Dynamics and Complexity Analysis of Fractional-Order Chaotic Systemswith Line EquilibriumBased onAdomianDecomposition

Shaanxi Engineering Research Center of Controllable Neutron Source, School of Science, Xijing University, Xi’an 710123, China Collaborative Innovation Center of Memristive Computing Application, Qilu Institute of Technology, Jinan 250200, Shandong, China Shandong Engineering Research Center of China-Germany Smart Factory Applied, Jinan 250200, Shandong, China School of Civil Engineering, Inner Mongolia University of Technology, Inner Mongolia, Huhehot 010051, China


Introduction
ree hundred years ago, fractional-order calculation was proposed as a classical mathematical problem. Fractional calculation has no practical background and has not been applied in engineering. erefore, researchers and scientists are not interested in fractional research. Recently, it have been found that many engineering and physical systems, electronic systems, etc. exhibit their own fractional-order characteristics, and the fractional order can more accurately reflect natural phenomena, such as material memory and damping characteristics [1][2][3][4]. In recent years, with the deepening and perfection of chaotic theory research, fractional-order chaotic systems have become a hot topic in the field. In particular, the complexity of fractional-order chaotic systems is not only related to the parameters of the system itself, but also related to the fractional order of the system. Many scholars have proposed several fractionalorder chaotic systems based on integer-order chaotic systems: fractional-order Lorenz system [5], factional-order Lorenz hyperchaotic system [6], four-wing fractional-order chaotic system [7], and so on [8,9].
In the numerical calculation of fractional chaotic systems, namely, discretization of fractional chaotic systems, many scholars have made some achievements based on frequency-domain method (FDM) [10], Adomian decomposition method (ADM) [5-7, 9, 11], and Adams-Bashforth-Moulton (ABM) algorithm [12,13]. Among them, FDM uses high-dimensional system close to fractional-order system through Laplace change, but the error is relatively large [14]. ABM is the most commonly used method, but its operation speed is slow. ADM scheme, on the other hand, has more accuracy and requires less computational resources compared with ABM algorithm [5][6][7]9]. Reference [15] shows that fractional-order chaotic systems show more complex chaotic behaviors than integerorder chaotic systems, which can be seen from complexity, Lyapunov index, bifurcation diagram, and other aspects. When the system order is smaller, the system is more complicated [9,16].
On the other hand, hot research on chaotic systems mainly focuses on chaos control [17][18][19][20], Lyapunov exponent calculation and analysis [21,22], and doubling and growth of attractors [23,24], and chaotic systems with hidden attractors are also hot research topics, because such systems are extremely prone to multistable phenomenon, which is a common phenomenon in nature. Hidden attractors originate from several types of dynamical systems; the dynamical systems with one stable equilibrium [25], a line or plane equilibrium [26,27], or no equilibrium [28] can all have hidden attractors. Hidden attractors are even found in some special chaotic systems, which have both unstable and stable equilibria. However, the fractional-order chaotic system with hidden attractor, which can be easily obtained by combining the hidden attractor with fractional-order systems, is rarely studied [29]. e system has high complexity, so it plays an important role in the field of communication encryption.
In this paper, we propose a new fractional-order chaotic system with line equilibrium. Discretization of the proposed system is performed using Adomian decomposition scheme. e rest of the paper is organized as follows. In Section 2, the equilibrium point and dynamic analysis of the integer-order chaotic equilibrium point system are presented. In Section 3, the fractional-order chaotic system is numerically calculated and simulated by Adomian decomposition method. In Section 4, the bifurcation diagram, phase diagrams, and Lyapunov exponents spectrum are employed to analyze the dynamics of the system. In Section 5, complexity of fractional-order chaotic system is analyzed. Finally, the obtained results are summarized in Section 6.

Analysis of Equilibrium Points of Systems.
Reference [29] proposes a class of T chaotic systems defined by A 3D integer-order chaotic system is defined by where, a, b, and c are system (2) parameters and x, y, and z are three state variables of system (2). When parameters a � 5, b � 2, and c � 34, the system has two equilibrium points: , (c/a)), and E 3 (0, h, 0). e eigenvalues at E 1 and E 2 are λ 11 � − 6.0739, λ 12 � − 0.4631 + 21.3673, and λ 13 � 0; E 3 (0, h, 0) is the line equilibrium point; and h is any real number. e Jacobian matrix linearized under equilibrium E 3 is as follows: e characteristic equation of matrix is e eigenvalue at the line equilibrium point E 3 is as follows: according to Routh-Hurwitz criterion, the equilibrium E 3 is stable.
according to Routh-Hurwitz criterion, the equilibrium E 3 is unstable.

Case
3. If h 2 � ab, then according to Routh-Hurwitz criterion, we obtain that the equilibrium E 3 is critical stable.
When a � 5, b � 2, and c � 34, using ODE45 function in MATLAB software to simulate (2), the chaotic attractor is shown in Figure 1. e orbits marked by the blue and red colors emerge form initial (1, 2, 3) and (1, 2, − 1), and the red attractor is periodic. e blue attractors are shown in chaotic states.

Dynamics
Analysis of the Chaotic System. Fix a � 5, b � 2, and c � 34; initial values of state variables [x 0 , y 0 , z 0 ] � [1, 2, z 0 ]; let the z 0 vary from − 15 to 15 with step size of 0.01. Bifurcation diagram and Lyapunov exponents spectrum of system (2) are shown in Figure 2. As can be seen from the bifurcation diagram of the system in Figure 2 Figure 3. In Figure 3 the chaotic region is green (labeled "C") and the periodic region is red (labeled "P"). We can see from Figure 3 that the system state is mainly affected by the initial condition z 0 .

Solution of Fractional Chaotic System Based on Adomian Decomposition
System (2) can generate complex chaotic attractors with line equilibrium point, and it has three typical parameter sets. Change the above system equations to the following fractional-order form: e initial condition is According to domains decomposition method [30] and fractional calculus properties, the following properties are obtained: , , where x 1 , y 1 , z 1 are the values of systems (3). h s � t − t 0 is iteration step size. Γ(·) is Gamma function. e corresponding variables are assigned to the corresponding value; let Complexity 5 us, the solution of systems (5) is defined as , .
When a � 5, b � 2, and c � 34, system (5) has stable equilibrium with line (0, h, 0). e analytical solution of the system can be obtained according to (14). Using MATLAB to simulate (14), the chaotic attractor is shown in Figure 4. e orbits marked by the blue and red colors emerge form initial (1, 2, 3) and (1, 2, − 1), and the red attractor is periodic. e blue attractors are shown in chaotic states.

Dynamical Analysis
Based on the ADM solution of the fractional-order chaotic system, we focus on the analysis of the influence of parameters on the system. In this paper, the system bifurcation diagram and Lyapunov exponent are used to analyze the dynamic system. e bifurcation diagram is obtained by using the maximum value of state variables. In this section, the influence of three parameters on the system is studied.  Figure 5. As can be seen from the bifurcation diagram of the system in Figure 5(a), when z 0 ∈ [− 15, − 0.7], the system is in a periodic state and stable at (0, 0, 0). When z 0 ∈ (− 0.7, 15], the system is in a chaos state. When z 0 ∈ [− 15, − 0.7], the largest Lyapunov exponents is equal to 0. When z 0 ∈ (− 0.7, 15], the largest Lyapunov exponents is positive, as shown in Figure 5(b).

Dynamics with q Varying.
Fix a � 5, b � 2, and c � 34; let the derivative order q vary from 0.6 to 1 with step size of 0.002 and the initial values of state variables [x 0 , y 0 , z 0 ] � [1, 2, 3]. It can be seen from Figure 6(a) that, under the variation of derivative order q, the system chaos and periodic variation are repeated and crossed and are not a process from period to chaos directly. At the same time, Figure 6 shows that the system is chaotic over the interval q ∈ [0.678, 0.824) ∩ [0.876, 1]. However, q ∈ [0.876, 1] also has a periodic window. It can be seen clearly that the maximum Lyapunov exponent decreases with the increase of derivative order q ∈ [0.678, 0.824]. In this case, the minimum order for chaos is q � 0.678. e maximum Lyapunov exponent value at this point (q � 0.678) is the largest. With the change of parameter q, the bifurcation diagram of the system is consistent with the Lyapunov exponent spectrum of the system.
For further research, influence of coexistence of attractors in fractional derivative q system. In this paper, the phase diagram is used to introduce. It is found that a � 5, b � 2, and c � 34. e integer-order system (q � 1) is different from the fractional-order system (q � 0.98). e attractors of the integerorder system are almost the same, while the attractors of the fractional-order system coexist. As shown in Figure 7, the red initial values are [1,2,3]. e blue initial values are [1,2,20].

Dynamics with a and b
Varying. e fractional chaotic system (5) has three system parameters besides the fractional derivative q. In this paper, the dynamic properties of the system parameters with a and b varying are analyzed. Firstly, bifurcation diagrams of the fractional-order system with a ∈ [3,10] is investigated as shown in Figure 8 Secondly, bifurcation diagrams of the fractional-order system (5) with b ∈ [0, 6] is investigated as shown in Figure 8 QR decomposition method [6,15] is effective in calculating the LEs, and its computational process is shown by Here, qr( ) represents QR decomposition function, m is the maximum iteration number, and J is the Jacobian matrix of 6 Complexity the given discrete formula as presented in (14). LCEs are calculated by Here, k � 1, 2, n (system dimension); n is the maximum number of iterations; and h is the iteration time step. According to the above basic formula, make n � 4, n � 7000, and h � 0.01, and use MATLAB to draw the Lyapunov exponent under parameter changes.

Completeness Calculation.
By calculating the energy distribution in the Fourier transform domain and combining it with the Shannon entropy, spectral entropy is obtained [31]. C0 complexity [32] is mainly to decompose the time series into regular and irregular parts and to test the proportion of irregular parts in the sequence. In practical applications, the chaos diagram of SE and C0 complexity can provide a basis to select parameter better. Sample entropy (SampEn) [16] measures the complexity of time series by measuring the probability of generating new patterns in signals. e greater the probability of generation, the greater the value of the complexity of the sequence.
(v) S5: similarly, change m to m + 1 and repeat step one with step three; get A m (r). (vi) S6: theoretically, the SampEn complexity can be calculated by

Complexity
Analysis. Fix a � 5, b � 2, c � 34; let q change from 0.6 to 1; step size is 0.002; and the initial value of state variables [x 0 , y 0 , z 0 ] � [1,2,3]. e complexity of the system is shown in Figure 11(a). We can see that the complexity decreases with the increase of q, and the change of complexity is consistent with the change of Lyapunov exponent in Figure 6(b). Fixed b � 2, q � 0.8; initial conditional is [x 0 , y 0 , z 0 ] � [1, 2, 3]; and let a vary. e complexity of the system is shown in Figure 11(b). e change of complexity is consistent with the change of Lyapunov exponent in Figure 10(a).
Fix a � 2, q � 0.8; initial conditional is [x 0 , y 0 , z 0 ] � [1, 2, 3]; and let b vary. e complexity of the system is shown in    Complexity Figure 11(c). e change of complexity is consistent with the change of Lyapunov exponent in Figure 10(b). Complexity is consistent with Lyapunov exponent, but Figure 11 also shows that sample entropy complexity is small; SE complexity is not obvious when some bifurcation points change; C 0 complexity is accurate; and C 0 complexity is faster than Lyapunov exponent, consumes less resources, and provides a good idea for analysis of chaotic systems.
Fix a � 5, b � 2, and c � 34; initial values of state variables [x 0 , y 0 , z 0 ] � [1, 2, z0]; let the z(0) vary from 0 to 25 with step size of 0.2 and q vary from 0.6 to 1 with step size of 0.004. e C0 complexity in the z(0)-q plane is shown in Figure 12. It can be seen that, in the fractional-order region, the system complexity under the change of z(0) is more complex than that of the integer order. e complexity of the system is calculated by the time sequence of a certain state variable. Compared with Lyapunov exponent calculation, the calculation of complexity is faster and saves resources, but there are also some errors in complexity, especially when SE complexity is in a periodic state.

Conclusion
In this paper, the accurate approximate solution for the fractional-order system with line equilibrium is obtained based on the Adomian decomposition method. Dynamical behaviors of the systems are analyzed using the chaotic attractor, bifurcation diagram, Lyapunov exponent spectrum, and largest Lyapunov exponent diagram. Chaotic range and periodic windows are determined. Both the system parameter and the fractional order can be taken as bifurcation parameters, which shows that the fractionalorder system has more complex and abundant dynamics than its integral-order counterpart. Besides, integer-order systems do not have the phenomenon of attractors coexistence, while fractional-order systems have it, so fractionalorder systems have better effects when applied to communication security systems.

Data Availability
e data used to support the findings of this study are available from the corresponding author upon request.

Conflicts of Interest
e authors declare that they have no conflicts of interest.