = 0.5$ .

DownLoad: Full-Size Img PowerPoint

Figure 5. Step responses of the controlled system using the PSO algorithm for

$\beta = 0.9$ .

DownLoad: Full-Size Img PowerPoint

Figure 6. Step responses of the controlled system using the BFO algorithm for

$\beta = 0.9$ .

DownLoad: Full-Size Img PowerPoint

In both optimization techniques, the size of the controllers when using El-Khazali approximation is much less than that of the other two ones, as depicted in and . Moreover, there is a significant improvement in the step responses when using the PSO algorithm over that of the BFO algorithm. More precisely, one could notice that the minimum overshoot of the closed-loop VCAV system has been achieved when the $PI^\rho D^\mu$ -controller was applying the PSO algorithm.

Table 3. The transfer functions of the

$PI^\rho D^\mu$ -controllers for

$\beta = 0.5$ .

$\textsf{Algorithm}$	$\textsf{Type of approximation}$	$\textsf{Controller transfer function, $ C(s) $}$
$\textsf{PSO}$	$\textsf{$ 1^{st} $-order El-Khazali approximation}$	$\textsf{$ C_{\{1^{st}-Kh\}}(s)=\frac{626.9 s^2 + 1135 s + 462.3}{1.67 s^2 + 23.58 s + 13.52} $}$
	$\textsf{$ 2^{nd} $-order El-Khazali approximation}$	$\textsf{$ C_{\{2^{nd}-Kh\}}(s)=\frac{344.4 s^4 + 1705 s^3 + 2852 s^2 + 1810 s + 371}{0.1917 s^4 + 8.814 s^3 + 28.9 s^2 + 25.83 s + 6.079} $}$
	$\textsf{Oustaloup approximation}$	$C_{Ous}(s)=\frac{\begin{aligned}1.92e05 s^{10} + 8.374e06 s^9 + 1.191e08 s^8 + 6.29e08 s^7 + 1.468e09 s^6 + 1.569e09 s^5\\ \ \ \ \ \ \ \ \ \ \ \ + 7.846e08 s^4 + 1.793e08 s^3 + 1.675e07 s^2+ 6.184e05 s + 6599 \end{aligned}} {\begin{aligned}43.85 s^{10} + 5837 s^9 + 1.847e05 s^8 + 2.14e06 s^7 + 9.029e06 s^6 + 1.54e07 s^5\\ \ \ \ \ \ \ \ \ \ \ \ + 9.936e06 s^4 + 2.592e06 s^3 + 2.463e05 s^2 + 8567 s+ 70.81 \end{aligned}}$
	$\textsf{The CFE approximation}$	$C_{CFE}(s)=\frac{\begin{aligned}1.637e04 s^{10} + 5.268e05 s^9 + 5.684e06 s^8 + 2.671e07 s^7 + 6.385e07 s^6 + 8.343e07 s^5\\ \ \ \ \ \ \ \ \ \ \ \ \ + 6.143e07 s^4 + 2.489e07 s^3 + 5.182e06 s^2+ 4.744e05 s + 1.467e04 \end{aligned}} {\begin{aligned}2.254 s^{10} + 543.4 s^9 + 1.462e04 s^8 + 1.312e05 s^7 + 5.014e05 s^6 + 9.102e05 s^5\\ \ \ \ \ \ \ \ \ \ \ \ \ \ + 8.165e05 s^4 + 3.595e05 s^3 + 7.334e04 s^2+ 6038 s + 142 \end{aligned}}$
$\textsf{BFO}$	$\textsf{$ 1^{st} $-order El-Khazali approximation}$	$\textsf{$ C_{\{1^{st}-Kh\}}(s)=\frac{379.1 s^2 + 303.8 s + 124.7}{14.97 s^2 + 44.52 s + 2.907} $}$
	$\textsf{$ 2^{nd} $-order El-Khazali approximation}$	$\textsf{$ C_{\{2^{nd}-Kh\}}(s)=\frac{301.3 s^4 + 1018 s^3 + 1314 s^2 + 772.4 s + 186.4}{0.3702 s^4 + 15.65 s^3 + 31.49 s^2 + 16.57 s + 0.555} $}$
	$\textsf{Oustaloup approximation}$	$C_{Ous}(s)=\frac{\begin{aligned}1132 s^{10} + 8.422e04 s^9 + 1.968e06 s^8 + 1.638e07 s^7 + 5.564e07 s^6 + 7.554e07 s^5\\ \ \ \ \ \ \ \ \ \ \ \ \ + 4.426e07 s^4 + 1.068e07 s^3 + 1.106e06 s^2+ 4.292e04 s + 550.1 \end{aligned}} {\begin{aligned}12.07 s^{10} + 1008 s^9 + 2.53e04 s^8 + 2.234e05 s^7 + 7.698e05 s^6 + 1.02e06 s^5\\ \ \ \ \ \ \ \ \ \ \ \ \ \ + 5.413e05 s^4 + 1.104e05 s^3 + 8797 s^2 + 246.4 s + 2.075 \end{aligned}}$
	$\textsf{The CFE approximation}$	$C_{CFE}(s)=\frac{\begin{aligned}184.2 s^{10} + 7443 s^9 + 1.039e05 s^8 + 5.885e05 s^7 + 1.588e06 s^6 + 2.15e06 s^5\\ \ \ \ \ \ \ \ \ \ \ \ + 1.499e06 s^4 + 5.243e05 s^3 + 8.724e04 s^2+ 5874 s + 137.9 \end{aligned}} {\begin{aligned}2.342 s^{10} + 141.1 s^9 + 2552 s^8 + 1.726e04 s^7 + 5.29e04 s^6 + 7.875e04 s^5\\ \ \ \ \ \ \ \ \ \ \ \ \ + 5.864e04 s^4 + 2.135e04 s^3 + 3575 s^2 + 232.4 s + 4.71 \end{aligned}}$

| Show Table

DownLoad: CSV

Table 4. The transfer functions of the

$PI^\rho D^\mu$ -controllers for

$\beta = 0.9$ .

$\textsf{Algorithm}$	$\textsf{Type of approximation}$	$\textsf{Controller transfer function, $ C(s) $}$
$\textsf{PSO}$	$\textsf{$ 1^{st} $-order El-Khazali approximation}$	$\textsf{$ C_{\{1^{st}-Kh\}}(s)=\frac{4944 s^2 + 6329 s + 548.8}{13.1 s^2 + 178.1 s + 13.52} $}$
	$\textsf{$ 2^{nd} $-order El-Khazali approximation}$	$\textsf{$ C_{\{2^{nd}-Kh\}}(s)=\frac{577.3 s^4 + 2012 s^3 + 2207 s^2 + 778.2 s + 40.67}{0.5624 s^4 + 15.49 s^3 + 31.31 s^2 + 16.83 s + 0.8531} $}$
	$\textsf{Oustaloup approximation}$	$C_{Ous}(s)=\frac{\begin{aligned}1.529e06 s^{10} + 7.401e08 s^9 + 9.569e10 s^8 + 1.819e12 s^7 + 1.058e13 s^6 + 1.81e13 s^5\\ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ + 1.14e13 s^4 + 1.881e12 s^3 + 8.537e10 s^2+ 5.842e08 s + 1.088e06 \end{aligned}} {\begin{aligned}3.126 s^{10} + 3.135e04 s^9 + 1.393e07 s^8 + 1.702e09 s^7 + 2.953e10 s^6 + 1.378e11 s^5\\ \ \ \ \ \ \ \ \ \ \ \ + 1.149e11 s^4 + 1.94e10 s^3 + 7.952e08 s^2 + 4.859e06 s+ 8017 \end{aligned}}$
	$\textsf{The CFE approximation}$	$C_{CFE}(s)=\frac{\begin{aligned}9.315e04 s^{10} + 2.947e06 s^9 + 3.076e07 s^8 + 1.401e08 s^7 + 3.212e08 s^6 + 3.971e08 s^5\\ \ \ \ \ \ \ \ \ \ \ \ \ + 2.709e08 s^4 + 9.988e07 s^3 + 1.848e07 s^2+ 1.468e06 s + 3.426e04 \end{aligned}} {\begin{aligned}2.254 s^{10} + 2084 s^9 + 6.151e04 s^8 + 5.866e05 s^7 + 2.343e06 s^6 + 4.418e06 s^5\\ \ \ \ \ \ \ \ \ \ \ \ \ \ + 4.103e06 s^4 + 1.87e06 s^3 + 3.946e05 s^2+ 3.376e04 s + 809.4 \end{aligned}}$
$\textsf{BFO}$	$\textsf{$ 1^{st} $-order El-Khazali approximation}$	$\textsf{$ C_{\{1^{st}-Kh\}}(s)=\frac{134.8 s^2 + 187.8 s + 66.78}{10.15 s^2 + 13.45 s + 1.226} $}$
	$\textsf{$ 2^{nd} $-order El-Khazali approximation}$	$\textsf{$ C_{\{2^{nd}-Kh\}}(s)=\frac{144.4 s^4 + 613.8 s^3 + 875.3 s^2 +475.1 s + 80.65}{0.1251 s^4 + 9.486 s^3 + 29.32 s^2 + 24.11 s + 4.499} $}$
	$\textsf{Oustaloup approximation}$	$C_{Ous}(s)=\frac{\begin{aligned}4.56e04 s^{10} + 6.061e07 s^9 + 6.876e09 s^8 + 1.95e11 s^7 + 1.091e12 s^6 + 1.394e12 s^5\\ \ \ \ \ \ \ \ \ \ \ \ \ \ + 6.238e11 s^4 + 7.936e10 s^3 + 2.854e09 s^2 + 2.559e07 s + 2.134e04 \end{aligned}} {\begin{aligned}274.6 s^{10} + 6.801e05 s^9 + 9.577e07 s^8 + 3.703e09 s^7 + 2.237e10 s^6 + 3.132e10 s^5\\ \ \ \ \ \ \ \ \ \ \ \ \ \ + 9.348e09 s^4 + 4.765e08 s^3 + 6.737e06 s^2 + 1.324e04 s + 6.861 \end{aligned}}$
	$\textsf{The CFE approximation}$	$C_{CFE}(s)=\frac{\begin{aligned}3.669e04 s^{10} + 9.368e05 s^9 + 8.288e06 s^8 + 3.18e07 s^7 + 6.075e07 s^6 + 6.215e07 s^5\\ \ \ \ \ \ \ \ \ \ \ \ + 3.635e07 s^4 + 1.26e07 s^3 + 2.545e06 s^2 + 2.556e05 s + 9712 \end{aligned}} {\begin{aligned}262.6 s^{10} + 2.016e04 s^9 + 3.028e05 s^8 + 1.752e06 s^7 + 4.556e06 s^6 + 5.762e06 s^5\\ \ \ \ \ \ \ \ \ \ \ \ \ + 3.585e06 s^4 + 1.068e06 s^3 + 1.365e05 s^2 + 6128 s + 16.91 \end{aligned}}$

| Show Table

DownLoad: CSV

and , however, list the performance results of the unit-step responses of the controlled system using the four approximations for $\beta = 0.5$ and $\beta = 0.9$ , respectively. It can be seen from that the three approximations, i.e., Oustaloup, CFE, and the $2^{nd}$ -order El-Khazali approximations, gave the same performance. However, the $2^{nd}$ -order El-Khazali approximation yields a $4^{th}$ -order $PI^\rho D^\mu$ , while the other two yield $10^{th}$ -order controllers. This is a significant reduction in the size of controller. A closer look at the rise time, settling time and the maximum overshoot justifies the use of the El-Khazali $2^{nd}$ -order approximation. Similarly, when using the BFO optimization algorithm, the $2^{nd}$ -order El-Khazali approximation was superior to the other three approximations. Therefore, both remarks satisfy the use of the biquadratic approximation described by Eqs (4.3) to (4.8).

Table 5. Performance of the controlled system step response using the PSO and the BFO algorithms for

$\beta = 0.5$ .

$\textsf{Step Response}$	$H_{\{1^{st}-Kh\}}^{PSO}$	$H_{\{2^{nd}-Kh\}}^{PSO}$	$H_{CFE}^{PSO}$	$H_{Ous}^{PSO}$	$H_{\{1^{st}-Kh\}}^{BFO}$	$H_{\{2^{nd}-Kh\}}^{BFO}$	$H_{CFE}^{BFO}$	$H_{Ous}^{BFO}$
$\textsf{Rise Time}$	$\textsf{0.2829}$	$\textsf{0.1858}$	$\textsf{0.2333}$	$\textsf{0.1516}$	$\textsf{1.3512}$	$\textsf{0.4828}$	$\textsf{0.5938}$	$\textsf{0.3225}$
$\textsf{Settling Time}$	$\textsf{0.4526}$	$\textsf{0.3260}$	$\textsf{0.3783}$	$\textsf{0.2514}$	$\textsf{14.6911}$	$\textsf{5.2044}$	$\textsf{3.3122}$	$\textsf{4.7305}$
$\textsf{Settling Min.}$	$\textsf{0.9001}$	$\textsf{0.9010}$	$\textsf{0.9012}$	$\textsf{0.9008}$	$\textsf{0.9017}$	$\textsf{0.9000}$	$\textsf{0.9079}$	$\textsf{0.7189}$
$\textsf{Settling Max.}$	$\textsf{0.9997}$	$\textsf{0.9998}$	$\textsf{0.9999}$	$\textsf{0.9994}$	$\textsf{1.1660}$	$\textsf{1.0741}$	$\textsf{1.2092}$	$\textsf{1.5552}$
$\textsf{Overshoot}$	$\textsf{5.2705e-04}$	$\textsf{0.0000}$	$\textsf{0.0000}$	$\textsf{0.0000}$	$\textsf{16.6236}$	$\textsf{7.4173}$	$\textsf{20.9616}$	$\textsf{55.5270}$
$\textsf{Peak}$	$\textsf{0.9997}$	$\textsf{0.9998}$	$\textsf{0.9999}$	$\textsf{0.9994}$	$\textsf{1.1660}$	$\textsf{1.0741}$	$\textsf{1.2092}$	$\textsf{1.5552}$
$\textsf{Peak Time}$	$\textsf{0.5622}$	$\textsf{0.5414}$	$\textsf{0.5430}$	$\textsf{0.3460}$	$\textsf{4.9523}$	$\textsf{2.2783}$	$\textsf{1.3415}$	$\textsf{0.8683}$

| Show Table

DownLoad: CSV

Table 6. Performance of the controlled system step response using the PSO and the BFO algorithms for

$\beta = 0.9$ .

$\textsf{Step Response}$	$H_{\{1^{st}-Kh\}}^{PSO}$	$H_{\{2^{nd}-Kh\}}^{PSO}$	$H_{CFE}^{PSO}$	$H_{Ous}^{PSO}$	$H_{\{1^{st}-Kh\}}^{BFO}$	$H_{\{2^{nd}-Kh\}}^{BFO}$	$H_{CFE}^{BFO}$	$H_{Ous}^{BFO}$
$\textsf{Rise Time}$	$\textsf{ 0.2817}$	$\textsf{ 0.1854}$	$\textsf{ 0.2200}$	$\textsf{ 0.1387}$	$\textsf{ 0.9100}$	$\textsf{ 0.9252}$	$\textsf{ 1.6685}$	$\textsf{ 0.4199}$
$\textsf{Settling Time}$	$\textsf{ 0.4506}$	$\textsf{ 0.3016}$	$\textsf{ 0.3499}$	$\textsf{ 0.2433}$	$\textsf{ 5.4000}$	$\textsf{ 2.5069}$	$\textsf{ 18.9348}$	$\textsf{ 3.9376}$
$\textsf{Settling Min.}$	$\textsf{ 0.9018}$	$\textsf{ 0.9005}$	$\textsf{ 0.9022}$	$\textsf{ 0.9014}$	$\textsf{ 0.9075}$	$\textsf{ 0.8997}$	$\textsf{ 0.9006}$	$\textsf{ 0.8633}$
$\textsf{Settling Max.}$	$\textsf{ 0.9998}$	$\textsf{ 0.9998}$	$\textsf{ 0.9996}$	$\textsf{ 0.9997}$	$\textsf{ 1.2930}$	$\textsf{ 0.9982}$	$\textsf{ 1.1533}$	$\textsf{ 1.4392}$
$\textsf{Overshoot}$	$\textsf{ 2.4138e-004}$	$\textsf{ 0.0000}$	$\textsf{ 0.0000}$	$\textsf{0.0000}$	$\textsf{ 29.3215}$	$\textsf{ 0.0000}$	$\textsf{ 15.3356}$	$\textsf{ 43.9248}$
$\textsf{Peak}$	$\textsf{ 0.9998}$	$\textsf{ 0.9998}$	$\textsf{ 0.9996}$	$\textsf{ 0.9997}$	$\textsf{ 1.2930}$	$\textsf{ 0.9982}$	$\textsf{ 1.1533}$	$\textsf{ 1.4392}$
$\textsf{Peak Time}$	$\textsf{ 0.5620}$	$\textsf{ 0.3872}$	$\textsf{ 0.4728}$	$\textsf{ 0.3791}$	$\textsf{ 2.3246}$	$\textsf{ 4.3336}$	$\textsf{ 6.2231}$	$\textsf{ 1.1085}$

| Show Table

DownLoad: CSV

and shows the control signals of the controlled system using the PSO and the BFO optimization algorithms for $\beta = 0.5$ , while and show the control signals for $\beta = 0.9$ . In addition, shows the values of the objective function (3.3) using the two optimization algorithms for $\beta = 0.5$ and $\beta = 0.9$ . One concludes that the PSO algorithm yields a better result than that of the BFO algorithm. Even though the initial magnitude of the control signal of the PSO algorithm is larger than that of the BFO algorithm, the value of the objective function $J$ described by (3.3) is smaller when using the PSO algorithm. This is also verified by the step responses of the controlled systems, as shown by and . It is clear that the $2^{nd}$ -order El-Khazali approximation yields a smaller controller signal than the other approximations and provides a competitive steady-state error compared to its other counterparts in both algorithms and for both $\beta = 0.5$ and $\beta = 0.9$ .

Figure 7.

$PI^\rho D^\mu$ -control signal using PSO algorithm for

$\beta = 0.5$ .

DownLoad: Full-Size Img PowerPoint

Figure 8.

$PI^\rho D^\mu$ -control signal using BFO algorithm for

$\beta = 0.5$ .

DownLoad: Full-Size Img PowerPoint

Figure 9.

$PI^\rho D^\mu$ -control signal using PSO algorithm for

$\beta = 0.9$ .

DownLoad: Full-Size Img PowerPoint

Figure 10.

$PI^\rho D^\mu$ -control signal using BFO algorithm for

$\beta = 0.9$ .

DownLoad: Full-Size Img PowerPoint

Table 7. The value of the objective function (3.3) using PSO and BFO algorithms for

$\beta = 0.5$ and

$\beta = 0.9$ .

$\textsf{Algorithm}$	$\textsf{Type of approximation}$	$\textsf{When $ \beta=0.5 $}$	$\textsf{When $ \beta=0.9 $}$
$\textsf{PSO}$	$\textsf{$ 1^{st} $-order El-Khazali approximation}$	$\textsf{0.2190}$	$\textsf{0.2199}$
	$\textsf{$ 2^{nd} $-order El-Khazali approximation}$	$\textsf{0.1247}$	$\textsf{0.1662}$
	$\textsf{Oustaloup approximation}$	$\textsf{0.1131}$	$\textsf{0.0949}$
	$\textsf{The CFE approximation}$	$\textsf{0.1336}$	$\textsf{0.1732}$
$\textsf{BFO}$	$\textsf{$ 1^{st} $-order El-Khazali approximation}$	$\textsf{0.8197}$	$\textsf{1.403}$
	$\textsf{$ 2^{nd} $-order El-Khazali approximation}$	$\textsf{0.3074}$	$\textsf{0.4393}$
	$\textsf{Oustaloup approximation}$	$\textsf{0.7388}$	$\textsf{0.8126}$
	$\textsf{The CFE approximation}$	$\textsf{0.6341}$	$\textsf{0.9395}$

| Show Table

DownLoad: CSV

6. Conclusions

A new $PI^\rho D^\mu$ -controller is developed to control and improve the behavior of the Volume-Controlled Artificial Ventilation (VCAV) system. Two optimization algorithms, Bacteria Foraging Optimization (BFO) and Particle Swarm Optimization (PSO), are successfully used in conjunction with the use of four different approximations of the fractional-order integro-differential Laplacian operators, $s^{\pm\gamma}$ , where $0 < \gamma < 1$ . These approximations are the $1^{st}$ - and the $2^{nd}$ -order El-Khazali approximations, Oustaloup's approximation and the Continued Fractional Expansion (CFE) approximation. The PSO algorithm yields a better performance compared to that of the BFO algorithm, especially when using the $2^{nd}$ -order El-Khazali approximations. It provided a significant improvement in terms of a smaller controller size, i.e., a $4^{th}$ -order one, and in terms of the overall performance of the controlled system step response.

Acknowledgments

This work is supported by Ajman University grant 2020-COVID-19-08.

Conflict of interest

The authors declare no conflict of interest.

Appendix A

Code C1: MATLAB code for calculating the $1^{st}$ -order El-Khazali approximation for $s^{\pm\gamma}$ , where $0 < \gamma < 1$ .

Code C2: MATLAB code for calculating the $2^{nd}$ -order El-Khazali approximation for $s^{\pm\gamma}$ , where $0 < \gamma < 1$ .

Code C3: MATLAB code for calculating the Oustaloup approximation for $s^{\pm\gamma}$ , where $0 < \gamma < 1$ .

Code C4: MATLAB code for calculating the CFE approximation for $s^{\pm\gamma}$ , where $0 < \gamma < 1$ .

References

[1]	G. Zaman, Y. H. Kang, I. H. Jung, Stability analysis and optimal vaccination of an SIR epidemic model, Biosystems, 93 (2008), 240â€"249. http://dx.doi.org/10.1016/j.biosystems.2008.05.004 doi: 10.1016/j.biosystems.2008.05.004
[2]	I. Cooper, A. Mondal, C. G. Antonopoulos, A SIR model assumption for the spread of COVID-19 in different communities, Chaos Soliton. Fract., 139 (2020), 110057. http://dx.doi.org/10.1016/j.chaos.2020.110057 doi: 10.1016/j.chaos.2020.110057
[3]	L. J. S Allen, Some discrete-time SI, SIR, and SIS epidemic models, Math. Biosci., 124 (1994), 83â€"105. http://dx.doi.org/10.1016/0025-5564(94)90025-6 doi: 10.1016/0025-5564(94)90025-6
[4]	Y. Enatsu, Y. Nakata, Y. Muroya, Global stability for a class of discrete SIR epidemic models, Math. Biosci. Eng., 7 (2010), 347â€"361. http://dx.doi.org/10.3934/mbe.2010.7.347 doi: 10.3934/mbe.2010.7.347
[5]	B. Shulgin, L. Stone, Z. Agur, Pulse vaccination strategy in the SIR epidemic model, Bull. Math. Biol., 60 (1998), 1123â€"1148. http://dx.doi.org/10.1006/s0092-8240(98)90005-2 doi: 10.1006/s0092-8240(98)90005-2
[6]	L. Stone, B. Shulgin, Z. Agur, Theoretical examination of the pulse vaccination policy in the SIR epidemic model, Math. Comput. Model., 31 (2000), 207â€"215. http://dx.doi.org/10.1016/S0895-7177(00)00040-6 doi: 10.1016/S0895-7177(00)00040-6
[7]	T. Zhao, Y. Xiao, Non-smooth plant disease models with economic thresholds, Math. Biosci., 241 (2013), 34â€"48. http://dx.doi.org/10.1016/j.mbs.2012.09.005 doi: 10.1016/j.mbs.2012.09.005
[8]	Z. Guo, L. Huang, X. Zou, Impact of discontinuous treatments on disease dynamics in an SIR epidemic model, Math. Biosci. Eng., 9 (2012), 97â€"110. http://dx.doi.org/10.3934/mbe.2012.9.97 doi: 10.3934/mbe.2012.9.97
[9]	Z. Guo, X. Zou, Impact of discontinuous harvesting on fishery dynamics in a stock-effort fishing model, Commun. Nonlinear Sci., 20 (2015), 594â€"603. http://dx.doi.org/10.1016/j.cnsns.2014.06.014 doi: 10.1016/j.cnsns.2014.06.014
[10]	L. Huang, H. Ma, J. Wang, C. Huang, Global dynamics of a Filippov plant disease model with an economic threshold of infected-susceptible ratio, J. Appl. Anal. Comput., 10 (2020), 2263â€"2277. http://dx.doi.org/10.11948/20190409 doi: 10.11948/20190409
[11]	W. Li, L. Huang, J. Wang, Global dynamics of Filippov-type plant disease models with an interaction ratio threshold, Math. Meth. Appl. Sci., 43 (2020), 6995â€"7008. http://dx.doi.org/10.1002/mma.6450 doi: 10.1002/mma.6450
[12]	Y. K. Xie, Z. Wang, A ratio-dependent impulsive control of an SIQS epidemic model with non-linear incidence, Appl. Math. Comput., 423 (2022), 127018. https://doi.org/10.1016/j.amc.2022.127018 doi: 10.1016/j.amc.2022.127018
[13]	Y. K. Xie, Z. Wang, Transmission dynamics, global stability and control strategies of a modified SIS epidemic model on complex networks with an infective medium, Math. Comput. Simul., 188 (2021), 23â€"34. https://doi.org/10.1016/j.matcom.2021.03.029 doi: 10.1016/j.matcom.2021.03.029
[14]	W. O. Kermack, A. G. McKendrick, A contribution to the mathematical theory of epidemics, P. Roy. Soc. A, 115 (1927), 700â€"721. http://dx.doi.org/10.1007/bf02464423 doi: 10.1007/bf02464423
[15]	W. O. Kermack, A. G. McKendrick, Contributions to the mathematical theory of epidemics â…¡.â€"-The problem of endemicity, P. Roy. Soc. A, 138 (1932), 55â€"83. http://dx.doi.org/10.1016/s0092-8240(05)80041-2 doi: 10.1016/s0092-8240(05)80041-2
[16]	W. O. Kermack, A. G. McKendrick, Contributions to the mathematical theory of epidemics â…¢.â€"-Further studies of the problem of endemicity, P. Roy. Soc. A, 141 (1933), 94â€"122. http://dx.doi.org/10.2307/96207 doi: 10.2307/96207
[17]	B Shulgin, L. W. Stone, Z. Agur, Pulse vaccination strategy in the SIR epidemic model, Bull. Math. Biol., 60 (1998), 1123â€"1148. http://dx.doi.org/10.1016/S0092-8240(98)90005-2 doi: 10.1016/S0092-8240(98)90005-2
[18]	A. B. Gumel, S. M. Moghadas, A qualitative study of a vaccination model with non-linear incidence, Appl. Math. Comput., 143 (2003), 409â€"419. http://dx.doi.org/10.1016/S0096-3003(02)00372-7 doi: 10.1016/S0096-3003(02)00372-7
[19]	S. Tang, J. Liang, Y. Xiao, R. A. Cheke, Sliding bifurcations of Filippov two stage pest control models with economic thresholds, SIAM J. Appl. Math., 72 (2012), 1061â€"1080. http://dx.doi.org/10.1137/110847020 doi: 10.1137/110847020
[20]	Y. Zhang, Y. Xiao, Global dynamics for a Filippov epidemic system with imperfect vaccination, Nonlinear Anal. Hybrid Syst., 38 (2020), 100932. http://dx.doi.org/10.1016/j.nahs.2020.100932 doi: 10.1016/j.nahs.2020.100932
[21]	A. P. Lemos-Paio, C. J. Silva, D. F. M. Torres, An epidemic model for cholera with optimal control treatment, J. Comput. Appl. Math., 318 (2017), 168â€"180. http://dx.doi.org/10.1016/j.cam.2016.11.002 doi: 10.1016/j.cam.2016.11.002
[22]	M. J. Jeger, L. V. Madden, F. V. D. Bosch, The effect of transmission route on plant virus epidemic development and disease control, J. Theor. Biol., 258 (2009), 198â€"207. http://dx.doi.org/10.1016/j.jtbi.2009.01.012 doi: 10.1016/j.jtbi.2009.01.012
[23]	V. C. C. Cheng, S. C. Wong, V. W. M. Chuang, S. Y. C. So, J. H. K. Chen, S. Sridhar, et al., The role of community-wide wearing of face mask for control of coronavirus disease 2019 (COVID-19) epidemic due to SARS-CoV-2, J. Infect., 81 (2020), 107â€"114. http://dx.doi.org/10.1016/j.jinf.2020.04.024 doi: 10.1016/j.jinf.2020.04.024
[24]	S. E. Eikenberry, M. Mancuso, E. Iboi, T. Phan, K. Eikenberry, Y. Kuang, et al., To mask or not to mask: Modeling the potential for face mask use by the general public to curtail the covid-19 pandemic, Infect. Dis. Model., 5 (2020), 293â€"308. http://dx.doi.org/10.1101/2020.04.06.20055624 doi: 10.1101/2020.04.06.20055624
[25]	C. R. MacIntyre, S. Cauchemez, D. E. Dwyer, H. Seale, P. Cheung, G. Browne, et al., Face mask use and control of respiratory virus transmission in households, Emerg. Infect. Dis., 15 (2009), 233. http://dx.doi.org/10.3201/eid1502.081167 doi: 10.3201/eid1502.081167
[26]	Z. Tai, T. Sun, Media dependencies in a changing media environment: The case of the 2003 SARS epidemic in China, New Media Soc., 9 (2007), 987â€"1009. http://dx.doi.org/10.1177/1461444807082691 doi: 10.1177/1461444807082691
[27]	S. Collinson, J. M. Heffernan, Modelling the effects of media during an influenza epidemic, BMC Public Health, 14 (2014), 1â€"10. http://dx.doi.org/10.1186/1471-2458-14-376 doi: 10.1186/1471-2458-14-376
[28]	R. M. Anderson, R. M. May, Infectious diseases of humans: Dynamics and control, Oxford: Oxford Science Publications, 1992. http://dx.doi.org/10.1126/science.254.5031.591
[29]	J. Cui, Y. Sun, H. Zhu, The impact of media on the spreading and control of infectious disease, J. Dyn. Differ. Equ., 20 (2008), 31â€"53. http://dx.doi.org/10.1007/s10884-007-9075-0 doi: 10.1007/s10884-007-9075-0
[30]	J. Deng, S. Tang, H. Shu, Joint impacts of media, vaccination and treatment on an epidemic Filippov model with application to COVID-19, J. Theor. Biol., 523 (2021), 110698. http://dx.doi.org/10.1016/j.jtbi.2021.110698 doi: 10.1016/j.jtbi.2021.110698
[31]	A. Wang, Y. Xiao, A Filippov system describing media effects on the spread of infectious diseases, Nonlinear Anal. Hybrid Syst., 11 (2014), 84â€"97. http://dx.doi.org/10.1016/j.nahs.2013.06.005 doi: 10.1016/j.nahs.2013.06.005
[32]	J. Wang, F. Zhang, L. Wang, Equilibrium, pseudoequilibrium and sliding-mode heteroclinic orbit in a Filippov-type plant disease model, Nonlinear Anal.-Real, 31 (2016), 308â€"324. http://dx.doi.org/10.1016/j.nonrwa.2016.01.017 doi: 10.1016/j.nonrwa.2016.01.017
[33]	W. Li, L. Huang, J. Wang, Dynamic analysis of discontinuous plant disease models with a non-smooth separation line, Nonlinear Dyn., 99 (2020), 1675â€"1697. http://dx.doi.org/10.1007/s11071-019-05384-w doi: 10.1007/s11071-019-05384-w
[34]	A. F. Filippov, Differential equations with discontinuous right-hand side, Dordrecht: Kluwer Academic, 1988. http://dx.doi.org/10.1016/0022-247X(91)90044-Z
[35]	X. Chen, L. Huang, A Filippov system describing the effect of prey refuge use on a ratio-dependent predator-prey model, J. Math. Anal. Appl., 428 (2015), 817â€"837. http://dx.doi.org/10.1016/j.jmaa.2015.03.045 doi: 10.1016/j.jmaa.2015.03.045
[36]	A. Wang, Y. Xiao, Sliding bifurcation and global dynamics of a Filippov epidemic model with vaccination, Int. J. Bifurcat. Chaos, 23 (2013), 1350144. http://dx.doi.org/10.1142/S0218127413501447 doi: 10.1142/S0218127413501447
[37]	M. di Bernardo, C. J. Budd, A. R. Champneys, P. Kowalczyk, A. B. Nordmark, Bifurcation in nonsmooth dynamical systems, SIAM Rev., 50 (2008), 629â€"701. http://dx.doi.org/10.1137/050625060 doi: 10.1137/050625060
[38]	Y. A. Kuznetsov, S. Rinaldi, A. Gragnani, One parameter bifurcations in planar Filippov systems, Int. J. Bifurcat. Chaos, 13 (2003), 2157â€"2188. http://dx.doi.org/10.1142/S0218127403007874 doi: 10.1142/S0218127403007874
[39]	P. V. D. Driessche, J. Watmough, Reproduction numbers and sub-threshold endemic equilibria for compartmental models of disease transmission, Math. Biosci., 180 (2002), 29â€"48. http://dx.doi.org/10.1016/S0025-5564(02)00108-6 doi: 10.1016/S0025-5564(02)00108-6
[40]	Z. Ma, Y. Zhou, C. Li, Qualitative and stability methods for ordinary differential equations, Beijing: Science Press, 2015.
[41]	R. M. Corless, G. H. Gonnet, D. E. G. Hare, D. J. Jeffrey, D. E. Kunth, On the Lambert W function, Adv. Comput. Math., 5 (1996), 329â€"359. http://dx.doi.org/10.1007/BF02124750 doi: 10.1007/BF02124750

This article has been cited by:

1.	Nina Huo, Bing Li, Yongkun Li, Global exponential stability and existence of almost periodic solutions in distribution for Clifford-valued stochastic high-order Hopfield neural networks with time-varying delays, 2022, 7, 2473-6988, 3653, 10.3934/math.2022202
2.	Wen Lv, Bing Li, Existence and Global Attractivity of Pseudo Almost Periodic Solutions for Clifford-Valued Fuzzy Neural Networks with Proportional Delays, 2021, 9, 2227-7390, 3306, 10.3390/math9243306
3.	Yongkun Li, Xiaohui Wang, Almost periodic solutions in distribution of Clifford-valued stochastic recurrent neural networks with time-varying delays, 2021, 153, 09600779, 111536, 10.1016/j.chaos.2021.111536
4.	Shiping Shen, Xiaofang Meng, Li Yang, Pseudo Almost Periodic Synchronization of OVCNNs with Time-Varying Delays and Distributed Delays on Time Scales, 2024, 23, 1575-5460, 10.1007/s12346-023-00885-x
5.	António J.G. Bento, José J. Oliveira, César M. Silva, Existence and stability of a periodic solution of a general difference equation with applications to neural networks with a delay in the leakage terms, 2023, 126, 10075704, 107429, 10.1016/j.cnsns.2023.107429

Reader Comments

Your name:*

Email:*
© 2022 the Author(s), licensee AIMS Press. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0)