Ordinary Differential Equations and Dynamic Systems

Introduction

Differential equations describe how a quantity changes in relation to other quantities. In machine technology, position changes with time, temperature changes through a component, and a control-system output responds to an input. These are dynamic processes: their present behavior depends not only on present conditions but often also on their history.

An ordinary differential equation (ODE) contains derivatives with respect to one independent variable, usually time t. ODE methods allow engineers to predict motion, determine transient and steady-state responses, assess stability, and design or evaluate machines and control systems.

Learning Objectives

After studying this chapter, you should be able to:

• classify ordinary differential equations by order, linearity, and homogeneity;
• solve important first-order and second-order equations;
• solve higher-order linear equations with constant coefficients;
• formulate and interpret systems of first-order equations;
• analyze free and forced vibration, damping, resonance, and transmissibility;
• model thermal response using a first-order equation;
• explain basic feedback dynamics using transfer functions and state variables;
• select appropriate analytical or numerical methods and evaluate their limitations.

1. Fundamental Concepts

An equation such as

is a first-order ODE. Here, y is the dependent variable, t is the independent variable, and f is a specified function. The order is the highest derivative present. For example,

is a second-order equation. In a mechanical model, x may be displacement in metres, t time in seconds, a may have units kg, b N·s/m, and c N/m.

An ODE is linear if the dependent variable and its derivatives occur only to the first power and are not multiplied together. Thus, y'' + 3y' + 2y = sin t is linear, whereas yy' + y = 0 is nonlinear. A linear equation is homogeneous when its forcing term is zero; otherwise it is nonhomogeneous.

A general solution contains arbitrary constants. An initial condition, such as y(0) = y₀, determines those constants and produces an initial-value solution. A boundary-value problem specifies conditions at two or more different values of the independent variable.

Equilibrium and stability

An equilibrium is a constant solution. For the autonomous equation dy/dt = f(y), an equilibrium yₑ satisfies

If a small displacement from yₑ decays with time, the equilibrium is locally stable; if it grows, the equilibrium is unstable. Linearization near yₑ uses

provided f is differentiable and the disturbance is sufficiently small. The equilibrium is locally asymptotically stable when f'(yₑ) < 0 and unstable when f'(yₑ) > 0. The case f'(yₑ) = 0 requires further analysis.

2. First-Order Differential Equations

2.1 Separable equations

A separable equation can be written as

When h(y) is nonzero, variables can be separated:

Integration gives an implicit or explicit solution. Any equilibrium lost by dividing by h(y) must be checked separately.

Worked Example 1: Cooling component. A component initially at 120 °C is placed in a 20 °C environment. Newton's cooling law is

where T is component temperature, Tₐ is ambient temperature, and k > 0 is a heat-transfer constant in s⁻¹. Separation and integration give

Suppose a measurement shows T = 70 °C after 300 s. Then

70 = 20 + 100e^(−300k), so e^(−300k) = 0.5 and k = ln 2/300 ≈ 0.00231 s^(−1).

The temperature after 600 s is therefore

T(600) = 20 + 100e^(−0.00231×600) ≈ 45.0 °C.

The model predicts approach to ambient temperature but never predicts an instantaneous temperature change.

2.2 First-order linear equations

The standard form is

where p and q are known functions. The integrating factor is

Multiplying the equation by μ gives

and hence

The formula assumes p and q are sufficiently continuous over the interval of interest.

Worked Example 2: First-order actuator response. Consider

τ dy/dt + y = K u₀, with y(0) = 0,

where τ is a time constant in seconds, K is a static gain, and u₀ is a constant input. Dividing by τ gives dy/dt + y/τ = Ku₀/τ. The solution is

The output reaches 63.2% of its final value at t = τ, approximately 95% at 3τ, and over 99% at 5τ. These percentages follow from the exponential model and assume a constant input and a linear, time-invariant system.

2.3 Exact equations and logistic growth

An equation M(t,y)dt + N(t,y)dy = 0 is exact if

on a suitable simply connected region. Then a potential function Φ(t,y) exists with dΦ = Mdt + Ndy, and Φ(t,y) = C is the solution.

A useful nonlinear model is the logistic equation:

where P is population or quantity, r is the intrinsic growth rate, and K is the limiting capacity. Its solution for P(0) = P₀ is

Although originally biological, the same mathematical form can represent constrained adoption of a manufacturing technology or accumulation of wear products under a saturation limit.

3. Higher-Order Linear Equations

A second-order linear equation with constant coefficients has the form

The complementary solution is found from the characteristic equation

For roots m₁ and m₂:

• distinct real roots produce y_c = C₁e^(m₁t) + C₂e^(m₂t);
• a repeated root m produces y_c = (C₁ + C₂t)e^(mt);
• complex roots α ± iβ produce y_c = e^(αt)(C₁ cos βt + C₂ sin βt).

The complete solution is y = y_c + y_p, where y_p is any particular solution. For polynomial, exponential, or sinusoidal forcing, undetermined coefficients is efficient. Variation of parameters is more general but usually more algebraically demanding.

Worked Example 3: Free damped response. A mass–spring–damper system satisfies

Let m = 2 kg, c = 4 N·s/m, k = 8 N/m, x(0) = 0.05 m, and x'(0) = 0. The characteristic equation is

2r² + 4r + 8 = 0, or r² + 2r + 4 = 0.

Thus r = −1 ± i√3. Therefore

x(t) = e^(−t)[C₁ cos(√3t) + C₂ sin(√3t).

From x(0) = 0.05, C₁ = 0.05. Differentiation and x'(0) = 0 give C₂ = 0.05/√3. Hence

The negative real part indicates a decaying response. The oscillation frequency is √3 rad/s, while the envelope decays as e^(−t).

Laplace-transform viewpoint

For initial-value problems, the Laplace transform converts derivatives into algebraic expressions. With zero initial conditions,

where s is the complex transform variable and Y(s) is the transform of y(t). With nonzero initial values,

and

Laplace methods are particularly useful for step, impulse, and piecewise inputs. The transfer function is defined under zero initial conditions as

where U(s) is the input transform. Poles, the roots of the denominator, determine natural modes and strongly influence stability.

4. Systems of Differential Equations

A system of first-order equations can be written in vector form:

where x is the state vector, u is the input, and the dot denotes differentiation with respect to time. A linear time-invariant model is

with output

Matrices A, B, C, and D describe state dynamics, input coupling, measured output, and direct input transmission. Units must be consistent; entries of A have units determined by the selected states.

For a two-degree-of-freedom mechanical system, x may contain two positions and two velocities. Newton's laws then generate four first-order equations. This conversion is useful for simulation because numerical solvers generally advance a vector of first-order states.

For a constant matrix A and zero input, the formal solution is

where e^(At) is the matrix exponential. Stability of a continuous linear system requires every eigenvalue of A to have a negative real part. A positive-real-part eigenvalue produces an unbounded mode; a zero-real-part case needs additional investigation.

5. Mechanical Vibration

5.1 Free undamped vibration

For a mass m attached to a linear spring of stiffness k, with displacement measured from static equilibrium,

The natural angular frequency is

in rad/s, and the frequency in hertz is fₙ = ωₙ/(2π). The ideal solution is x = X cos(ωₙt + φ). Real systems contain damping, friction, nonlinear stiffness, and external disturbances, so the ideal model is an approximation.

5.2 Damped vibration

The viscous damping model is

Define the damping ratio

and the undamped natural frequency ωₙ = √(k/m). The cases are:

For 0 < ζ < 1, the damped natural frequency is

and the response contains the factor e^(−ζωₙt). The decay rate is therefore ζωₙ, measured in s⁻¹.

5.3 Forced vibration and resonance

With harmonic force F₀ cos ωt,

The steady-state amplitude is

where r = ω/ωₙ. The phase lag φ satisfies

with the quadrant selected consistently using an atan2-type interpretation. The amplitude formula assumes linear stiffness, viscous damping, harmonic forcing, and steady-state conditions. The transient component has been omitted.

Near r = 1, low damping can produce resonance and large amplitudes. Engineers avoid resonance by changing mass or stiffness, adding damping, shifting operating speed, or using isolation. The undamped formula must not be used to predict a real resonant amplitude because it becomes unbounded at r = 1; damping and nonlinear effects become important.

6. Thermal and Process Response

A lumped thermal model treats a body as having uniform temperature. An energy balance gives

where C_th is thermal capacitance in J/K, G_th is thermal conductance in W/K, Tₐ is ambient temperature, and P is heat input in watts. The thermal time constant is

For a step heat input P₀ and initial temperature Tₐ,

The final temperature rise is P₀/G_th. This model is valid when internal temperature gradients are small, commonly when the Biot number is sufficiently small. If conduction within the body is slow relative to surface heat transfer, a distributed-parameter heat equation is required instead.

7. Feedback Dynamics

An open-loop system applies an input without using the output for correction. In negative feedback, the error is

where r is the reference and y_m is the measured output. For forward transfer function G(s) and feedback transfer function H(s), the closed-loop transfer function is

for negative feedback and zero initial conditions. The characteristic equation is 1 + G(s)H(s) = 0. Its roots determine closed-loop stability.

Feedback can reduce sensitivity to parameter variation and disturbances, but excessive gain may cause oscillation or instability. Sensor noise, actuator saturation, delays, quantization, and unmodeled dynamics limit practical performance.

A proportional controller is u(t) = K_p e(t). A proportional–integral controller is

where Kp and Ki are controller gains. Integral action can remove steady-state error for suitable systems, but it can cause windup when an actuator saturates. Anti-windup logic and output limits are therefore important in industrial implementation.

8. Numerical Solution and Model Selection

Closed-form solutions are valuable but unavailable for many nonlinear or time-varying models. Euler's method approximates

where h is the time step. It is simple but can be inaccurate or unstable for large h. Fourth-order Runge–Kutta methods improve accuracy by evaluating the slope several times per step. Adaptive solvers alter h according to estimated error.

Numerical results should be checked through dimensional analysis, limiting cases, step-size refinement, conservation laws, and comparison with measurements. A simulation is not automatically valid merely because it produces a smooth graph. Model uncertainty, parameter identification error, sensor bias, and neglected nonlinearities must be assessed.

9. Illustrative Case: Speed Regulation of a Conveyor Drive

Background

Consider an illustrative conveyor driven by a motor and gearbox. The required speed is 1,500 rpm, but load changes produce speed deviations. The drive controller measures speed and adjusts motor torque.

Problem

A simplified rotational model is

where J is reflected rotational inertia, b is viscous rotational resistance, ω is angular speed, τm is motor torque, and τL is load torque. Assume J = 0.20 kg·m² and b = 0.40 N·m·s/rad. The mechanical time constant is

τ = J/b = 0.20/0.40 = 0.50 s.

A sudden constant torque increase of 2 N·m, with τL unchanged, produces the speed change

Thus the final speed increase is 2/0.40 = 5 rad/s, and after 0.50 s it is approximately 5(1 − e^(−1)) ≈ 3.16 rad/s.

Approach

A speed sensor supplies feedback. A proportional controller increases torque when measured speed falls below the reference. If integral action is added, persistent speed error caused by a constant load torque can be eliminated in the ideal linear model. The designer must still check actuator limits, gearbox backlash, sampling delay, sensor filtering, and mechanical resonance.

Outcome and interpretation

The first-order model predicts a smooth exponential response, but an actual conveyor may show flexible-shaft modes and friction dead zones. A commissioning engineer should identify parameters from a safe step test, compare measured and simulated responses, and tune gains conservatively. A model validated at one load is not automatically valid at every speed and product mass.

Learning points

The case demonstrates how a physical balance becomes an ODE, how a time constant describes transient speed response, how feedback changes regulation, and why model limitations matter in professional decisions.

10. Limitations, Safety, and Professional Practice

Differential-equation models are abstractions. Linearization is local; coefficients may vary with temperature, wear, speed, or load. A lumped model may hide spatial gradients. Viscous damping may approximate friction over one operating range but fail near reversal or low speed.

Engineers should document assumptions, units, parameter sources, uncertainty, validation tests, and applicable safety requirements. Simulations used for machine protection must not replace independent safety functions where failure could injure people. Controllers should be tested against sensor failure, actuator saturation, communication loss, and unexpected disturbances. Data-driven and AI-based models can assist parameter estimation or anomaly detection, but they require representative data, traceability, cybersecurity controls, and human oversight. Physical constraints and conservation laws should remain part of the engineering design.

Chapter Summary

Ordinary differential equations provide a mathematical language for dynamic machine behavior. First-order equations describe processes such as cooling, actuator response, and simple speed regulation. Higher-order linear equations are solved through characteristic roots and particular solutions; Laplace transforms provide an efficient framework for initial-value and feedback problems. Systems of first-order equations represent multiple interacting states and support numerical simulation.

The mass–spring–damper equation explains natural frequency, damping ratio, transient decay, forced response, resonance, and isolation. Thermal balances lead to first-order exponential responses governed by thermal capacitance, conductance, and time constant. Feedback systems are described through error signals, transfer functions, characteristic equations, and closed-loop stability. Numerical methods are essential when analytical solutions are unavailable, but results must be checked against units, limiting behavior, measurements, and model assumptions.

Key Terms

Ordinary differential equation: An equation involving derivatives with respect to one independent variable.

Order: The highest derivative appearing in an equation.

Initial condition: A specified value of a variable or derivative at an initial time.

Equilibrium: A constant state satisfying the governing dynamic equation.

Stability: The tendency of a system to remain near, or return toward, an equilibrium after a disturbance.

Integrating factor: A function used to convert a first-order linear ODE into an exactly integrable form.

Characteristic equation: The algebraic equation used to find natural modes of a constant-coefficient linear ODE.

Complementary solution: The solution of the associated homogeneous differential equation.

Particular solution: One solution produced by a nonzero forcing function.

Laplace transform: An integral transform that converts differential operations into algebraic operations in the s-domain.

Transfer function: The zero-initial-condition ratio of output transform to input transform.

State variable: A variable whose present values, together with the input, determine future system behavior.

Natural frequency: The characteristic oscillation frequency of an undamped system.

Damping ratio: A dimensionless measure of damping relative to critical damping.

Resonance: Large forced response occurring near a system's natural frequency.

Time constant: The exponential response scale; a first-order system reaches 63.2% of a step change after one time constant.

Feedback: The return of measured output information to influence the input.

Closed-loop stability: Stability determined by the poles of the feedback-connected system.

Numerical integration: Approximation of a differential-equation solution using discrete computational steps.

Model validation: Comparison of model predictions with independent physical observations.

Review Questions

Short Answer Questions

1. What distinguishes a linear ODE from a nonlinear ODE?
2. Define an equilibrium solution and state the local stability test for dy/dt = f(y) when f'(yₑ) ≠ 0.
3. What physical meaning does the time constant have in a first-order response?
4. Define damping ratio and identify the condition for critical damping.
5. Why can feedback improve regulation yet also create instability?

Long Answer Questions

1. Derive the solution of the first-order cooling equation and explain the meaning of each parameter.
2. Explain the characteristic-root method for second-order constant-coefficient equations, including real, repeated, and complex roots.
3. Derive the natural frequency and damping ratio for a mass–spring–damper system.
4. Explain how Laplace transforms are used to obtain transfer functions and analyze initial-value problems.
5. Compare lumped thermal models with distributed thermal models, including assumptions and limitations.

Analytical / Application Questions

1. A machine axis has m = 5 kg, c = 10 N·s/m, and k = 500 N/m. Calculate its undamped natural frequency, damping ratio, and classify its response.
2. A thermal unit has Cth = 12,000 J/K and Gth = 60 W/K. For a 300 W step input, calculate the thermal time constant and final temperature rise above ambient.
3. A motor-speed model has G(s) = 10/(s+2) and unity negative feedback. Determine the closed-loop transfer function and comment on stability.

Multiple Choice Questions with Answers

Question 1

Which equation is a first-order linear ODE?

A. y' + 3y = t

B. yy' + 3 = t

C. y'' + y = 0

D. (y')² + y = 0

Correct Answer: A. y' + 3y = t

Question 2

For dy/dt = −k(T−Tₐ), with k > 0, what happens as time becomes large?

A. T diverges to infinity

B. T approaches Tₐ

C. T approaches zero regardless of Tₐ

D. T oscillates permanently

Correct Answer: B. T approaches Tₐ

Question 3

A first-order step response reaches approximately what percentage of its final change after one time constant?

A. 25.0%

B. 50.0%

C. 63.2%

D. 99.9%

Correct Answer: C. 63.2%

Question 4

For a mass–spring system with m = 4 kg and k = 100 N/m, what is the undamped natural angular frequency?

A. 5 rad/s

B. 20 rad/s

C. 25 rad/s

D. 400 rad/s

Correct Answer: A. 5 rad/s

Question 5

Which condition represents critical damping?

A. ζ = 0

B. 0 < ζ < 1

C. ζ = 1

D. ζ > 2 only

Correct Answer: C. ζ = 1

Question 6

For a stable continuous-time linear system, where must all eigenvalues of the system matrix lie?

A. On the positive real axis

B. In the left half of the complex plane

C. Only at the origin

D. On the unit circle

Correct Answer: B. In the left half of the complex plane

Question 7

The closed-loop transfer function for negative feedback with forward path G(s) and feedback H(s) is:

A. G(s)H(s)

B. G(s)/(1−G(s)H(s))

C. G(s)/(1+G(s)H(s))

D. H(s)/G(s)

Correct Answer: C. G(s)/(1+G(s)H(s))

Question 8

If a thermal system has Cth = 1,000 J/K and Gth = 20 W/K, its thermal time constant is:

A. 0.02 s

B. 20 s

C. 50 s

D. 20,000 s

Correct Answer: C. 50 s

Question 9

In the forced vibration amplitude formula, the frequency ratio is defined as:

A. r = ωₙ/ω

B. r = ω/ωₙ

C. r = c/k

D. r = m/k

Correct Answer: B. r = ω/ωₙ

Question 10

What is a principal limitation of the Euler numerical method?

A. It cannot represent any first-order equation

B. It always gives an exact solution

C. It may be inaccurate or unstable for a large step size

D. It requires a matrix exponential for every step

Correct Answer: C. It may be inaccurate or unstable for a large step size