Bifurcation analysis of a composite cantilever beam via 1:3 internal resonance

In this paper, we study a multiple scales perturbation and numerical solution for vibrations analysis and control of a system which simulates the vibrations of a nonlinear composite beam model. System of second order differential equations with nonlinearity due to quadratic and cubic terms, excited by parametric and external excitations, are presented. The controller is implemented to control one frequency at primary and parametric resonance where damage in the mechanical system is probable. Active control is applied to the system. The multiple scales perturbation (MSP) method is implemented to obtain an approximate analytical solution. The stability analysis of the system is obtained by frequency response (FR). Bifurcation analysis is conducted using various control parameters such as natural frequency (ω1), detuning parameter (σ1), feedback signal gain (β), control signal gain (γ), and other parameters. The dynamic behavior of the system is predicted within various ranges of bifurcation parameters. All of the stable steady state (point attractor), stable periodic attractors, unstable steady state, and unstable periodic attractors are determined efficiently using bifurcation analysis. The controller’s influence on system behavior is examined numerically. To validate our results, the approximate analytical solution using the MSP method is compared with the numerical solution using the Runge-Kutta (RK) method of order four.


Introduction
An active control system is defined necessarily in terms of that a mass of external power or energy is required. Fanson and Caughey [1] were the first to introduce the Positive Position Feedback (PPF) technique. Vibrations can be abolished via a movement of feedback signals conducted by the mentioned technique. Active constrained layer damping has been effectively implemented as an efficient method to control the vibration of various flexible mechanical structures [2][3][4][5][6][7][8][9]. Eissa and Sayed [10][11][12] examined the active controller's effect on both spring and simple pendulum at the primary resonance using negative velocity feedback. El-Bassiouny [13], Eissa et al. [14][15][16][17], and Jaensch [18] confirmed how active control is functional in vibration attenuation at resonance for different vibration modes. They approved that the active control method has several advantages with respect to the passive one. Sadek et al. [19] employed piezoelectric-patch-actuators as active control to control the vibration of simply supported flexible plates. An optimal control law was deduced utilizing the maximum principle theory to calculate the voltage of the actuators. Wang et al. [20] used an inverse feed-forward controller method for directing polymer actuator displacement control. Dong et al. [21] investigated the active vibration control method to control piezoelectric smart structures using a system identification technique. Kapuria and Yasin [22] studied the active vibration repression of metal laminated plates through piezoelectric fiber-reinforced composite materials. Warminski et al. [23] studied the experimental and theoretical investigations of vibrations of an autoparametric system composed of two beams with rectangular cross-sections. The experimental response of the system, tuned at 1:4 internal resonance case, is performed for random and harmonic excitations. Warminski et al. [24] investigated the numerical and experimental solutions for various active controllers which applied to a nonlinear mechanical system. El-Ganaini et al. [25] considered the PPF controller to put down the vibration amplitude of a nonlinear dynamic model at primary resonance with 1 : 1 internal resonance. Hamed and Amer [26] studied an active controller for damping out the nonlinear vibration composite beam with parametric excitation force in the case of 1 : 2 internal resonances. Sayed and Kamel [27] examined the effect of a linear controller to control the vibration of each of the saturation control of a linear absorber and the vibrating system. Sayed and Kamel [28] applied an active control method with on 1 : 2 and 1 : 3 internal resonance to control the vibrations of a nonlinear vibrating system. Large deformation analysis for a cantilever beam with a variable bending stiffness under both static and dynamic loads is investigated by the direct integration scheme [29]. A simple boundary feedback control moment is proposed to stabilize a nonhomogeneous flexible beam with a tip mass [30]. They obtained the exponential stability, spectrum-determined growth condition, and optimal decay rate. Chentouf [31] considered the stabilization problem of a variant of the SCOLE model. He shows that the system uniformly stabilized by the proposed feedback law as soon as the boundary control force and control moment were presented in the feedback. Chentouf [32] studied a model that consists of a nonhomogeneous flexible beam clamped at its left end to a rigid disk and free at the right end, where another rigid body is attached. They treated different physical situations and provided accordingly appropriate feedback control laws. Bağdatlı et al. [33] investigated nonlinear transverse vibrations of a tensioned Euler-Bernoulli beam resting on multiple supports. A perturbation technique was applied to the equations of motion to obtain approximate analytical solutions. He considered 3 : 1 internal resonance. Hegazy [34] studied the dynamic behavior and chaotic motion of a string-beam coupled system subjected to parametric excitation. He considered and examined the case of 3 : 1 internal resonance between the modes of the beam and the string, in the presence of subharmonic resonance for the beam. Cemil Tunç [35,36] investigated the asymptotic stability and boundedness of all solutions of certain differential equations of fourth order.
In this paper, the solutions of nonlinear differential equations that represent composite beam vibrations are obtained. We control the vibration of the system via 1 : 3 active control techniques. We applied the MSP technique to solve those nonlinear equations. We extract the corresponding FR equations and present graphically at different system parameters. We apply the RK algorithm to confirm the obtained curves numerically.
We applied Lyapunov's first method to investigate the stability of the controlled system. To validate our results, we compare approximate solutions using the MSP method and numerical solutions using the RK method of order four. We predict the dynamic behavior of the system at various ranges of bifurcation parameters. We use bifurcation analysis to determine the efficiently stable steady state (point attractor), stable periodic attractor, unstable steady state, and unstable periodic attractors.

Equations of motion
The modified nonlinear differential equations describe the oscillations of the nonlinear composite beam model [23,24,26] are given by: with initial conditions uð0Þ ¼ 0:02;uð0Þ ¼ 0; νð0Þ ¼ 0:2;νð0Þ ¼ 0; where u denotes the composite beam system displacement and ν denotes the controller displacement, and the derivatives with respect to the time t are denoted by dots. μ 1 and μ 2 are the parameters of the viscous damping coefficients, ω 1 is the natural frequency accompanied by the composite beam, and ω 2 is the natural frequencies associated with controller modes, Ω 1 and Ω 2 are the excitation frequencies, f 1 and f 2 are the excitation forces amplitudes, and γ and β are the control and feedback signal gains. δ, α 1 are the main system nonlinear stiffness coefficient which may be due to material of the beam nonhomogeneity,α 2 nonlinear stiffness coefficient of the controller. The nonlinear parameters, linear viscous damping, and excitation forces are as follows: where the parameter ε has a low value denoted as perturbation parameter and 0 < ε ≪ 1. Substituting Eq. (3) into Eqs. (1) and (2), we can obtain: The parametersα 1 ;α 2 ;μ 1 ;μ 2 ;δ;β;γ;f 1 , andf 2 are of the order 1. This means that all these parameters appear when comparing the coefficients of the parameter ε.
Theorem 2 [37]. (Uniqueness theorem) Suppose that f and ∂ f ∂x n are continuous function in R (defined in the existence theorem). Hence, both the f and ∂ f ∂x n are bounded in R, i.e., (a) |f(t, x n )| ≤ K and (b) ∂ f ∂x n ≤ L ∀ðt; x n Þ∈R. Then, the IVP (*) has at most one solution Combining with existence theorem, the IVP (*) has a unique solution x n = x n (t) defined in the interval |t − 0| ≤ α. By using Theorems (1) and (2), both the functions f(t, x n ) and its partial derivative are defined and continuous at all pointsfð

Mathematical analysis
A first-order approximate solution of Eqs. (4) and (5) is conducted using the method of MSP [38,39].
where T n = ε n t is the fast time scales for n = 0 and slow time scales for n = 1. In terms of T 0 , T 1 , the time derivatives are converted to the D operator according to: where D n ¼ ∂ ∂T n : After substituting Eqs. (6) and (7) and (8) into Eqs. (4) and (5) and comparing the coefficients of similar powers of the parameter ε in both sides, we obtain the following: Order (ε 0 ): Order (ε): The solutions of Eqs. (9) and (10) are represented in the complex form: where A 1 and A 2 are complex function with T 1 as a complex argument, which assigned by removing of secular terms and small-divisor terms from the 1st order of approximation. Therefore, we investigate the case where Ω 1 ≅ ω 1 , Ω 2 ≅ 2ω 1 and ω 1 ≅ 3ω 2 . To examine quantitatively the closeness of the resonances, the detuning parameters σ 1 , σ 2 , and σ 3 are represented by Ω 1 = ω 1 + εσ 1 , Ω 2 = 2ω 1 + εσ 2 and ω 1 = 3ω 2 + εσ 3 . Substituting Eqs. (13) and (14) into Eqs. (11) and (12), we get: We get the solvability conditions of the first-order expansion by eliminating the secular terms means the term that has some singularity (an unbounded term arising in time-dependent perturbation theory): Substituting Eqs. (17) and (18) into Eqs. (19) and (20), we obtain To analyze Eqs. (21) and (22), we can express A 1 and A 2 in polar form as where a 1 and a 2 are the steady state amplitudes, and γ s (s = 1, 2) are phases of the motion. Substituting Eq. (23) into Eqs. (21) and (22) and separating the real and imaginary parts, we get the following: where Then, it follows from Eq. (28) thaṫ Substituting Eq. (29) into Eqs. (25) and (27), we obtaiṅ From Eqs. (24), (26), (30), and (31), the autonomous system of equations is:

Steady state solution
In this work, the steady state motions are examined as follows: Then, it follows from Eq. (29) thaṫ Then, the steady state solutions of Eqs. (32)-(35) are as follows: In addition to a trivial one, two possible cases which will be described as follows exist.
Case (1): We consider a 1 ≠ 0 and a 2 = 0 (system without control); the FR equation is described as follows: Case (2): It is considered a 1 ≠ 0 and a 2 ≠ 0 (system with control); the FR equations are described as follows:

Stability of the system
To derive the stability criteria, we assume that where a 10 , a 20 , and θ n0 satisfy Eqs.  41) and keeping only the linear terms in a 11 , a 21 , and θ n1 , we obtain: The system stability is analyzed by assessing the eigenvalues of the Jacobian matrix of the right-hand sides of the system of Eqs. (46)-(49). The equilibrium solution is considered stable when the real parts of the Jacobian matrix eigenvalues are negative. While the relevant solution is unstable if the corresponding real part of any of the eigenvalues is positive, the fixed points stability is investigated according to the following cases: Case (1): Uncontrolled system, let a 21 = θ 31 = 0 in Eqs. (46)-(49); then, the above system is written in the matrix form as follows a 11 : where Then, the characteristic equation is , and the obtained eigenvalues are as follows: According to the Routh-Hurwitz criterion, the necessary and sufficient conditions of the roots of Eq. (51) have negative real parts, such that Then, the fixed points are stable; otherwise, they are unstable. Case (2): Controlled system when a 11 ≠ 0, θ 11 ≠ 0, a 21 ≠ 0, θ 31 ≠ 0 in Eqs. (46)-(49); then, we can write the above system in the matrix form as shown below: The eigenvalue of the above system is as follows: i.e. λ 4 + R 1 λ 3 + R 2 λ 2 + R 3 λ + R 4 = 0, (54) where R 1 , R 2 , R 3 , and R 4 are functions of L s (s = 1, 2, …, 16) which are given above. Based on the Routh-Hurwitz criterion, the necessary and sufficient conditions of the roots of Eq. (54) have negative real parts, such that In the following figures, the dashed lines refer to unstable solutions, while the solid lines refer to stable solutions.

Numerical results
Simulation results are illustrated in graphical forms such as the steady state amplitude of the composite beam system versus time history and detuning parameter σ 1 or the response of the composite beam system and controller at 1 : 3 internal resonance. The different parameters of the composite beam system shown in Fig. 1 are μ 1 = 0.001, f 1 = 0.01, f 2 = 0.05, α 1 = 0.002, ω 1 = 4, δ = 0.0001. Figure 1 declares that the steady state amplitude of the composite beam system is about 90 times that of the external excitation amplitude f 1 .This is conducted when the controller is idle at the simultaneous primary and principal parametric resonance. Figure 2 shows that the time history for both composite beam system and controller amplitudes at Ω 1 ≅ ω 1 , Ω 2 ≅ 2ω 1 at internal resonance ω 1 ≅ 3ω 2 . From Fig. 2, it is observed that the energy is transmitted from the composite beam to the controller. To measure the effectiveness of the controller, we define E a as flows (E a = uncontrolled system amplitude/controlled system amplitude). Then, the effectiveness of the controller E a is about 35. Figures 3, 4, 5, 6, and 7 indicate numerical results of the frequency response curves of the controlled system. Figure 3 illustrates that the response of the controlled system against σ 1 . We note that the minimum response of the beam occurs when σ 1 = 0, which approves that the controller is able to eliminate the simultaneous Fig. 1 System behavior without a controller at resonance Ω 1 ≅ ω 1 , Ω 2 ≅ 2ω 1 for initial conditions uð0Þ ¼ :02;uð0Þ ¼ 0 primary and parametric resonance vibration effectively at the existence of the 1:3 internal resonance. Figure 4 shows that the influence of f 1 on the system amplitude response against the detuning parameter σ 1 . From these figures, it is noticed that the increase in excitations force f 1 leads to an increase in the region of instability system. Figure 5 shows the effects of changing the control signal γ on the FR curves of both the controller and the beam system. Also, for decreasing γ, the effective frequency bandwidth of the saturation controller induces lowering the vibration impact of the composite beam. On the other hand, the controller's excessive risk is decreased when the control signal is raised. Figure 6 demonstrates the effect of feedback signal β variation on the FR curves. Form this figure, we can see that decreasing the feedback signal obtained a better vibration lowering influence on the composite beam. Figure 7 indicates that the impacts of the natural frequency on the amplitude response curves against σ 1 . Also, it is observed that for increasing the natural frequency, the effective frequency bandwidth of the controller and the better vibration damping influence of the composite beam.

Bifurcation analysis
Dynamic bifurcation analysis is performed using the XPPAUT 5 software package and AUTO-07p. Furthermore, MATLAB is used for investigating the dynamic behavior where the dynamic results are obtained at specific parameter values defined from AUTO 2007 [40][41][42][43]. The bifurcation analysis is performed for most of the controlling parameters such as parametric excitations parameters (f 2 ), nonlinear control parameter Fig. 2 System behavior with a controller at resonance Ω 1 ≅ ω 1 , Ω 2 ≅ 2ω 1 , ω 1 ≅ 3ω 2 for initial conditions uð0Þ ¼ 0:02;uð0Þ ¼ 0; νð0Þ ¼ 0:2;νð0Þ ¼ 0:  Table 1. There is only Hopf bifurcation (HB) point at f 2 = 0.06839, where a 1 = 0.02473 and a 2 = 0.5691. In the region of (f 2 ≤ 0.06839), the system is dominated by a stable steady state, where a 1 is around 0.025 and a 2 is around 0.57. In the region of (0.06839 ≤ f 2 ≤ 0.09), the system is characterized by stable periodic attractor, where a 1 oscillated in the range of [0.02, 0.28] while a 2 oscillated in the range of [0.4, 1.32]; there is the unstable steady state in the same region (0.06839 ≤ f 2 ≤ 0.09) and unstable periodic attractor in (0.085 ≤ f 2 ≤ 0.09). Effect of the nonlinear control parameter (α 2 ) Figure 9 a and b indicate the effects of the nonlinear control parameter (α 2 ) as a bifurcation parameter on a 1 and a 2 at γ = 30 based on the two-parameter continuation technique [44], and the rest of the parameter's values are shown in Table 1. There are two Hopf bifurcation points: the first Hopf HB 1 appears at the two points (α 2 , a 1 ) = (0.478, 0.2006) and (α 2 , a 2 ) = (0.478, 0.09285), while second Hopf HB 2 appears at the two points (α 2 , a 1 ) = (0.1008, 0.1447) and (α 2 , a 2 ) = (0.1008, 0.09622). The system behavior can be characterized as follow: Region (α 2 ≤ 0.1008): The bifurcation diagrams obtained in Fig. 9 a,  Region (0.1008 ≤ α 2 ≤ 0.478): The system is characterized by the unstable steady state through this region in the range (0.1008 ≤ α 2 ≤ 0.18); the system is dominated by the stable periodic attractor. Figure 10 a-c indicate the dynamics of the system at α 2 = 0.12 at the corresponding initial conditions. Figure 10 Fig. 10b. The phase plane is shown in Fig. 10c confirms the oscillatory attractor which is the only attractor at α 2 = 0.12 as shown through the regular limit cycles of a 1 and a 2 .
Region (0.478 ≤ α 2 ): The system is dominated by a stable steady state. There is also an unstable periodic attractor. Figure 11 a and b show time traces of a 1 and a 2 at α 2 = 0.8 at the corresponding initial conditions where a 1 settles down at a 1 = 0.27 while a 2 = Fig. 6 Effects of feedback signal gain β 0.0923. Phase plane (Fig. 11c) confirms the time traces in Fig. 11a, b where the system approached a point attractor.
Effect of the detuning parameter (σ 1 ) Figure 12 a and b show the bifurcation analysis using the detuning parameter (σ 1 ) as the bifurcation parameter at γ = 50 based on the two-parameter continuation technique [44], and the rest of the parameter's values are indicated in Table 1. Two Hopf bifurcation points are available: HB 1 appears at σ 1 = 5.95 × 10 −5 , a 1 = 0.1722, and a 2 = 0.08094, while HB 2 is at σ 1 = − 0.001926, a 1 = 0.1875,and a 2 = 0.08292 as shown in Fig. 12 a, b. Region (5.95 × 10 −5 ≤ σ 1 ): The system is dominated by the point attractor where there is only a stable steady state.
Region (−0.001926 ≤ σ 1 ≤ 5.95 × 10 −5 ): In this region, the system is dominated by the only periodic attractor. There is also an unstable steady state. The maximum frequency for a 1 and a 2 appears close to HB 1 at σ 1 = 0, where a 1 oscillates in the range of [0.02, 0.45] as shown in Fig. 12a, and a 2 oscillates in the range of [0.045, 0.135] as shown in Fig. 12b. However, this frequency decreases continuously as σ 1 decreases until reaching HB 2 as shown in Fig. 12a, b.
Region (σ 1 ≤ − 0.001926): In this region, the system is dominated by the point attractor where a stable steady state is available, where a 1 changes in the range [0.172, 0.19] while a 2 changes in the range [0.082, 0.09] as shown in Fig. 12a, b. Effect of the feedback signal gain (β) Figure 13 shows the effect of feedback signal β as the main bifurcation parameter where other parameter values are shown in Table 1. There is only one HB at β = 0.1776 and a 1 = 0.1544, and a 2 = 0.5116 as shown in Fig. 13a, b, Fig. 13b.
Region (β ≤ 0.133): The system has no stable behavior where there are unstable steady state and unstable periodic behavior as shown in Fig. 13a, b.
Effect of the control signal gain (γ) Figure 14 shows the effect of control signal γ as the main bifurcation parameter where the rest of the parameter's values are indicated in Table 1. There is only one HB at γ = 47.56 and a 1 = 0.1691, and a 2 = 0.08244 as shown in Fig. 14a, b, respectively. The behavior of the system can be described as follows:   Fig. 14b.

Effect of the natural frequency (ω 1 )
The effect of ω 1 as the main bifurcation parameter is indicated in Fig. 15 where the other parameter values are taken from Table 1. There is only HB point at ω 1 = 3.283, a 1 = 0.01664, and a 2 = 0.5767 as shown in Fig. 15a, b. The system behavior due to changing ω 1 can be described as follows: Region (3.283 ≤ ω 1 ): In this region, the system is controlled by point attractor, where a 1 changes in the range of [0.017, 0.02], and a 2 changes in the range of [0.57, 0.6] as shown in Fig. 15a, b, respectively.    Figure 15b shows that the a 2 has the maximum frequency range of [0.5, 1.2] at ω 1 = 2.85.
Region (ω 1 ≤ 2.85): In this region, the system has unstable oscillations and unstable steady state. There is neither point nor periodic attractors as shown in Fig. 15a, b. Figure 16 shows time trace of the controlled system Eqs. (1) and (2) and the approximate modulated amplitudes of Eqs. (34)-(37) which obtained from perturbation analysis of the controlled system. From the comparison, we concluded that the predictions obtained from approximated analytical solutions using the MSP method agree well with the numerical solutions found by the RK. Figures 17 and 18 show a good agreement between the numerical solution of the original system (1) and (2) using the RK method and approximate solution using the MSP method. Also, the results obtained in Fig. 17 agreement with the results obtained in Fig. 5. The saturation phenomenon occurs for increasing feedback signal gain β as shown in Fig. 18.

Comparison with previous work
In [26], the active vibration controller 1:2 for suppressing the vibration of the nonlinear composite beam subjected to parametric excitation force only has been studied. In this paper, an active vibration controller 1:3 for suppressing the vibration of the nonlinear   composite beam subjected to parametric and external excitation forces has been studied. Bifurcation analysis is conducted using various control parameters such as natural frequency (ω 1 ), detuning parameter (σ 1 ), feedback signal gain (β), control signal gain (γ), and other parameters. The dynamic behavior of the system is predicted within various ranges of bifurcation parameters. All of the stable steady state (point attractor), stable periodic attractor, unstable steady state, and unstable periodic attractors were determined efficiently using bifurcation analysis. The controller's influence on system behavior is examined numerically.

Conclusions
The main objective of the work is to control the vibration of a nonlinear composite cantilever beam with external and parametric excitation forces. We applied the MSP technique to solve those nonlinear equations. We applied Lyapunov's first method to investigate the stability of the controlled system. The corresponding FR equations have been extracted and presented graphically at different system parameters. The obtained curves were confirmed numerically applying the RK algorithm. Bifurcation analysis was conducted using many bifurcation parameters such as f 2 , α 2 , σ 1 , β, γ, and ω 1 . The nonlinear controller has been studied at Ω 1 ≅ ω 1 , Ω 2 ≅ 2ω 1 , ω 1 ≅ 3ω 2 . To validate our results, the approximated analytical solution using the MSP perturbation method is compared with the numerical solution using the RK method of order four. From a previous comprehensive study, the following points are concluded: 1. The composite beam amplitude in the case of no controller at the simultaneous principal and primary parametric resonance Ω 1 ≅ ω 1 , Ω 2 ≅ 2ω 1 is multiplied to about 30 times of the parametric excitation amplitude f 2 .  2. The controller 1 : 3 is efficient for lowering the high steady state amplitude vibration of the composite beam. 3. Increasing the feedback signal β obtained the better of the vibration damping effect of the composite beam. 4. For increasing the linear damping μ 1 and natural frequency ω 1 , the efficient frequency bandwidth of the controller and the better of vibration damping the composite beam influence. 5. For decreasing the control signal γ and excitation forces f 1 and f 2 , the effective frequency bandwidth of the saturation controller can be improved and increase the vibration reduction action of the composite beam. 6. The overload controller risk decreases as the control signal γ increases. 7. The saturation phenomenon occurs for increasing feedback signal gain β. 8. The excitation force amplitude is increased as the steady state amplitude increased and lost its stability. 9. The composite beam amplitude is inversely proportional to the damping coefficient μ 1 and natural frequency ω 1 . 10. The validation of our results has been done, where the approximate analytical solutions are compatible with the numerical solutions. 11. The dynamic system behavior was determined at different regions of the bifurcation parameters where all stable steady state, stable periodic, unstable steady state, and periodic attractors were evaluated.