Parametric Study of Nonlinear Beam Vibration Resting on Linear Elastic Foundation

Nonlinear beam resting on linear elastic foundation and subjected to harmonic excitation is investigated. The beam is simply supported at both ends. Both linear and nonlinear analyses are carried out. Hamilton’s principle is utilized in deriv ing the governing equations. Well known forced duffing oscillator equation is obtained. The equation is analyzed numerically using Runk-Kutta technique. Three main parameters are investigated: the damping coefficient, the natural frequency, and the coefficient of the nonlinearity. Stability reg ions for first mode analyses are unveiled. Comparison between the linear and the nonlinear model is presented. It is shown that first mode shape the natural frequency could be approximated as square root of the sum of squares of both natural frequency of the beam and the foundation. The stretching potential energy is proved to be responsible for generating the cubic nonlinearity in the system.


Introduction
Beams on elastic foundations received great attention of researches due to its wide applicat ions in engineering. The model of the beam resting on a nonlinear support represents a large class of mechanical systems, such as, vibrating mach ines on elastic foundations, pipelines transversally supported, disc brake pad, shafts supported on ball, roller, or journal bearings, network of beams in the construction of floor systems for ships, buildings, bridges, submerged floating tunnels, railroad tracks etc. The elastic foundation for the beam part is supplied by the resilience of the adjoining portions of a continuous elastic structure. Hetenyi [1] and Timoshenko [2] p resented an analytical solution for beams on elastic supports using classical differential equation approach, and considering several loading and boundary conditions. It is well known in engineering that a beam supported by discrete elastic supports spaced at equal intervals acts analogously to a beam on an elastic foundation and that the appropriateness of that analogy depends on the flexural rig idity of the beam as well as the stiffness and spacing of the supports. Ellington investigated conditions under which a beam on discrete elastic supports could be treated as equivalent to a beam on elastic foundation [3].
Beams resting on elastic foundations have been studied extensively over the years due to the wide application of this system in engineering. This system according to the literature can be d ivided at least into three categories.
The first category is "linear beam on linear elastic founda tion". Examp le of this type can be found in references [4]- [1 5]. The applications in this category include but not limited to Eu ler -Bernoulli beam, Timoshenko beam, Winkler foundation, Pasternak foundation, tensionless foundation, single parameter or t wo parameter foundation, static loading, harmonic loading and moving loading.
For example the Winkler foundation model represents the simp lest form of these types of beams. In this model the foundation is treated essentially as an array of closely spaced but non-interacting springs, each having a spring stiffness that equals the foundation modulus divided by the spacing between springs. The Pasternak foundation is extension of Winkler foundation by introducing shear interaction between springs. It is assumed that the top ends of the springs are connected to an incompressible layer that resists only transverse shear deformation. Tensionless foundation is similar to Winkler foundation but the springs do not carry any tension load.
The second category is "linear beam on nonlinear elastic foundation [16]- [20]. In this category the foundation is considered to have nonlinear stiffness. Also this type includes different boundary and loading conditions according to the engineering application.
The third category is nonlinear beam on linear elastic foundation [21]- [33]. Usually the beam nonlinearity means large deflections. Most of the studies related to this category have analyzed the system either using boundary element method or boundary integral equation method. Similar to the above two categories, there is wide variety of boundary and loading conditions being applied to such system according to the application.
Nonlinear beam subjected to harmonic d istributed load resting on linear elastic foundation is investigated in this research. The study is carried out in the view of the linearized model of the system. Well known duffing equation is obtained using Hamilton's princip le. Th ree main parameters are investigated: the damping coefficient, the natural frequency, and the coefficient o f the nonlinear term. The effect of these parameters on the system stability is unveiled. Up to the author's knowledge, this work is not published in the literature.

Problem Statement
Nonlinear beam resting on elastic foundation that is shown in Fig. 1 is subjected to the following conditions: 1. The beam material properties are linear. 2. The damping () and stiffness (k f ) of the foundation are linear.
3. The beam is slender and pris matic. 4. The beam is simp ly supported (pin-pin ends) 5. The load applied is harmonic and distributed over the length of the beam.

Kinetic Energ y
The rotary inertia of the beam will be neglected since the beam is slender.
Where : material density, A: beam cross sectional area, L: beam length, w=w(x,t): beam transverse displacement (in y-direction).

Potenti al Energy
The potential energy due to bending can be calculated as the following: Where P: amp litude of excitation and  e : excitation frequency

Deri vati on Of Governing Equation
The lagrangian is defined as the following: By applying Hamilton's princip le Integrating the first and the second term by parts with respect to x the result is the following equation: Since w is arbitrary, the fo llo wing can be concluded fro m the above equation: The governing equation comes fro m setting the expression within the brackets in Equation (6) equal to zero. Upon carrying out the indicated differentiations, the governing can be rewritten as It is obvious that Equation (7) is the duffing oscillator equation. This equation is going to be recasted into a mo re familiar form in the next section. The boundary and init ial conditions can be obtained from the remain ing terms in Equation (6).
The boundary conditions at x=0 and x=L are Either EI w ′′ is zero or w ′ is prescribed (8a) Either EI w ′′ is zero or w is prescribed Boundary conditions Equation (8a) correspond to end mo ments and slopes respectively. In Equation (8b), w corresponds to end displacement, and in Equation (8c) the first condition corresponds to pre-stretching. For the pinned ends, the boundary conditions are: These boundary conditions must be satisfied by the mode shapes of the system. This fact will be used in the following sections as the criteria fo r selecting the form of the mode shape equation.
Finally the in itial conditions for t = t 1 and t = t 2 are

Either 2 ( ) 2 is zero or is prescribed
In this case, it will be assumed that the system starts from rest i.e. the initial d isplacement and velocity is zero.

Discretization and linearization
The follo wing exp ression is used for w(x, t) in order to discretize the problem , = =1 For simp licity the limits of the above summat ion, the subscript of w, and the time dependence of w will be implied in the equations that follow. It is evident from Equation (9) that the pinned ends boundary condition Equation (8a) are satisfied since transverse displacements at 0 and L are zero, and the end slopes are free (imp lying zero bending mo ments at the ends). Equation (9) represents series summation of N modes each has time dependent amp litude response, ( ) with spatial sine function. Substituting Equation (9) Lagrangian's equation for each mode can be written as the following: Substituting Equation (10) in Equation (11) Where Writing Equation (13) for a single mode and inserting the linear damping term g ives, Where µ is the damping coefficient This makes it clear that the above equation represent unforced damped duffing oscillator. Recasting Equation (14) into the follo wing:

4 2
In order to linearize the system for the first mode (n=1) the system is converted into first order ordinary differential equations by the follo wing substitution = → = = → = Applying this to Equation (15) = Fro m the above equations it is obvious that (0, 0) is the only critical point for the system. So the equivalent linear system is obtained by expanding the above equation using Taylor series about (0, 0), so the remain ing linear terms are = , = − 2μ − 2 The corresponding Jacobi matrix is The general solution of the linearized unforced system is Applying the init ial conditions (0) = 0 , (0) = 0 the constants of integration are going to be as the following:

Simulation of the nonlinear system
It is obvious that the strength of the nonlinearity is inversely proportional to the square of the radius of gyration of the beam. This indicates that the nonlinearity remains weak as long as the beam is relat ively slender as assumed in this study. Finally, the frequency equation can be simplified to 2 The apparent natural frequency of the system  is the square root of the sum of the squares of the natural frequencies of the beam and the elastic foundation.
The nonlinear second order ordinary d ifferential equation is converted into a system of first order o rdinary d ifferential equations. This is suitable fo r nu merical study using Runge-Kutta Techniques.

Results and Discussion
The results for simply supported beam on elastic foundation are presented in Figures 2 through 6, Table 1 and Table 2. Figures 2 through 6 present the results of the linearized model whereas Table 1, 2 and 3 represent the results of the nonlinear model. Figure 2 present the linear system behavior for µ>0 and ∆≥0. It is obvious that the system is stable attractive node. Table 1 represents sample phase diagrams of the studied ranges. Figure 3 shows stable centre for the linear system of µ=0 and ∆≥0. Figure 4 presents stable spiral for the linear system behavior of µ>0 and ∆<0. In Figures 2, 3 and 4 the system is stable however in Figures 5 and 6 it is not stable. The behavior in both Figures 5 and 6 is unstable node. Table 2 shows the time response for those cases that are presented in Table 1, whereas the corresponding frequency spectra (Fast Fourier Transform-FFT) for those cases presented in both Tables 1 and 2. The phase portraits and the time response are collected after long period of t ime to be sure that the system has passed the transient range. The duffing Equation (16) is converted into system o f first order ord inary d ifferential equation as shown in Equation (17) and solved using MATLAB package by utilizing the Runga-Kutta ODE (Ordinary Differential Equation) solver. The equation which represents the system under investigation is of cubic nonlinearity with harmonic excitation.
The sample results present the effect of damping when the system has weak, mediu m and strong nonlinearity for excitation frequencies below, at and above resonance. The whole study is considering weak nonlinearity that does not exceed α =0.1 and those levels of weak, med iu m and strong within that range. On ly the first mode is considered in this study. The parameters range covered in this investigation are for =0.0 through 0.1, α = 0.001 through 0.1 and natural frequency = 0.7 through 1.4. It can be seen fro m Table 1 that when there is no damp ing the system is tending toward chaos however when little damp ing is applied the system is tending towards limit cycle.
It is obvious that the damping and the nonlinearity are the most effective parameters in controlling the chaotic behavior of the system. As long as the radius of gyration for the beam under consideration is large i.e., the beam is mo re towards slender, the nonlinearity is going to be weak. This means that the contribution of the stretching energy to the behavior of the system is going to be low. The damping system dissipates the oscillat ing energy and provides a control over the system behavior. For the linear system, as long as damping coefficient is positive, the transient response is going to decrease exponentially and the forcing excitation response is bounded even at resonance. For undamped linear system the forcing excitation response is not bounded and the response is increasing with time. For the nonlinear system, the response is tending toward chaos as it can be seen in Tab le 1 for =0.0. However when the damping increases the system is transferring fro m chaos to limit cycle. It is obvious in Table 3 that all cases of =0.0 have Double period, Trip le period, or, Chaotic response. The response of those double period cases shows amp litude modulation whereas those cases where their frequencies are very close show beating phenomenon. For the resonance case of no damping i.e., the excitation frequency equals the natural frequency, the linear system has increasing amp litude response whereas the nonlinear system is tending towards chaos with bounded amplitude. It is obvious that the nonlinear cases (within the range of investigation) are tending towards chaos as long as there is no damping in the system.

Conclusions
The behavior of nonlinear beam on elastic foundation is unveiled. It is found that the system is stable and controllable as long as the damping coefficient is non zero and positive. As the nonlinearity increases more damping is required to prevent it fro m moving towards chaos. For first mode shape the natural frequency could be calculated as square root of the sum of squares of both natural frequency of the beam and the foundation. The strength of the nonlinearity is inversely proportional to the square of the radius of gyration, i.e. as long as the beam more towards slender the nonlinearity is weaker. The stretching potential energy is responsible for generating the cubic nonlinearity in the system.