9. Nonlinear and Parametric Vibrations

Introduction and Context

Most introductory vibration models assume that restoring force and damping are proportional to displacement and velocity. Such linearization is valuable, but it is not universally valid. Large deformation, material nonlinearity, clearance, friction, geometric effects, magnetic forces, fluid interactions, and amplitude-dependent damping can make the equation of motion nonlinear. In other systems, the parameters themselves vary with time, producing parametric vibration. These effects may generate phenomena absent from linear theory: multiple equilibrium positions, amplitude-dependent natural frequencies, jump phenomena, subharmonic response, limit cycles, bifurcations, internal resonance, and deterministic chaos.

A representative single-degree-of-freedom oscillator is

m ẍ + c(x,ẋ,t)ẋ + f(x,t) = F(x,ẋ,t),

where the restoring force need not be proportional to x and the damping need not be proportional to ẋ. A parametrically excited system may instead be written as

ẍ + 2ζω_n ẋ + ω_n²[1+h cos(Ωt)]x = 0.

The first equation is externally forced in a general sense; the second can vibrate even without an applied force because its stiffness varies with time. Understanding the distinction is essential in rotating machinery, aerospace structures, MEMS resonators, vehicle suspensions, vibration absorbers, and energy-harvesting devices.

Learning Objectives

After studying this chapter, the reader should be able to:

1. distinguish linear, nonlinear, and parametrically excited vibration models;

2. formulate systems with nonlinear stiffness and damping;

3. interpret phase-plane trajectories, equilibrium stability, and limit cycles;

4. apply harmonic balance and perturbation methods to weakly nonlinear systems;

5. explain resonance bending, jump phenomena, bifurcation, and hysteresis;

6. analyze internal and parametric resonance conceptually and mathematically;

7. assess stability using linearization, averaged equations, and Floquet ideas;

8. identify the mechanisms and signatures of deterministic chaos;

9. select suitable analytical, numerical, and experimental tools for engineering decisions.

From Linear Oscillation to Nonlinear Dynamics

For a linear single-degree-of-freedom oscillator,

m ẍ+c ẋ+kx=F_0 cos Ωt,

frequency response is single-valued for specified parameters, superposition applies, and the natural frequency is independent of amplitude. Nonlinear systems violate one or more of these properties. A common model is the Duffing oscillator:

ẍ+2ζω_n ẋ+ω_n²x+αx³=f cos Ωt.

If α is positive, the system is hardening: its effective frequency increases with amplitude. If α is negative, it is softening: the effective frequency decreases with amplitude. The cubic term may arise from geometric stretching, nonlinear springs, or magnetic forces. A quadratic term, βx², destroys symmetry about the equilibrium and may generate a static shift and even harmonics.

Nonlinear damping is similarly diverse. Viscous damping is c ẋ, whereas Coulomb friction is approximately F_c sgn(ẋ), and quadratic fluid drag is c_q|ẋ|ẋ. Van der Pol damping is represented by

μ(x²−1)ẋ.

For small |x| this term supplies negative damping, while for large |x| it supplies positive damping. The result can be a self-excited stable oscillation, or limit cycle, without an external periodic force.

A useful energy description follows from

E=½ẋ²+V(x),

where V′(x)=f(x). For the undamped Duffing oscillator,

V(x)=½ω_n²x²+¼αx⁴.

The shape of V indicates whether the equilibrium is a single-well or double-well system. When the quadratic stiffness is negative and the quartic stiffness positive, the potential has two stable wells separated by an unstable equilibrium. Oscillations may remain in one well or cross between wells, depending on energy and forcing.

Phase Plane and Equilibrium Stability

Introducing y=ẋ converts a second-order autonomous equation into

ẋ=y,

ẏ=−g(x,y).

The phase plane is the x-y plane. A trajectory represents the state of the oscillator rather than displacement alone. Closed curves correspond to periodic motions; spirals indicate decay or growth; a fixed point represents equilibrium; and a separatrix may divide qualitatively different motions.

For an autonomous system ẋ=f(x), an equilibrium x_e satisfies f(x_e)=0. Linearizing about it gives

δẋ=J(x_e)δx,

where J is the Jacobian matrix. If all eigenvalues have negative real parts, the equilibrium is locally asymptotically stable. A positive real part indicates instability. Purely imaginary eigenvalues require higher-order analysis because linearization alone does not determine nonlinear stability.

In a conservative double-well oscillator, the central equilibrium is a saddle and the two outer equilibria are centers in the idealized undamped model. Damping changes centers into stable spirals or nodes. Under periodic forcing, the system is nonautonomous; the phase portrait is then often examined using a stroboscopic or Poincaré section, sampling the state once per forcing period.

Weakly Nonlinear Analysis

Harmonic Balance

Harmonic balance assumes an approximate periodic response and equates coefficients of selected harmonics. For the forced Duffing equation

ẍ+2ζω_nẋ+ω_n²x+αx³=f cos Ωt,

assume x≈a cos(Ωt−φ). Using

cos³θ=¼(3cos θ+cos 3θ),

and retaining the fundamental harmonic gives the amplitude relation

[(ω_n²−Ω²+¾αa²)²+(2ζω_nΩ)²]a²=f²,

and

tan φ=(2ζω_nΩ)/(ω_n²−Ω²+¾αa²).

This equation explains resonance bending. The term ¾αa² shifts the effective stiffness. For hardening behavior, the response curve bends toward higher frequency; for softening behavior, it bends toward lower frequency. In a sufficiently strong nonlinear response, one excitation frequency may correspond to three mathematical amplitudes. Usually the middle branch is unstable, while the upper and lower branches can be stable. Slow variation of frequency can then produce sudden transitions called jumps and a difference between upward and downward sweeps called hysteresis.

Harmonic balance is efficient but approximate. It may fail when higher harmonics are large, when motion is strongly nonperiodic, or near separatrix crossings. Including third and higher harmonics improves accuracy at greater algebraic cost.

Perturbation and Multiple Scales

Perturbation methods express the solution as a series in a small parameter, for example

x=x_0+εx_1+ε²x_2+….

A direct expansion may produce secular terms such as t sin t, which grow without bound and invalidate the approximation at long times. The method of multiple scales introduces fast and slow times, T_0=t and T_1=εt, allowing amplitude and phase to evolve slowly. Solvability conditions eliminate secular terms and produce modulation equations.

For weakly forced, weakly damped oscillators, the resulting amplitude and phase equations often have the form

da/dT=−ζa+forcing\ term,

dφ/dT=detuning+nonlinear\ frequency\ shift.

Their steady solutions reproduce the frequency-response relation while also revealing stability and slow transient behavior. Averaging is closely related: rapidly oscillating terms are averaged over one cycle, leaving equations for slowly varying amplitude and phase.

Bifurcation, Limit Cycles, and Stability

A bifurcation occurs when a qualitative property of a solution changes as a parameter varies. Important examples include saddle-node, pitchfork, Hopf, period-doubling, and torus bifurcations. In forced nonlinear vibration, saddle-node bifurcation commonly creates or destroys periodic response branches and produces jumps. A Hopf bifurcation occurs when an equilibrium loses stability and a periodic orbit emerges, although the detailed type depends on nonlinear coefficients.

A limit cycle is an isolated closed trajectory in phase space. Unlike a family of closed orbits in a conservative system, it attracts or repels neighboring trajectories. Self-excited oscillators, such as idealized Van der Pol systems, possess attracting limit cycles. Limit-cycle amplitude is determined by the balance between energy supplied at small amplitude and energy dissipated at large amplitude.

For periodic solutions, stability can be assessed by perturbing the solution and examining the perturbation after one period. The associated multipliers are Floquet multipliers. A periodic orbit is stable if all nontrivial multipliers lie inside the unit circle. A multiplier crossing −1 often signals period doubling; crossing +1 is associated with a saddle-node or related loss of periodic stability.

Internal Resonance

Internal resonance occurs when two or more natural frequencies satisfy an approximately integer relationship, such as ω_2≈2ω_1 or ω_2≈3ω_1. Unlike ordinary resonance, energy is exchanged between modes through nonlinear coupling. A simplified two-mode model is

q̈_1+ω_1²q_1+ε(aq_1q_2+bq_1³)=0,

q̈_2+ω_2²q_2+ε(cq_1²+dq_2³)=0.

Near a 2:1 relationship, the square of the first modal coordinate contains a component near the second modal frequency. Energy can therefore flow from one mode to another, causing amplitude saturation, modal beating, or sudden changes in response. Internal resonance is important in aircraft panels, rotor systems, nonlinear microbeams, and large flexible structures.

Designers should not rely only on isolated linear modal frequencies. Operating conditions, temperature, preload, and amplitude can shift frequencies into an internal-resonance relationship. Nonlinear modal analysis and time-domain simulation are useful for identifying these interactions.

Parametric Vibration

In direct forcing, energy enters through an additive force. In parametric excitation, a system parameter such as stiffness, length, or preload varies with time. The classical Mathieu equation is

ẍ+2ζω_nẋ+ω_n²[1+h cos(Ωt)]x=0.

The modulation can amplify motion when Ω is close to 2ω_n, because the parameter variation reinforces the oscillation twice during each cycle. This is the principal parametric resonance region. Instability regions, often called tongues, occur in the parameter plane of excitation frequency and modulation amplitude. Damping narrows or raises the threshold for instability.

Parametric excitation appears in pendulums with periodically varying support, rotating shafts with stiffness variation, planetary gears, vibrating strings, and parametrically driven MEMS devices. A system can become unstable even when the mean stiffness is positive and no direct force is applied.

A practical distinction is important: changing stiffness at the wrong phase removes energy, whereas changing it at the right phase adds energy. Thus parametric excitation is fundamentally a phase-sensitive energy-transfer mechanism.

Deterministic Chaos

Chaos is bounded, aperiodic motion generated by deterministic equations and characterized by sensitive dependence on initial conditions. It is not equivalent to random noise. A chaotic vibration may display a broadband-looking spectrum, a complicated Poincaré section, a positive largest Lyapunov exponent, and an attractor with fractal structure.

Routes to chaos include period doubling, intermittency, quasiperiodic breakdown, and crises. In a periodically forced oscillator, the Poincaré map is especially useful. A period-one motion appears as one point, period-two motion as two points, and a chaotic attractor as a structured cloud or fractal set.

Chaos can be harmful by producing unpredictable loads, fatigue, noise, and loss of precision. It can also be useful in secure communication, mixing, weak-signal detection, and broadband energy harvesting. Claims of chaos require care: a noisy measurement, inadequate sampling, or an unmodeled transient can be mistaken for deterministic chaos. At least several complementary indicators should be used.

Engineering Analysis Workflow

A sound nonlinear vibration investigation normally proceeds through model selection, nondimensionalization, equilibrium analysis, local linearization, approximate analysis, numerical continuation, and validation. Nondimensionalization identifies governing ratios such as damping, forcing, detuning, and nonlinear strength. Equilibrium and energy analysis reveal wells, barriers, and possible escape.

For weak nonlinearity, harmonic balance or multiple scales can provide design-level formulas. For strongly nonlinear or multimode systems, numerical integration, shooting methods, continuation, and Floquet analysis are more appropriate. Parameter sweeps should be performed in both directions to detect hysteresis. Initial conditions must be varied because coexisting attractors may exist.

Experimental identification requires controlled amplitude and frequency sweeps, phase measurement, and careful separation of sensor noise from nonlinear harmonics. Digital acquisition must avoid aliasing, and the excitation system must not introduce an unrecognized parametric variation. Reduced-order models should be compared with measured backbone curves, modal interactions, and transient responses.

Applied Case: Illustrative Nonlinear Vibration of a Flexible Robotic Joint

Background

Consider a lightweight robotic joint driven through an elastic transmission. The joint is modeled as a torsional oscillator with inertia J, viscous damping c, linear stiffness k, and cubic stiffness α. Motor torque is harmonic during a repetitive task:

Jθ̈+cθ̇+kθ+αθ³=T_0 cos Ωt.

The transmission stiffens at large angular displacement because of geometric constraints.

Situation and Analysis

A linear model predicts a resonance near the small-amplitude natural frequency. During testing, however, the measured resonance shifts upward as torque amplitude increases, and the frequency sweep shows a sudden drop in response during the upward sweep. The behavior suggests hardening stiffness and coexisting stable solutions.

After scaling the equation and applying first-harmonic balance, the predicted amplitude relation contains the effective term k−JΩ²+¾αa². Positive α bends the response curve toward higher frequency. Continuation of the periodic solutions identifies a folded response branch. Floquet analysis shows that the middle branch is unstable, explaining why the experiment jumps between the lower and upper branches.

Engineering Approach

The design team considers reducing excitation near the folded region, increasing damping, and modifying the transmission geometry to reduce α. Increasing damping suppresses the jump but reduces efficiency. Changing stiffness moves the nonlinear resonance but may lower positioning accuracy. A controller based only on a linear resonance frequency is rejected because it cannot predict amplitude-dependent detuning.

Outcome

A revised operating schedule avoids slow frequency sweeps through the unstable region, while moderate damping is added to reduce sensitivity. The final model is validated against upward and downward sweeps and transient tests from different initial conditions.

Learning Points

The case demonstrates that a frequency-response curve is not always single-valued, that stability must be assessed in addition to solving an amplitude equation, and that nonlinear parameters influence both performance and control strategy. It also shows why experimental sweep direction and initial conditions are essential parts of vibration testing.

Challenges, Limitations, and Professional Considerations

Nonlinear models are sensitive to parameter uncertainty. A cubic coefficient fitted over one amplitude range may be invalid at another. Coulomb friction is nonsmooth and may require special numerical treatment. Modal truncation can omit important internal-resonance pathways. Harmonic balance may conceal transient instability, while time integration can miss narrow instability regions if the parameter grid is too coarse.

Engineering decisions should distinguish mathematical possibility from physically reachable motion. A high-amplitude branch may exist but be inaccessible under the available excitation. Conversely, a small perturbation or impact may move the system across a separatrix into a different potential well.

Professional practice requires transparent reporting of assumptions, parameter sources, uncertainty, sampling rate, initial conditions, and numerical tolerances. Safety-critical designs should not rely on a single reduced-order simulation. Stability margins, worst-case parameter combinations, fatigue implications, and fail-safe operating limits should be documented. Automated optimization and machine-learning surrogates can accelerate parameter studies, but they do not replace physical validation or stability analysis.

Current Developments and Best Practices

Modern nonlinear vibration research increasingly combines experimental modal analysis, continuation software, digital twins, and data-driven identification. Physics-informed machine learning can assist in estimating nonlinear forces from measured data, while sparse regression may identify governing terms. These methods are most reliable when constrained by conservation laws, dimensional consistency, and independently measured responses.

For professional work, the following practices are particularly valuable:

• retain nonlinear terms that have a defensible physical origin;
• nondimensionalize before interpreting parameter effects;
• analyze equilibria and energy barriers before forced response;
• use both upward and downward frequency sweeps;
• test multiple initial conditions;
• verify periodic solutions with stability analysis;
• compare approximate solutions with time-domain simulation and experiment;
• report uncertainty and distinguish noise from deterministic irregularity.

Chapter Summary

Nonlinear vibration arises when stiffness, damping, geometry, friction, or forcing cannot be represented adequately by linear terms. Nonlinear stiffness produces amplitude-dependent frequency and may create hardening, softening, multiple solutions, jumps, and hysteresis. Nonlinear damping can generate self-excited limit cycles. Phase-plane methods reveal equilibria, separatrices, periodic motions, and attraction. Harmonic balance provides practical frequency-response equations, while multiple scales and averaging describe slow amplitude and phase evolution. Bifurcation and Floquet analysis explain changes in periodic-solution stability. Internal resonance transfers energy between modes whose frequencies are nearly commensurate. Parametric excitation varies a system parameter and can cause instability, especially near twice the natural frequency. Deterministic chaos is bounded but aperiodic motion requiring careful identification. Reliable engineering analysis combines physical modeling, approximate theory, numerical continuation, stability assessment, experiments, and uncertainty-aware decision-making.

Key Terms

Nonlinear stiffness: Restoring stiffness whose force is not proportional to displacement.

Nonlinear damping: Damping force whose dependence on velocity or state is nonproportional.

Duffing oscillator: An oscillator containing a cubic stiffness term.

Hardening: Increase of effective natural frequency with amplitude.

Softening: Decrease of effective natural frequency with amplitude.

Phase plane: State-space plot, commonly displacement against velocity.

Separatrix: Trajectory separating qualitatively different classes of motion.

Harmonic balance: Approximate method equating selected harmonic components.

Multiple scales: Perturbation method using fast and slow independent time variables.

Averaging: Replacement of rapidly oscillating equations by slow mean-flow equations.

Limit cycle: Isolated periodic orbit in state space.

Bifurcation: Qualitative change in solutions caused by parameter variation.

Hysteresis: Dependence of response on the direction or history of parameter variation.

Internal resonance: Nonlinear interaction between modes with nearly commensurate frequencies.

Parametric excitation: Excitation produced by time variation of a system parameter.

Mathieu equation: Canonical equation for a parametrically excited oscillator.

Floquet multiplier: Eigenvalue describing perturbation evolution over one period.

Poincaré section: Sampled representation of a dynamical system at fixed phase or time intervals.

Deterministic chaos: Aperiodic, sensitive motion generated by deterministic equations.

Backbone curve: Amplitude–frequency relation of a weakly damped nonlinear oscillator.

Review Questions

Short Answer Questions

1. What distinguishes direct forcing from parametric excitation?

2. Define hardening and softening stiffness.

3. What information does a phase-plane trajectory provide?

4. Why can a nonlinear oscillator have multiple steady-state amplitudes at one frequency?

5. What is a limit cycle?

Long Answer Questions

1. Derive the first-harmonic-balance amplitude relation for the forced Duffing oscillator and interpret each term.

2. Explain how multiple-scales or averaging methods eliminate secular terms and describe slow modulation.

3. Discuss bifurcation and Floquet stability in periodically forced nonlinear systems.

4. Explain internal resonance and the mechanisms of energy exchange between two modes.

5. Describe the principal features, diagnostic tools, and engineering implications of deterministic chaos.

Analytical / Application Questions

1. A system has positive cubic stiffness and displays a resonance curve bending toward higher frequency. Explain the likely response branches and propose two design modifications.

2. A structure has a time-varying stiffness near twice its natural frequency. Explain why instability may occur without an external additive force and identify factors that affect the threshold.

3. An experimental response appears irregular and broadband. Develop a procedure to determine whether it is deterministic chaos, random noise, or an unresolved transient.

22. Multiple Choice Questions with Answers

Question 1

For the forced Duffing oscillator, a positive cubic stiffness coefficient generally produces which behavior?

A. Hardening and resonance bending toward higher frequency

B. Softening and resonance bending toward lower frequency

C. Amplitude-independent natural frequency

D. Parametric instability only

Correct Answer: A. Hardening and resonance bending toward higher frequency

Question 2

Why can a frequency sweep through a nonlinear resonance produce a sudden jump in amplitude?

A. The forcing frequency becomes zero

B. A stable periodic branch terminates at a fold bifurcation

C. Damping necessarily becomes negative

D. The system becomes linear at high amplitude

Correct Answer: B. A stable periodic branch terminates at a fold bifurcation

Question 3

Which method is especially useful for obtaining an approximate steady-state amplitude relation in a weakly nonlinear periodically forced oscillator?

A. Dimensional inspection alone

B. Harmonic balance

C. Static equilibrium analysis only

D. Fourier filtering without a model

Correct Answer: B. Harmonic balance

Question 4

In a phase plane, an isolated attracting closed trajectory represents a:

A. Separatrix

B. Saddle point

C. Stable limit cycle

D. Linear resonance frequency

Correct Answer: C. Stable limit cycle

Question 5

Internal resonance is most likely when:

A. All modal frequencies are widely separated

B. Two modal frequencies are approximately in an integer ratio

C. Damping is exactly zero in every mode

D. The forcing contains no harmonic components

Correct Answer: B. Two modal frequencies are approximately in an integer ratio

Question 6

The principal parametric resonance of the Mathieu equation commonly occurs when the parameter-modulation frequency is near:

A. Zero only

B. The square of the natural frequency

C. Twice the natural frequency

D. Half the damping coefficient

Correct Answer: C. Twice the natural frequency

Question 7

Which observation most directly supports period-doubling in a Poincaré section?

A. One point changing into two distinct points

B. A single equilibrium point remaining fixed

C. A purely sinusoidal time record

D. A response with no dependence on initial conditions

Correct Answer: A. One point changing into two distinct points

Question 8

What is the main purpose of Floquet multipliers for a periodic solution?

A. To calculate static displacement

B. To determine perturbation stability over one period

C. To remove nonlinear terms from the model

D. To measure material density

Correct Answer: B. To determine perturbation stability over one period

Question 9

The Van der Pol damping term can generate a limit cycle because it:

A. Always dissipates energy at every amplitude

B. Supplies energy at small amplitude and dissipates it at large amplitude

C. Eliminates all restoring stiffness

D. Makes the system parametrically forced

Correct Answer: B. Supplies energy at small amplitude and dissipates it at large amplitude

Question 10

Which practice is most appropriate when testing for nonlinear hysteresis experimentally?

A. Use only one initial condition and one sweep direction

B. Perform both upward and downward frequency sweeps

C. Ignore phase measurements

D. Assume every irregular response is chaotic

Correct Answer: B. Perform both upward and downward frequency sweeps