Measurement data intrusion detection in industrial control systems based on unsupervised learning

Sohrab Mokhtari; Kang K Yen; Sohrab Mokhtari; Kang K Yen

doi:10.3934/aci.2021004

Applied Computing and Intelligence

2021, Volume 1, Issue 1: 61-74. doi: 10.3934/aci.2021004

Previous Article Next Article

Research article

Measurement data intrusion detection in industrial control systems based on unsupervised learning

Sohrab Mokhtari ^,,
Kang K Yen

Electrical and Computer Engineering Department, Florida International University, 11200 SW 8th St, Miami, FL 33199, USA

Academic Editor: Chih-Cheng Hung

Received: 23 September 2021 Accepted: 06 December 2021 Published: 10 December 2021

Anomaly detection strategies in industrial control systems mainly investigate the transmitting network traffic called network intrusion detection system. However, The measurement intrusion detection system inspects the sensors data integrated into the supervisory control and data acquisition center to find any abnormal behavior. An approach to detect anomalies in the measurement data is training supervised learning models that can learn to classify normal and abnormal data. But, a labeled dataset consisting of abnormal behavior, such as attacks, or malfunctions is extremely hard to achieve. Therefore, the unsupervised learning strategy that does not require labeled data for being trained can be helpful to tackle this problem. This study evaluates the performance of unsupervised learning strategies in anomaly detection using measurement data in control systems. The most accurate algorithms are selected to train unsupervised learning models, and the results show an accuracy of 98% in stealthy attack detection.

Keywords:

Citation: Sohrab Mokhtari, Kang K Yen. Measurement data intrusion detection in industrial control systems based on unsupervised learning[J]. Applied Computing and Intelligence, 2021, 1(1): 61-74. doi: 10.3934/aci.2021004

Related Papers:

[1]	Anil Chavada, Nimisha Pathak . Transmission dynamics of breast cancer through Caputo Fabrizio fractional derivative operator with real data. Mathematical Modelling and Control, 2024, 4(1): 119-132. doi: 10.3934/mmc.2024011
[2]	Erick Manuel Delgado Moya, Diego Samuel Rodrigues . Fractional order modeling for injectable and oral HIV pre-exposure prophylaxis. Mathematical Modelling and Control, 2023, 3(2): 139-151. doi: 10.3934/mmc.2023013
[3]	C. Kavitha, A. Gowrisankar . Fractional integral approach on nonlinear fractal function and its application. Mathematical Modelling and Control, 2024, 4(3): 230-245. doi: 10.3934/mmc.2024019
[4]	Chunyu Tian, Lei Sun . Partitioning planar graphs with girth at least $9$ into an edgeless graph and a graph with bounded size components. Mathematical Modelling and Control, 2021, 1(3): 136-144. doi: 10.3934/mmc.2021012
[5]	Zhen Lin . On the sum of powers of the $A_{\alpha}$ -eigenvalues of graphs. Mathematical Modelling and Control, 2022, 2(2): 55-64. doi: 10.3934/mmc.2022007
[6]	Jiaming Dong, Huilan Li . Hopf algebra of labeled simple graphs arising from super-shuffle product. Mathematical Modelling and Control, 2024, 4(1): 32-43. doi: 10.3934/mmc.2024004
[7]	Qian Lin, Yan Zhu . Unicyclic graphs with extremal exponential Randić index. Mathematical Modelling and Control, 2021, 1(3): 164-171. doi: 10.3934/mmc.2021015
[8]	Zhen Lin, Ting Zhou . Degree-weighted Wiener index of a graph. Mathematical Modelling and Control, 2024, 4(1): 9-16. doi: 10.3934/mmc.2024002
[9]	Ihtisham Ul Haq, Nigar Ali, Shabir Ahmad . A fractional mathematical model for COVID-19 outbreak transmission dynamics with the impact of isolation and social distancing. Mathematical Modelling and Control, 2022, 2(4): 228-242. doi: 10.3934/mmc.2022022
[10]	Yunhao Chu, Yansheng Liu . Approximate controllability for a class of fractional semilinear system with instantaneous and non-instantaneous impulses. Mathematical Modelling and Control, 2024, 4(3): 273-285. doi: 10.3934/mmc.2024022

Abstract

1. Introduction

In various fields of science, nonlinear evolution equations practically model many natural, biological and engineering processes. For example, PDEs are very popular and are used in physics to study traveling wave solutions. They have played a crucial role in illustrating the nature of nonlinear problems. PDEs are collected to control the diffusion of chemical reactions. In biology, they play a fundamental role in describing various phenomena, such as population growth. In addition, natural phenomena such as fluid dynamics, plasma physics, optics and optical fibers, electromagnetism, quantum mechanics, ocean waves, and others are studied using PDEs. The qualitative and quantitative characteristics of these phenomena can be identified from the behaviors and shapes of their solutions. Therefore, finding the analytic solutions to such phenomena is a fundamental topic in mathematics. Scientists have developed sparse fundamental approaches to find analytic solutions for nonlinear PDEs. Among these techniques, I present integration methods from ^[1] and ^[2], the modified F-expansion and Generalized Algebraic methods, respectively. Bekir and Unsal ^[3] proposed the first integral method to find the analytical solution of nonlinear equations. Kumar, Seadawy and Joardar ^[4] used the improved Kudryashov technique to extract fractional differential equations. Adomian ^[5] proposed the Adomian decomposition technique to find the solution of frontier problems of physics. ^[6] uses an exploratory method to find explicit solutions of non-linear PDEs. Many different methods of solving equations arising from natural phenomena and some of their analytic solutions, such as dark and light solitons, non-local rogue waves, an occasional wave and mixed soliton solutions, are exhibited and can be found in ^{[7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40]}.

The Novikov-Veselov (NV) system ^[41,42] is given by

$\begin{equation} \begin{aligned} \Psi_t& +\alpha \Gamma_{xxy}+\beta \Phi_{xyy}+\gamma \Gamma_y \Phi+\gamma \left(\dfrac{\Psi^2}{2}\right)_y+\lambda\, \Gamma\, \Phi_x+\lambda\, \left(\dfrac{\Psi^2}{2}\right)_x = 0, \\ \Gamma_y& = \Psi_x, \\ \Psi_y& = \Phi_x, \\ \end{aligned} \end{equation}$

(1.1)

where $\alpha, \, \beta, \, \gamma$ and $\lambda$ are constants. Barman ^[42] declared that Eq (1.1) is involved to represent tidal and tsunami waves, electro-magnetic waves in communication cables and magneto-sound and ion waves in plasma. In ^[42], the generalized Kudryashov method was utilized to have traveling wave solutions for Eq (1.1). According to Croke ^[43], the Novikov-Veselov system is generalized from the KdV equations which were examined by Novikov and Veselov. Croke ^[43] used several approaches, (the extended mapping, the Hirota and the extended tanh-function approaches) in the proposed system to achieve numerous soliton solutions, such as breathers, and constrained analytic solutions. Boiti, Leon, and Manna ^[44] applied the inverse dispersion technique to solve (1.1) for a particular type of initial value. Numerical solutions and a study of the stability of solutions for the proposed equation were presented by Kazeykina and Klein ^[45]. The Nizhnik-Novikov-Veselov system for two dimensions was also solved using the Kansa technique to find the numerical results ^[46]. To the best of my knowledge, the stability and error analysis of the numerical scheme presented here has not yet been discussed for system (1.1). Therefore, this has motivated me enormously to do so. The primary purpose is to obtain multiple analytic solutions to system (1.1) by using both the modified F-expansion and Generalized Algebraic methods. In connection with the numerical solution, the method of finite differences is utilized to achieve numerical results for the studied system. I graphically and analytically compare the traveling wave solutions and numerical results. Undoubtedly, the presented results strongly contribute to describing physical problems in practice.

The outline of this article is provided in this paragraph. Section 2 summarizes the employed methods. All the analytic solutions are extracted in Section 3. The shooting and BVP results for the proposed system are presented in Section 4. In addition, I examine the numerical solution of the system (1.1) in Section 5. Sections 6 and 7 study the stability and error analysis of the numerical scheme, respectively. Section 8 presents the results and discussion.

2. Summary of proposed methods

Considering the development equation with physical fields $\Psi(x, y, t)$ , $\Phi(x, y, t)$ and $\Gamma(x, y, t)$ in the variables $x,$ $y$ and $t$ is given in the following form:

$\begin{equation} Q_1(\Psi, \Psi_t, \Psi_x, \Psi_y, \Gamma, \Gamma_y, \Gamma_{xxy}, \Phi, \Phi_{xyy}, \Phi_{x}, \dots) = 0. \end{equation}$

(2.1)

Step 1. We extract the traveling-wave solutions of System (1.1) that are formed as follows:

$\begin{equation} \begin{aligned} \Phi(x, y, t) = &\phi(\eta) , &&\qquad&\eta = x+y-wt, \\ \Psi(x, y, t) = &\psi(\eta), \\ \Gamma(x, y, t) = &\Theta(\eta), \end{aligned} \end{equation}$

(2.2)

where $w$ is the wave speed.

Step 2. The nonlinear evolution (2.1) is reduced to the following ODE:

$\begin{equation} Q_2(\psi, \psi_\eta, \Theta, \Theta_\eta, \Theta_{\eta\eta\eta}, \phi, \phi_{\eta\eta\eta}, \phi_{\eta}, \dots) = 0, \end{equation}$

(2.3)

where $Q_2$ is a polynomial in $\psi(\eta), \; \phi(\eta),$ $\Theta(\eta)$ and their total derivatives.

2.1. The modified F-expansion method

According to the modified F-expansion method, the solutions of (2.3) are given by the form

$\begin{equation} \psi(\eta) = \rho_0+\sum\limits_{k = 1}^{N} \left( \rho_{k} F(\eta)^{k}+\dfrac{ q_{k}}{F(\eta)^{k}} \right) , \end{equation}$

(2.4)

and $F(\eta)$ is a solution of the following differential equation:

$\begin{equation} F^{\prime}(\eta) = \mu_0+\mu_1 F(\eta)+\mu_2 F(\eta)^{2}, \end{equation}$

(2.5)

where $\mu_0$ , $\mu_1$ , $\mu_2$ , are given in ^[1], and $\rho_k$ , $q_k$ are to be determined later.

Table 1. The relations among

$\mu_0$ ,

$\mu_1$ ,

$\mu_2$ and the function

$F(\eta)$ .

$\mu_0$	$\mu_{1}$	$\mu_{2}$	$F(\eta)$
$\mu_0=0,$	$\mu_1=1,$	$\mu_2=-1,$	$F(\eta)=\dfrac{1}{2}+\dfrac{1}{2} \tanh \left(\dfrac{1}{2} \eta\right).$
$\mu_0=0,$	$\mu_1=-1,$	$\mu_2=1,$	$F(\eta)=\dfrac{1}{2}-\dfrac{1}{2} \operatorname{coth}\left(\dfrac{1}{2} \eta\right).$
$\mu_0=\dfrac{1}{2},$	$\mu_1=0,$	$\mu_2=\dfrac{-1}{2},$	$F(\eta)=\operatorname{coth}(\eta) \pm \operatorname{csch}(\eta),$ $\tanh (\eta) \pm \operatorname{sech}(\eta).$
$\mu_0=1,$	$\mu_1=0,$	$\mu_2=-1,$	$F(\eta)=\tanh (\eta),$ $\operatorname{coth}(\eta).$
$\mu_0=\dfrac{1}{2},$	$\mu_1=0,$	$\mu_2=\dfrac{1}{2},$	$F(\eta)=\sec (\eta)+\tan (\eta),$ $\csc (\eta)-\cot (\eta).$
$\mu_0=\dfrac{1}{2},$	$\mu_1=0,$	$\mu_2=\dfrac{1}{2},$	$F(\eta)=\sec (\eta)-\tan (\eta),$ $\csc (\eta)+\cot (\eta).$
$\mu_0=\pm1,$	$\mu_1=0,$	$\mu_2=\pm1,$	$F(\eta)=\tan (\eta),$ $\cot (\eta).$

| Show Table

DownLoad: CSV

2.2. The generalized algebraic method

According to the generalized direct algebraic method, the solutions of (2.3) are given by

$\begin{equation} \psi(\eta) = \nu_0+\sum\limits_{k = 1}^{N} \left( \nu_{k} G(\eta)^{k} +\dfrac{ r_{k}}{G(\eta) ^{k}} \right) , \end{equation}$

(2.6)

and $G(\eta)$ is a solution of the following differential equation:

$\begin{equation} G^{\prime}(\eta) = \varepsilon \sqrt{\sum\limits_{k = 0}^{4} \delta_{k} G^{k}(\eta)}, \end{equation}$

(2.7)

where $\nu_k,$ and $r_k$ are to be determined, and $N$ is an integer number obtained by the highest degree of the nonlinear terms and the highest order of the derivatives. $\varepsilon$ is user-specified, usually taken with $\varepsilon = \pm1$ , and $\delta_k$ , $k = 0, 1, 2, 3, 4,$ are given in Table 2 ^[2].

Table 2. The relations among

$\delta_k$ ,

$k = 0, 1, 2, 3, 4,$ and the function

$G(\eta)$ .

$\delta_{0}$	$\delta_{1}$	$\delta_{2}$	$\delta_{3}$	$\delta_{4}$	$G(\eta)$
$\delta_{0}=0$ ,	$\delta_{1}=0$ ,	$\delta_{2} > 0$ ,	$\delta_{3}=0$ ,	$\delta_{4} < 0$ ,	$G(\eta)= \varepsilon$ $\sqrt{-\dfrac{\delta_{2}}{\delta_{4}}} \operatorname{sech}\left(\sqrt{\delta_{2}} \eta\right)$ .
$\delta_{0}=\dfrac{\delta_{2}^{2}}{4 c 4}$ ,	$\delta_{1}=0$ ,	$\delta_{2} < 0$ ,	$\delta_{3}=0$ ,	$\delta_{4} > 0$ ,	$G(\eta)=\varepsilon \sqrt{-\dfrac{\delta_{2}}{2 \delta_{4}}} \tanh \left(\sqrt{-\dfrac{\delta_{2}}{2}} \eta\right)$ .
$\delta_{0}=0$ ,	$\delta_{1}=0$ ,	$\delta_{2} < 0$ ,	$\delta_{3}=0$ ,	$\delta_{4} > 0$ ,	$G(\eta)=\varepsilon$ $\sqrt{-\dfrac{\delta_{2}}{\delta_{4}}}$ $\sec$ $\left(\sqrt{-\delta_{2}} \eta\right)$ .
$\delta_{0}=\dfrac{\delta_{2}^{2}}{4 \delta_{4}},$	$\delta_{1}=0$ ,	$\delta_{2} > 0$ ,	$\delta_{3}=0,$	$\delta_{4} > 0,$	$G(\eta)=\varepsilon$ $\sqrt{\dfrac{\delta_{2}}{2 \delta_{4}}} \tan \left(\sqrt{\dfrac{\delta_{2}}{2}} \eta\right)$ .
$\delta_{0}=0,$	$\delta_{1}=0,$	$\delta_{2}=0,$	$\delta_{3}=0,$	$\delta_{4} > 0,$	$G(\eta)=-\dfrac{\varepsilon}{\sqrt{\delta_{4}} \eta}$ .
$\delta_{0}=0,$	$\delta_{1}=0,$	$\delta_{2} > 0,$	$\delta_{3}\neq 0,$	$\delta_{4}=0,$	$G(\eta)=-\dfrac{\delta_{2}}{\delta_{3}}. \operatorname{sech}^{2}\left(\dfrac{\sqrt{\delta_{2}}}{2} \eta\right)$ .

| Show Table

DownLoad: CSV

3. Methodology

Consider the Novikov-Veselov (NV) system

(3.1)

a system of PDEs in the unknown functions $\Psi = \Psi(x, y, t), \; \Phi = \Phi(x, y, t)$ , $\Gamma = \Gamma(x, y, t)$ and their partial derivatives. I plug the transformations

$\begin{equation} \begin{aligned} \Phi(x, y, t) = &\phi(\eta) , &&\qquad&\eta = x+y-wt, \\ \Psi(x, y, t) = &\psi(\eta), \\ \Gamma(x, y, t) = &\Theta(\eta), \end{aligned} \end{equation}$

(3.2)

into Eq (3.1) to reduce it to a system of ODEs given by

$\begin{equation} \begin{aligned} -w\; \psi_\eta& +\alpha \Theta_{\eta\eta\eta}+\beta \phi_{\eta\eta\eta}+\gamma \Theta_\eta \phi+\gamma \left(\dfrac{\psi^2}{2}\right)_\eta+\lambda\, \Theta\, \phi_\eta+\lambda\, \left(\dfrac{\psi^2}{2}\right)_\eta = 0, \\ \Theta_\eta& = \psi_\eta, \\ \psi_\eta& = \phi_\eta.\\ \end{aligned} \end{equation}$

(3.3)

Integrating $\Theta_\eta = \psi_\eta$ and $\phi_\eta = \psi_\eta$ yields

$\begin{equation} \Theta = \psi, \qquad \text{and } \qquad \phi = \psi. \end{equation}$

(3.4)

Substituting (3.4) into the first equation of (3.3) and integrating once with respect to $\eta$ yields

$\begin{equation} \begin{aligned} -w\; \psi& +\left( \alpha +\beta\right) \psi_{\eta\eta}+\left( \gamma+\lambda\right) \psi^2 = 0. \end{aligned} \end{equation}$

(3.5)

Balancing $\psi_{\eta\eta}$ with $\psi^2$ in (3.5) calculates the value of $N = 2.$

3.1. Application of modified F-expansion method

According to the modified F-expansion method with $N = 2$ , the solutions of (3.5) are

$\begin{equation} \psi(\eta) = \rho_0+ \rho_{1} F(\eta)+\dfrac{ q_{1}}{F(\eta)}+ \rho_{2} F(\eta)^{2}+\dfrac{ q_{2}}{F(\eta)^{2}} , \end{equation}$

(3.6)

and $F(\eta)$ is a solution of the following differential equation:

$\begin{equation} F^{\prime}(\eta) = \mu_0+\mu_1 F(\eta)+\mu_2 F(\eta)^{2}, \end{equation}$

(3.7)

where $\mu_0$ , $\mu_1$ , $\mu_2$ are given in Table 1. To explore the analytic solutions to (3.5), I ought to follow the subsequent steps.

$\textbf{Step 1.}$ Placing (3.6) along with (3.7) into Eq (3.5) and gathering the coefficients of $F(\eta)^j$ , $j$ = $-4, -3, -2, -1, 0, 1, 2, 3, 4,$ to zeros gives a system of equations for $\rho_0, \; \rho_k, \; q_k$ , $k = 1, 2.$

$\textbf{Step 2.}$ Solve the resulting system using mathematical software: for example, Mathematica or Maple.

$\textbf{Step 3.}$ Choosing the values of $\mu_0, \; \mu_1$ and $\mu_2$ and the function $F(\eta)$ from and substituting them along with $\rho_0, \; \rho_k, \; q_k$ , $k = 1, 2,$ in (3.6) produces a set of trigonometric function and rational solutions to (3.5).

Applying the above steps, I determine the values of $\rho_0, \; \rho_1, \; \rho_2, \; q_1, \; q_2$ and $w$ as follows:

$(1).$ When $\mu_0 = 0,$ $\mu_1 = 1$ and $\mu_2 = -1$ , I have two cases.

$\textbf{Case 1.}$

$\begin{equation} \begin{aligned} \rho_0 = &0, && \rho_1 = \dfrac{6 (\alpha +\beta)}{\gamma +\lambda}, &\rho_2 = & -\dfrac{6 (\alpha +\beta)}{\gamma + \lambda}, & q_1 = q_2 = 0, \quad \text{and}\quad w = \alpha+\beta. \end{aligned} \end{equation}$

(3.8)

The solution is given by

$\begin{equation} \begin{aligned} \Psi_1(x, y, t) = \dfrac{ 3(\alpha+\beta)}{2(\gamma+\lambda)} \operatorname{sech}^2\left(\dfrac{1}{2} \left( x+y-(\alpha +\beta) t+x0\right) \right). \end{aligned} \end{equation}$

(3.9)

$\textbf{Case 2.}$

$\rho_0 = -\dfrac{\alpha+\beta}{\gamma+\lambda}, \rho_1 = \dfrac{6 (\alpha +\beta)}{\gamma +\lambda}, \\ \rho_2 = -\dfrac{6 (\alpha +\beta)}{\gamma + \lambda}, q_1 = q_2 = 0, \quad \text{and}\quad w = -\left( \alpha+\beta\right) .$

(3.10)

The solution is given by

$\begin{equation} \begin{aligned} \Psi_2(x, y, t) = -\dfrac{ (\alpha+\beta)}{2(\gamma+\lambda)} \left( 3 \tanh ^2\left(\dfrac{1}{2} (x+y+t (\alpha+\beta))+x0\right)-1 \right) . \end{aligned} \end{equation}$

(3.11)

presents the time evolution of the analytic solutions $(a)$ $\Psi_1$ and $(b)$ $\Psi_2$ with $t = 0, 10, 20$ . The parameter values are $x0 = -20, \ \alpha = 0.50, \beta = 0.6, \ \gamma = -1.5,$ and $\lambda = 1.$ presents the wave behavior by changing a certain parameter value and fixing the values of the others. presents the behavior of $\Psi_1$ when I change the values of (a) $\alpha$ or $\beta$ and (b) $\gamma$ or $\lambda$ . In it can also be seen that the value of $\alpha$ or $\beta$ affects the direction and amplitude of the wave, such that a negative value always makes the wave negative, its amplitude decreases when $\alpha, \, \beta \to 0$ , and its amplitude increases when $\alpha, \, \beta \to \infty.$ In the value of $\gamma$ or $\lambda$ affects the direction and amplitude of the wave, such that a negative value always makes the wave negative, and its amplitude decreases when the value of $\gamma$ or $\lambda$ increases. In and present the wave behavior of $\Psi_2.$

Figure 1. Time evolution of the analytic solutions

$(a)$

$\Psi_1$ and

$(b)$

$\Psi_2$ with

$t = 0, 10, 20$ . The parameter are given by

$x0 = -20, \ \alpha = 0.50, \beta = 0.6, \ \gamma = -1.5,$ and

$\lambda = 1$ .

DownLoad: Full-Size Img PowerPoint

Figure 2. This figure present the wave behavior when changing a certain parameter value and fixing the values of the others.

$(a)$ presents the behavior when I change the value of

$\alpha$ or

$\beta,$ and

$(b)$ presents when I change the value of

$\gamma$ or

$\lambda$ for the solution

$\Psi_1.$

$(c)$ and

$(d)$ are for

$\Psi_2$ .

DownLoad: Full-Size Img PowerPoint

$(2).$ When $\mu_0 = 0,$ $\mu_1 = -1,$ and $\mu_2 = 1$ , I have two cases.

$\textbf{Case 3.}$ The solution is given by

$\begin{equation} \begin{aligned} \Psi_3(x, y, t) = -\dfrac{ 3(\alpha+\beta)}{2(\gamma+\lambda)} \operatorname{csch}^2\left(\dfrac{1}{2} \left( x+y-(\alpha +\beta) t\right) \right). \end{aligned} \end{equation}$

(3.12)

$\textbf{Case 4.}$ The solution is given by

$\begin{equation} \begin{aligned} \Psi_4(x, y, t) = -\dfrac{ (\alpha+\beta)}{2(\gamma+\lambda)} \left( 3 \coth ^2\left(\dfrac{1}{2} (x+y+t (\alpha+\beta))\right)-1 \right) . \end{aligned} \end{equation}$

(3.13)

$(3).$ When $\mu_0 = 1,$ $\mu_1 = 0,$ and $\mu_2 = -1$ , I have

$\textbf{Case 5.}$ The solution is given by

$\begin{equation} \begin{aligned} \Psi_5(x, y, t) = -\dfrac{8 (\alpha+\beta) }{\gamma+\lambda}(\cosh (4 (16 t (\alpha+\beta)+x+y))+2) \operatorname{csch}^2(2 (16 t (\alpha+\beta)+x+y)). \end{aligned} \end{equation}$

(3.14)

$(4).$ When $\mu_0 = \pm 1,$ $\mu_1 = 0,$ and $\mu_2 = \pm 1$ , I have one case.

$\textbf{Case 6.}$ The solution is given by

$\begin{equation} \begin{aligned} \Psi_6(x, y, t) = -\dfrac{24 (\alpha+\beta) }{\gamma+\lambda}\csc ^2(2 (16 t (\alpha+\beta)+x+y)) . \end{aligned} \end{equation}$

(3.15)

3.2. Application of the generalized algebraic method

According to the generalized algebraic method, the solutions of (3.5) are given by the form

$\begin{equation} \psi(\eta) = \nu_0+ \nu_{1} G(\eta)+ \nu_{2} G(\eta)^{2}+\dfrac{ r_{1}}{G(\eta)}+\dfrac{ r_{2}}{G(\eta)^{2}}, \end{equation}$

(3.16)

where $\nu_k$ , $r_k$ are to be determined later. $G(\eta)$ is a solution of the following differential equation:

$\begin{equation} G^{\prime}(\eta) = \varepsilon \sqrt{\sum\limits_{k = 0}^{4} \delta_{k} G^{k}(\eta)}, \end{equation}$

(3.17)

where $\delta_k$ , $k = 0, 1, 3, 4$ , are given in . In all the cases mentioned above and the subsequent solutions, I used the mathematical software Mathematica to find the values of the constants $\nu_0, \, \nu_1, \, \nu_2, \, r_1, \, r_2$ and $w$ . Thus, the analytic solutions to (3.5) using the generalized algebraic method will be presented here with different values of the constants $\delta_k$ , $k = 0, 1, 3, 4$ .

$(5).$ When $\delta_0 = \dfrac{\delta^2_2}{4 \delta_4},$ $\delta_1 = \delta_3 = 0,$ $\delta_2 < 0$ , $\delta_4 > 0$ , and $\varepsilon = \pm1,$

$\begin{equation} \begin{aligned} \nu_0 = &\frac{\pm\sqrt{4 \delta_2^2 \varepsilon^4 \left(\alpha+\beta\right)^2 \left(\gamma+\lambda\right)^2-3 \delta_2 \varepsilon^4 \left(\alpha+\beta\right)^2 \left(\gamma+\lambda\right)^2}-2 \delta_2 \varepsilon^2 \left(\alpha+\beta\right) \left(\gamma+\lambda\right)}{\left(\gamma+\lambda\right)^2}, \\ \nu_2 = &-\frac{6 \delta_4 \varepsilon^2 \left(\alpha+\beta\right)}{\left(\gamma+\lambda\right)}, \\ \nu_1 = &r_1 = r_2 = 0, \quad\varepsilon = \pm1.\\ w = &\pm\frac{2 \sqrt{\left(4 \delta_2^2-3 \delta_2\right) \varepsilon^4 \left(\alpha+\beta\right)^2 \left(\gamma+\lambda\right)^2}}{\left(\gamma+\lambda\right)}. \end{aligned} \end{equation}$

(3.18)

$\textbf{Case 7.}$ The solution is given by

$\Psi_7(x, y, t) = -\frac{1}{{(\gamma+\lambda)^2}}\\ \left( -3 \delta_2 \varepsilon^4 (\alpha+\beta) (\gamma+\lambda) \tanh ^2\left(\frac{\sqrt{-\delta_2} \left(\frac{2 t \sqrt{\delta_2 (4 \delta_2-3) \varepsilon^4 (\alpha+\beta)^2 (\gamma+\lambda)^2}}{\gamma+\lambda}+x+y\right)}{\sqrt{2}}\right)\right.\\ \left.+2 \delta_2 \varepsilon^2 (\alpha+\beta) (\gamma+\lambda)+\sqrt{\delta_2 (4 \delta_2-3) \varepsilon^4 (\alpha+\beta)^2 (\gamma+\lambda)^2}\right) .$

(3.19)

shows the time evolution of the analytic solutions. shows $\Psi_7$ with $t$ = $0:2:6$ . The parameter values are $\delta_2 = -1$ , $\delta_4 = 1,$ $\epsilon = -1,$ $\alpha = 0.50,$ $\beta = 0.6,$ $\gamma = -1.5,$ $\lambda = 1.8$ and $x0 = -10.$ shows $\Psi_8$ with $t = 0:2:8$ . The parameter values are $\delta_2 = 1$ , $\delta_4 = -1,$ $\epsilon = -1,$ $\alpha = 0.50,$ $\beta = 0.6,$ $\gamma = -1.5,$ $\lambda = 1.8$ and $x0 = -10.$ – present the $3D$ time evolution of the analytic solutions $\Psi_2$ (left) and the numerical solutions (right) obtained employing the scheme 5.1 with $t = 5, 15, 25$ , $M_x = 1600,$ $N_y = 100,$ $x = 0\to60$ and $y = 0\to1.$

Figure 3. Time evolution of the analytic solutions.

$(a)$

$\Psi_7$ with

$t = 0:2:6$ . The parameter values are

$\delta_2 = -1$ ,

$\delta_4 = 1,$

$\epsilon = -1,$

$\alpha = 0.50,$

$\beta = 0.6,$

$\gamma = -1.5,$

$\lambda = 1.8$ and

$x0 = -10.$

$(b)$

$\Psi_8$ with

$t = 0:2:8$ . The parameter values are

$\delta_2 = 1$ ,

$\delta_4 = -1,$

$\epsilon = -1,$

$\alpha = 0.50,$

$\beta = 0.6,$

$\gamma = -1.5,$

$\lambda = 1.8$ and

$x0 = -10$ .

DownLoad: Full-Size Img PowerPoint

Figure 4. 3D graphs presenting the analytic (left) and the numerical (right) solutions of

$\Psi_2(x, y, t)$ at

$t = 5.$ The figures present the strength of agreement between analytic and numerical solutions.

DownLoad: Full-Size Img PowerPoint

Figure 5. 3D graphs presenting the analytic (left) and the numerical (right) solutions of

$\Psi_2(x, y, t)$ at

$t = 15.$ The figures present the strength of agreement between analytic and numerical solutions.

DownLoad: Full-Size Img PowerPoint

Figure 6. 3D graphs presenting the analytic (left) and the numerical (right) solutions of

$\Psi_2(x, y, t)$ at

$t = 25.$ The figures present the strength of agreement between analytic and numerical solutions.

DownLoad: Full-Size Img PowerPoint

$(6).$ When $\delta_0 = 0,$ $\delta_1 = \delta_3 = 0,$ $\delta_2 > 0$ , $\delta_4 < 0$ , and $\varepsilon = \pm1,$

$\begin{equation} \begin{aligned} \nu_2 = &-\frac{6 \delta_4 \varepsilon^2 (\alpha+\beta)}{\gamma+\lambda}, \quad \nu_1 = \nu_0 = r_1 = r_2 = 0, \quad w = &4 \delta_2 \varepsilon^2 (\alpha+\beta). \end{aligned} \end{equation}$

(3.20)

$\textbf{Case 8.}$ The solution is given by

$\begin{equation} \begin{aligned} \Psi_8(x, y, t) = &\frac{6 \delta_2\epsilon^4 (\alpha+\beta) \operatorname{sech}^2\left(\sqrt{\delta_2} \left(-4 \delta_2\epsilon^2 t (\alpha+\beta)+x+y\right)\right)}{\gamma+\lambda} \end{aligned} \end{equation}$

(3.21)

$(7).$ When $\delta_0 = 0,$ $\delta_1 = \delta_4 = 0,$ $\delta_3\neq0,$ $\delta_2 > 0$ , $\varepsilon = \pm1$

$\begin{equation} \begin{aligned} \text{Set 1.}\, \nu_1 = & -\tfrac{3 \delta_3 \varepsilon^2 \left(\alpha+\beta\right)}{2 \left(\gamma+\lambda\right)}, \, \nu_0 = \nu_2 = r_1 = r_2 = 0, \, w = \delta_2 \varepsilon^2 \left(\alpha+\beta\right). \\ &\\ \text{Set 2.}\, \nu_0 = & -\tfrac{\delta_2 \varepsilon^2 \left(\alpha+\beta\right)}{\left(\gamma+\lambda\right)}, \, \nu_1 = -\tfrac{3 \delta_3 \varepsilon^2 \left(\alpha+\beta\right)}{2 \left(\gamma+\lambda\right)}, \quad\nu_2 = r_1 = r_2 = 0, \, w = -\delta_2 \varepsilon^2 \left(\alpha+\beta\right). \end{aligned} \end{equation}$

(3.22)

$\textbf{Case 9.}$ The solution is given by

$\begin{equation} \begin{aligned} \Psi_9(x, y, t) = \frac{3 \delta_2 (\alpha+\beta) \operatorname{sech}^2\left(\frac{1}{2} \sqrt{\delta_2} (\delta_2 t (-(\alpha+\beta))+x+y)\right)}{2 (\gamma+\lambda)}. \end{aligned} \end{equation}$

(3.23)

$\textbf{Case 10.}$ The solution is given by

$\begin{equation} \begin{aligned} \Psi_{10}(x, y, t) = \frac{3 \delta_2 (\alpha+\beta) \operatorname{sech}^2\left(\frac{1}{2} \sqrt{\delta_2} (\delta_2 t (\alpha+\beta)+x+y)\right)}{2 (\gamma+\lambda)}-\frac{\delta_2 (\alpha+\beta)}{\gamma+\lambda}. \end{aligned} \end{equation}$

(3.24)

4. Numerical solution

In this section I extract numerical solutions to the resulting ODE system (3.5) using several numerical methods. The purpose of this procedure is to guarantee the accuracy of the analytic solutions. I picked one of the analytic solutions above to be a sample, (3.11). The nonlinear shooting and BVP methods, at $t = 0,$ are used by taking the value of $\psi$ at the right endpoint of the domain $\eta = 0$ with guessing the initial value for $\psi_\eta.$ The new target is to obtain the second boundary condition of $\psi$ at the left endpoint of a particular domain. Once the numerical result is obtained, I compare it with the analytic solution (3.11). The MATLAB solver ODE15s and FSOLVE ^[47] are used to get the numerical solution. The resulting ODE (3.5) is discretized as

$\begin{equation} f(\psi) = 0, \qquad f(\psi_i) = -w\; \psi_i +\dfrac{\alpha +\beta}{\Delta_\eta} \left( \psi_{i+1}-2\psi_i+\psi_{i-1}\right) +\dfrac{ \gamma+\lambda}{2\Delta_\eta} \left( \psi^2_{i+1}-\psi_{i-1}^2\right), \end{equation}$

(4.1)

for the BVP method and

$\begin{equation} \psi_{\eta\eta} = \dfrac{1}{\alpha+\beta}\left( w\psi-\left( \gamma+\lambda\right) \psi^2\right), \end{equation}$

(4.2)

for the shooting method. Figure 7 presents the comparison between the numerical solutions obtained using the above numerical methods and the analytic solution. Figure 7 shows that the solutions are identical to the analytic solution.

Figure 7. Comparing the numerical solutions resulting from the shooting and BVP methods with the analytic solution (3.11) at

$t = 0.$ The parameter values are taken as

$\alpha = 0.50$ ,

$\beta = 0.6$ ,

$\gamma = -1.5, \lambda = 1.8$ , with

$N = 600$ .

DownLoad: Full-Size Img PowerPoint

Thus, it is possible to verify the correctness of the analytic solution. I also accept the obtained numerical solution as an initial condition for the numerical scheme in the next section.

5. Numerical scheme on a fixed mesh

In this section, I use the finite-difference method to obtain the numerical results of system (1.1) over the domain $[a, \, b]\times [c, \, d].$ Here, $a, \; b, \; c$ and $d$ represent the endpoints of the rectangular domain in the $x$ and $y$ directions, respectively, and $T_f$ is a certain time. The domain $[a, \, b]\times [c, \, d]$ is split into $(M_x+1)\times (N_y+1)$ mesh points:

$\begin{equation*} \label{} \begin{aligned} x_m = &a+m\, \Delta_x, &&& m = 0, \, 1, \, 2, \, \dots, \, M_x, \\ y_n& = c+n\, \Delta_y, &&& n = 0, \, 1, \, 2, \, \dots, \, N_y, \end{aligned} \end{equation*}$

where $\Delta_x$ and $\Delta_y$ are the step-sizes of the $x$ and $y$ domains, respectively. The system (1.1) is converted to an ODE system by discretizing the space derivatives while keeping the time derivative continuous. Completing this yields

$\begin{equation} \begin{aligned} &\Psi_t|_{m, n}^k + \dfrac{\alpha}{2\Delta_y \Delta_x^2} \delta_x^2\left( \Gamma^{k+1}_{m, n+1}-\Gamma^{k+1}_{m, n-1}\right) +\dfrac{\beta}{2\Delta_x\Delta_y^2} \delta_y^2\left( \Phi^{k+1}_{m+1, n}-\Phi^{k+1}_{m-1, n}\right)\\ &\qquad \quad- \frac{ \gamma}{4\Delta_y} \left( \left( \Phi^{k+1}_{m, n+1}+\Phi^{k+1}_{m, n} \right) \Gamma^{k+1}_{m, n+1}- \left( \Phi^{k+1}_{m, n}+\Phi^{k+1}_{m, n-1} \right)\Gamma^{k+1}_{m, n-1}\right)\\ &\qquad \quad- \frac{ \lambda}{4\Delta_x} \left( \left( \Gamma^{k+1}_{m+1, n}+\Gamma^{k+1}_{m, n} \right) \Phi^{k+1}_{m+1, n}- \left( \Gamma^{k+1}_{m, n}+\Gamma^{k+1}_{m-1, n} \right)\Phi^{k+1}_{m-1, n}\right)\\ &\qquad \quad+\frac{ \gamma}{4\Delta_y} \left( \left( \Psi^{k+1}_{m, n+1}\right)^2- \left( \Psi^{k+1}_{m, n-1}\right)^2\right) +\frac{ \lambda}{4\Delta_x} \left( \left( \Psi^{k+1}_{m+1, n}\right)^2- \left( \Psi^{k+1}_{m-1, n}\right)^2\right) = 0, \\ &\frac{1}{2\Delta_y}(\Gamma_{m, n+1}^{k+1}-\Gamma_{m, n-1}^{k+1}) = \frac{1}{2\Delta_x}(\Psi_{m+1, n}^{k+1}-\Psi_{m-1, n}^{k+1}), \\ &\frac{1}{2\Delta_y}(\Psi_{m, n+1}^{k+1}-\Psi_{m, n-1}^{k+1}) = \frac{1}{2\Delta_x}(\Phi_{m+1, n}^{k+1}-\Phi_{m-1, n}^{k+1}), \end{aligned} \end{equation}$

(5.1)

where

$\begin{aligned} \delta_x^2\Gamma_{m, n}^{k+1} = &\left(\Gamma^{k+1}_{m+1, n}-2 \Gamma^{k+1}_{m, n}+ \Gamma^{k+1}_{m-1, n}\right), \\ \delta_y^2\Phi_{m, n}^{k+1} = &\left(\Phi^{k+1}_{m, n+1}-2 \Phi^{k+1}_{m, n}+ \Phi^{k+1}_{m, n-1}\right), \end{aligned}$

subject to the boundary conditions:

$\begin{equation} \begin{array}{l} \Psi_{x}(a, \, y, \, t) = \Psi_{x}(b, \, y, \, t) = 0, \, \, \, \, \, \, \, \, \forall y\in [c, \, d], \\ \\ \Psi_{y}(x, \, c, \, t) = \Psi_{y}(x, \, d, \, t) = 0, \, \, \, \, \, \forall x\in [a, \, b]. \end{array} \end{equation}$

(5.2)

Equation (5.2) permits us to use fictitious points in estimating the space derivatives at the domain's endpoints. The initial conditions are generated by

$\begin{equation} \begin{aligned} \Psi_2(x, y, 0) = -\dfrac{ (\alpha+\beta)}{2(\gamma+\lambda)} \left( 3 \tanh ^2\left(\dfrac{1}{2} (x+y+x0\right)-1 \right) , \end{aligned} \end{equation}$

(5.3)

where $\alpha, \, \beta, \, \gamma$ and $\lambda$ are user-defined parameters. In all the numerical results shown in this section, the parameter values are fixed as $\alpha = 0.50, \, \beta = 0.6, \, \gamma = -1.50, \, \lambda = 1.80, \, x0 = -45.0$ , $y = 0\to1,$ $x = 0\to60$ and $t = 0\to 25.$ The above system is solved by using an ODE solver in FORTRAN called the DDASPK solver ^[48]. This solver used a backward differentiation formula. Since I do not have the initial conditions for the space derivatives, I approximate the Jacobian matrix of the linearized system by using LU-Factorization. The obtained numerical results are acceptable. This can be observed from the Figures 8 and 9.

Figure 8. Time change for the numerical results while holding

$y = 0.5$ and

$M_x = 1600$ at

$t = 0:5:25.$ The wave at

$t = 25$ illustrates that the numerical and the analytic solutions are quite identical.

DownLoad: Full-Size Img PowerPoint

Figure 9. The convergence histories of the scheme with the fixation of both

$y = 0.5$ and

$M_x = 1600$ at

$t = 5$ .

DownLoad: Full-Size Img PowerPoint

6. Stability of the numerical scheme

The von Neumann analysis is used to examine the stability of the scheme (5.1). The von Neumann analysis is occasionally called Fourier analysis and is utilized exclusively when the scheme is linear. Hence, I suppose that the linear version is given by

$\begin{equation} \begin{aligned} \Psi_t& +\alpha \Gamma_{xxy}+\beta \Phi_{xyy}+s_0\, \Gamma_y+s_1\, \Psi_y+s_2\, \Phi_x+s_3\, \Psi_x = 0, \\ \Gamma_y& = \Psi_x, \\ \Psi_y& = \Phi_x, \\ \end{aligned} \end{equation}$

(6.1)

where $s_0 = \gamma \Phi$ , $s_1 = \gamma \Psi$ , $s_2 = \lambda \Gamma$ , $s_3 = \lambda \Psi$ are constants. Since $\Gamma_y = \Psi_x,$ and $\Psi_y = \Phi_x,$ the first equation of (6.1) is given by

$\begin{equation} \Psi_t +\alpha \Psi_{xxx}+\beta \Psi_{yyy}+ s_0\, \Psi_y +s_1\, \Psi_y+s_2\, \Psi_x+s_3\, \Psi_x = 0, \end{equation}$

(6.2)

where $\alpha, \, \beta, \, \gamma, \, \lambda, \, s_0, \, s_1, \, s_2, \, s_3, \, l_4$ are constants. I set directly

$\begin{equation} \begin{aligned} \Psi_{m, n}^k = &\mu^k\, \exp(\iota\, \pi\, \xi_0 \, n\, \Delta_x)\exp(\iota\, \pi\, \xi_1 \, m\, \Delta_y), \quad\\ \end{aligned} \end{equation}$

(6.3)

and also I can have

$\begin{aligned} \Psi_{m, n}^{k+1} = &\mu \, \Psi_{m, n}^k, \quad \Psi_{m+1, n}^{k} = \exp(\iota\, \pi\, \xi_0 \, \Delta_x) \, \Psi_{m, n}^k, \quad \Psi_{m, n+1}^{k} = \exp(\iota\, \pi\, \xi_1 \, \Delta_y) \, \Psi_{m, n}^k, \\ & \qquad\qquad\, \, \, \Psi_{m-1, n}^{k} = \exp(-\iota\, \pi\, \xi_0 \, \Delta_x) \, \Psi_{m, n}^k, \quad \Psi_{m, n-1}^{k} = \exp(-\iota\, \pi\, \xi_1 \, \Delta_y) \, \Psi_{m, n}^k , \\ m = &1, \, 2, \dots, \, N_x-1, \quad \quad n = 1, 2, \dots, \, N_y-1. \end{aligned}$

Substituting (6.3) into (6.2) and doing some operations, I have

$\begin{array}{l} 1 = \mu\left( 1-\iota\Delta_t\left( \frac{\sin(\xi_0\pi \Delta_x)\, }{\Delta_x} \left(\frac{4\, \alpha}{\Delta_x^2}\, \sin^2(\tfrac{\xi_0\pi\, \Delta_x}{2})-s_2-s_3\right)+\frac{ \sin(\xi_1\pi \Delta_y)\, }{\Delta_y} \left(\frac{4\, \beta}{\Delta_y^2}\, \sin^2(\tfrac{\xi_1\pi\, \Delta_y}{2})-s_0-s_1\right)\right)\right). \end{array}$

Hence,

$\begin{equation} \begin{array}{l} \mu = \dfrac{1}{ 1-a\iota}, \end{array} \end{equation}$

(6.4)

where

$a = \Delta_t\left( \frac{\sin(\xi_0\pi \Delta_x)\, }{\Delta_x} \left(\frac{4\, \alpha}{\Delta_x^2}\, \sin^2(\tfrac{\xi_0\pi\, \Delta_x}{2})-s_2-s_3\right)+\frac{ \sin(\xi_1\pi \Delta_y)\, }{\Delta_y} \left(\frac{4\, \beta}{\Delta_y^2}\, \sin^2(\tfrac{\xi_1\pi\, \Delta_y}{2})-s_0-s_1\right)\right).$

Thus,

$\begin{equation} \begin{array}{l} |\mu|^2 = \frac{1}{1+a^2}\leq 1.\\ \, \end{array} \end{equation}$

(6.5)

The stability condition of the von Neumann analysis is fulfilled. Consequently, from Eq (6.5), the scheme is unconditionally stable.

7. Error analysis

To examine the accuracy of the numerical scheme (5.1), I study the truncation error utilizing Taylor expansions. Suppose that the error is

$\begin{equation} \begin{array}{l} e^{k+1}_{m, n} = \Psi_{m, n}^{k+1}-\Psi(x_m, y_n, t_{k+1}), \end{array} \end{equation}$

(7.1)

where $\Psi(x_m, y_n, t_{k+1})$ and $\Psi_{m, n}^{k+1}$ are the analytic solution and an approximate solution, respectively. Substituting (7.1) into (5.1) gives

$\begin{equation*} \label{} \begin{aligned} \frac{e^{k+1}_{j, k}-e^{k}_{j, k}}{\Delta_t} = & T^{k+1}_{m, n}- \left( \alpha \frac{1}{2\Delta_x^3 } \delta_x^2\left( e^{k+1}_{m+1, n}-e^{k+1}_{m-1, n}\right) +\beta \frac{1}{2\Delta_y^3 } \delta_y^2\left( e^{k+1}_{m, n+1}-e^{k+1}_{m, n-1}\right)\right.\\ &\left.+ \frac{ s_2+s_3}{2\Delta_x} \left( e^{k+1}_{m+1, n}-e^{k+1}_{m-1, n}\right)+ \frac{ s_0+s_1}{2\Delta_y} \left( e^{k+1}_{m, n+1}-e^{k+1}_{m, n-1}\right)\right), \end{aligned} \end{equation*}$

where

$T^{k+1}_{m, n} = \frac{\alpha}{2\Delta_x^3 } \delta_x^2\left( \Psi(x_{m+1}, y_n, t_{k+1})-\Psi(x_{m-1}, y_n, t_{k+1})\right) \\ +\frac{\beta }{2\Delta_y^3 } \delta_y^2\left( \Psi(x_{m}, y_{n+1}, t_{k+1})-\Psi(x_{m}, y_{n-1}, t_{k+1})\right)\\ + \frac{ s_2+s_3}{2\Delta_x} \left( \Psi(x_{m+1}, y_n, t_{k+1})-\Psi(x_{m-1}, y_n, t_{k+1})\right)+\\ \frac{ s_0+s_1}{2\Delta_y} \left( \Psi(x_{m}, y_{n+1}, t_{k+1})-\Psi(x_{m}, y_{n-1}, t_{k+1})\right).$

Hence,

$\begin{equation*} \label{} \begin{aligned} T^{k+1}_{m, n}\leq& \frac{\Delta_t}{2}\frac{\partial^2\Psi(x_{m}, y_{n}, \xi_{k+1})}{\partial t^2} -\frac{\Delta_x^2}{2}\frac{\partial^5\Psi(\zeta_{m}, y_{n}, t_{k+1})}{\partial x^5} -\frac{\Delta_y^2}{2}\frac{\partial^5\Psi(x_{m}, \eta_{n}, t_{k+1})}{\partial x^5}\\ & -\frac{\Delta_y^2}{6}\frac{\partial^3\Psi(x_{m}, \eta_{n}, t_{k+1})}{\partial x^3}-\frac{\Delta_x^2}{6}\frac{\partial^3\Psi(\zeta_{m}, y_{n}, t_{k+1})}{\partial x^3}. \end{aligned} \end{equation*}$

Accordingly, the truncation error of the numerical scheme is

$\begin{equation*} \label{} \begin{array}{l} T^{k+1}_{m, n} = O(\Delta_t, \Delta_x^2, \Delta_y^2). \end{array} \end{equation*}$

8. Results and discussion

I have prosperously employed several analytical methods to extract the traveling wave solutions to the two-dimensional Novikov-Veselov system, confirming the solutions with numerical results obtained using the numerical scheme (5.1). The major highlights of the results are shown in Table 3 and –, which allow immediate comparison of the analytic solutions with the numerical results. Through these, I can notice that the solutions are identical to a large extent, and the error approaches zero whenever the value of $\Delta_x, \Delta_y\to 0$ . The numerical schemes are unconditionally stable for fixing the parameter values $\alpha = 0.50, \, \beta = 0.6, \, \gamma = -1.50, \, \lambda = 1.80, \, x0 = -45.0$ , $y = 0\to1,$ $x = 0\to60$ and $t = 0\to 25.$

Table 3. The relative error with

$L_2$ norm and CPU at

$t = 20.$ .

$\Delta_x$	The Relative Error	CPU
$0.6000 \qquad\qquad$	$5.600\times 10^{-3}\qquad$	$0.063\times10^{3}\, \text{m} \qquad$
$0.3000$	$2.100\times 10^{-3}$	$0.1524\times10^{3} \, s$
$0.1500$	$6.700\times 10^{-4}$	$0.3564\times10^{3}\, s$
$0.0750$	$2.100\times 10^{-4}$	$0.8892\times10^{3}\, s$
$0.0375$	$6.610\times 10^{-5}$	$1.7424\times10^{3}\, s$
$0.0187$	$2.310\times 10^{-5}$	$4.0230\times10^{3}\, s$

| Show Table

DownLoad: CSV

Figure 10. The convergence histories measured utilizing the relative error with

$l_2$ norm as a function of

$\Delta_x$ (see Table 3). Here, I picked a certain value of the variable

$y = 0.5$ at

$t = 20$ and

$x = 0\to60$ .

DownLoad: Full-Size Img PowerPoint

presents the time evolution of the analytic solutions $(a)$ $\Psi_1$ and $(b)$ $\Psi_2$ with $t = 0, 10, 20$ . The parameter values are $x0 = -20, \ \alpha = 0.50, \ \beta = 0.6, \ \gamma = -1.5,$ and $\lambda = 1.$ presents the wave behavior by changing a certain parameter value and fixing the values of the others. presents the behavior of $\Psi_1$ when I change the values of (a) $\alpha$ or $\beta$ and (b) $\gamma$ or $\lambda$ . In it can also be seen that the value of $\alpha$ or $\beta$ affects the direction and amplitude of the wave, such that a negative value always makes the wave negative, its amplitude decreases when $\alpha, \, \beta \to 0$ , and its amplitude increases when $\alpha, \, \beta \to \infty.$ In the value of $\gamma$ or $\lambda$ affects the direction and amplitude of the wave, such that a negative value always makes the wave negative, and its amplitude decreases when the value of $\gamma$ or $\lambda$ increases. In present the wave behavior of $\Psi_2.$ shows the time evolution of the analytic solutions. shows $\Psi_7$ with $t = 0:2:6$ . The parameter values are $\delta_2 = -1$ , $\delta_4 = 1,$ $\epsilon = -1,$ $\alpha = 0.50,$ $\beta = 0.6,$ $\gamma = -1.5,$ $\lambda = 1.8$ and $x0 = -10.$ shows $\Psi_8$ with $t = 0:2:8$ . The parameter values are $\delta_2 = 1$ , $\delta_4 = -1,$ $\epsilon = -1,$ $\alpha = 0.50,$ $\beta = 0.6,$ $\gamma = -1.5,$ $\lambda = 1.8$ and $x0 = -10.$ – present the $3D$ time evolution of the analytic solutions $\Psi_2$ (left) and the numerical solutions (right) obtained employing the scheme 5.1 with $t = 5, 15, 25$ , $M_x = 1600,$ $N_y = 100,$ $x = 0\to60$ and $y = 0\to1.$ These figures provide us with an adequate answer that the numerical and analytic solutions are quite identical. Barman et al. ^[42] accepted several traveling wave solutions for (1.1) as hyperbolic functions. The authors employed other parameters to develop new forms for the accepted solution. They proposed that Eq (1.1) describes tidal and tsunami waves, electromagnetic waves in transmission cables and magneto-sound and ion waves in plasma. In comparison, I have found numerous solutions also as hyperbolic functions. Furthermore, I obtained the numerical solutions to enhance the assurance that the solutions presented here are correct and accurate.

9. Conclusions

I have successfully utilized the generalized algebraic and modified F-expansion methods to acquire the soliton solutions for the two-dimensional Novikov-Veselov system, verifying these solutions with numerical results obtained by employing the numerical scheme (5.1). The major highlights of the results shown in Figures 8– and , which allow immediate comparison of the analytic solutions with the numerical results. Through these, I can notice that the solutions are identical to a large extent, and the error approaches zero whenever the value of $\Delta_x, \Delta_y\to 0$ . The numerical schemes are unconditionally stable for fixing the parameter values $\alpha = 0.50, \, \beta = 0.6, \, \gamma = -1.50, \, \lambda = 1.80,$ $x0 = -45.0$ , $y = 0\to1,$ $x = 0\to60$ and $t = 0\to 25.$ The Jacobi elliptic functions have effectively deteriorated to hyperbolic functions. The applied numerical schemes have provided reliable numerical solutions when using a small value of $\Delta_x, \Delta_y\to0$ .

Ultimately, I can deduce that the methods used are valuable and applicable to extract soliton solutions for other nonlinear evolutionary systems found in chemistry, engineering, physics and other sciences.

Conflict of interest

The author declares that he has no potential conflict of interest in this article.

References

[1]	A. Abbaspour, S. Mokhtari, A. Sargolzaei, K. K. Yen, A survey on active fault-tolerant control systems, Electronics, 9 (2021), 1513. doi: 10.3390/electronics9091513 doi: 10.3390/electronics9091513
[2]	N. Sultana, N. Chilamkurti, W. Peng, R. Alhadad, Survey on SDN based network intrusion detection system using machine learning approaches, Peer-to-Peer Netw. Appl., 12 (2019), 493–501. doi: 10.1007/s12083-017-0630-0 doi: 10.1007/s12083-017-0630-0
[3]	S. Mokhtari, A. Abbaspour, K. K. Yen, A. Sargolzaei, A machine learning approach for anomaly detection in industrial control systems based on measurement data, Electronics, 10 (2021), 407. doi: 10.3390/electronics10040407 doi: 10.3390/electronics10040407
[4]	V. Chandola, A. Banerjee, V. Kumar, Anomaly detection: A survey, ACM comput. surv. (CSUR), 41 (2009), 1–58. doi: 10.1145/1541880.1541882 doi: 10.1145/1541880.1541882
[5]	K. Paridari, N. O'Mahony, A. Mady, R. Chabukswar, M. Boubekeur, H. Sandberg, A framework for attack-resilient industrial control systems: Attack detection and controller reconfiguration, P. IEEE, 106 (2017), 113–128. doi: 10.1109/JPROC.2017.2725482 doi: 10.1109/JPROC.2017.2725482
[6]	J. Goh, S. Adepu, M. Tan, Z. S. Lee, Anomaly detection in cyber physical systems using recurrent neural networks, 2017 IEEE 18th International Symposium on High Assurance Systems Engineering (HASE), (2017), 140–145. doi: 10.1109/HASE.2017.36 doi: 10.1109/HASE.2017.36
[7]	A. Javaid, Q. Niyaz, W. Sun, M. Alam, A deep learning approach for network intrusion detection system, Eai Endorsed Transactions on Security and Safety, 3 (2016). doi: 10.4108/eai.3-12-2015.2262516 doi: 10.4108/eai.3-12-2015.2262516
[8]	H. Choi, M. Kim, G. Lee, W. Kim, Unsupervised learning approach for network intrusion detection system using autoencoders, The Journal of Supercomputing, 75 (2019), 5597–5621. doi: 10.1007/s11227-019-02805-w doi: 10.1007/s11227-019-02805-w
[9]	M. Masson, Darktrace OT threat finds: Detecting an advanced ICS attack targeting an international airport, Aug., 2007. Available from: http://shorturl.at/nuJ19.
[10]	Y. Liu, Y. Mu, K. Chen, Y. Li, J. Guo, Daily activity feature selection in smart homes based on pearson correlation coefficient, Neural Process. Lett., (2020), 1–17. doi: 10.1007/s11063-019-10185-8 doi: 10.1007/s11063-019-10185-8
[11]	S. Bornelöv, J. Komorowski, Selection of significant features using Monte Carlo feature selection, Challenges in Computational Statistics and Data Mining, (2016), 25–38. doi: 10.1007/978-3-319-18781-5_2 doi: 10.1007/978-3-319-18781-5_2
[12]	Q. Guo, W. Wu, D. L. Massart, C. Boucon, S. De Jong, Feature selection in principal component analysis of analytical data, Chemometr. Intell. Lab., 61 (2002), 123–132. doi: 10.1016/S0169-7439(01)00203-9 doi: 10.1016/S0169-7439(01)00203-9
[13]	J. G. Dy, C. E. Brodley, Feature selection for unsupervised learning, J. Mach. Learn. Res., 5 (2004), 845–889. doi: 10.5555/1005332.1016787 doi: 10.5555/1005332.1016787
[14]	Y. Li, B. Lu, Z. Wu, Hierarchical fuzzy filter method for unsupervised feature selection, J. Intell. Fuzzy Syst., 18 (2007), 157–169. doi: 10.5555/1368376.1368381 doi: 10.5555/1368376.1368381
[15]	Z table website provides all required information for using Z-test. Aug., 2021. Available from: https://www.ztable.net.
[16]	F. T. Liu, K. M. Ting, Z. Zhou, Isolation forest, 2008 eighth ieee international conference on data mining, (2008), 413–422. doi: 10.1109/ICDM.2008.17 doi: 10.1109/ICDM.2008.17
[17]	M. Wu, C. Jermaine, Outlier detection by sampling with accuracy guarantees, Proceedings of the 12th ACM SIGKDD international conference on Knowledge discovery and data mining, (2006), 767–772. doi: 10.1145/1150402.1150501 doi: 10.1021/ie302042c
[18]	H. Ma, Y. Hu, H. Shi, Fault detection and identification based on the neighborhood standardized local outlier factor method, Ind. Eng. Chem. Res., 52 (2013), 2389–2402. doi: 10.1021/ie302042c doi: 10.1021/ie302042c
[19]	V. Kotu, B. Deshpande, Chapter 13 - Anomaly Detection, Data Science (Second Edition), (2019), 447–465. doi: 10.1016/B978-0-12-814761-0.00013-7 doi: 10.1016/B978-0-12-814761-0.00013-7
[20]	M. Ashrafuzzaman, S. Das, A. A. Jillepalli, Y. Chakhchoukh, F. T. Sheldon, Elliptic Envelope Based Detection of Stealthy False Data Injection Attacks in Smart Grid Control Systems, 2020 IEEE Symposium Series on Computational Intelligence (SSCI), (2020), 1131–1137. doi: 10.1109/SSCI47803.2020.9308523 doi: 10.1109/SSCI47803.2020.9308523
[21]	P. J. Rousseeuw, K. V. Driessen, A fast algorithm for the minimum covariance determinant estimator, Technometrics, 41 (1999), 212–223. doi: 10.1080/00401706.1999.10485670 doi: 10.1080/00401706.1999.10485670
[22]	P. C. Mahalanobis, On the generalized distance in statistics, National Institute of Science of India, 1936. doi: 10.1007/s13171-019-00164-5 doi: 10.1007/s13171-019-00164-5
[23]	M. Sokolova, G. Lapalme, A systematic analysis of performance measures for classification tasks, Inform. process. manag., 45 (2009), 427–437. doi: 10.1016/j.ipm.2009.03.002 doi: 10.1016/j.ipm.2009.03.002
[24]	N. V. Chawla, K. W. Bowyer, L. O. Hall, W. P. Kegelmeyer, SMOTE: synthetic minority over-sampling technique, J. artif. intell. res., 16 (2002), 321–357. doi: 10.5555/1622407.1622416 doi: 10.5555/1622407.1622416
[25]	H. Shin, W. Lee, J. Yun, H. Kim, HAI 1.0: HIL-based Augmented ICS Security Dataset, 13th $USENIX$ Workshop on Cyber Security Experimentation and Test ( $CSET$ 20), 2020. doi: 10.5555/3485754.3485755 doi: 10.5555/3485754.3485755
[26]	S. Choi, HIL-based Augmented ICS (HAI) Security Dataset, The Affiliated Institute of ETRI, South Korea, 2020. Available from: https://github.com/icsdataset/hai.
[27]	H. Shin, W. Lee, J. Yun, H. Kim, Implementation of programmable CPS testbed for anomaly detection, 12th USENIX Workshop on Cyber Security Experimentation and Test (CSET 19), 2019. doi: 10.5555/3359012.3359014 doi: 10.5555/3359012.3359014
[28]	W. Hwang, J. Yun, J. Kim, H. Kim, Time-series aware precision and recall for anomaly detection: considering variety of detection result and addressing ambiguous labeling, Proceedings of the 28th ACM International Conference on Information and Knowledge Management, (2019), 2241–2244. doi: 10.1145/3357384.3358118 doi: 10.1145/3357384.3358118

Reader Comments

Your name:*

Email:*
© 2021 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)

通讯作者: 陈斌, bchen63@163.com

1.
沈阳化工大学材料科学与工程学院沈阳 110142

Applied Computing and Intelligence

Metrics

Article views(4352) PDF downloads(134) Cited by(5)

Preview PDF

Download XML

Export Citation

Article outline

Show full outline

Applied Computing and Intelligence

Measurement data intrusion detection in industrial control systems based on unsupervised learning

Related Papers:

Abstract

1. Introduction

2. Summary of proposed methods

2.1. The modified F-expansion method

2.2. The generalized algebraic method

3. Methodology

3.1. Application of modified F-expansion method

3.2. Application of the generalized algebraic method

4. Numerical solution

5. Numerical scheme on a fixed mesh

6. Stability of the numerical scheme

7. Error analysis

8. Results and discussion

9. Conclusions

Conflict of interest

References

Reader Comments

通讯作者: 陈斌, bchen63@163.com

Metrics

Figures and Tables

Other Articles By Authors

Related pages

Tools

Export File

Citation

Format

Content

Catalog

Abstract

1. Introduction

2. Summary of proposed methods

2.1. The modified F-expansion method

2.2. The generalized algebraic method

3. Methodology

3.1. Application of modified F-expansion method

3.2. Application of the generalized algebraic method

4. Numerical solution

5. Numerical scheme on a fixed mesh

6. Stability of the numerical scheme

7. Error analysis

8. Results and discussion

9. Conclusions

Conflict of interest

References