Partial Differential Equations and Boundary-Value Problems

Introduction and Context

A partial differential equation (PDE) relates an unknown function of two or more independent variables to its partial derivatives. PDEs are central to mechatronics because temperature fields, elastic displacement, pressure, electric potential, magnetic fields, diffusion, and vibrating structures vary in both space and time. Unlike an ordinary differential equation, which describes change along one independent variable, a PDE describes how a distributed system evolves over a domain.

Typical examples include a temperature distribution T(x,t), a vibrating string displacement u(x,t), and an electrostatic potential V(x,y,z). The governing equation alone rarely determines a unique solution. Initial conditions specify the state at an initial time, while boundary conditions specify behavior at the spatial boundaries. The combination is called an initial-boundary-value problem when both time and boundary data occur.

Learning Objectives

By the end of this chapter, a student should be able to:

• derive first- and second-order PDEs from physical or mathematical descriptions;
• classify second-order linear PDEs as elliptic, parabolic, or hyperbolic;
• formulate appropriate initial and boundary conditions;
• solve standard heat, wave, and Laplace equations by separation of variables;
• use eigenfunctions and Fourier series to represent nonuniform initial or boundary data;
• interpret solutions physically and assess their limitations in mechatronic systems.

1. Formation and Classification of PDEs

1.1 Basic terminology

If u = u(x,y,t), then x, y, and t are independent variables and u is the dependent variable. Notation includes

The order of a PDE is the highest derivative present. Its degree is the power of the highest-order derivative when the equation is polynomial in derivatives. A PDE is linear if the dependent variable and all its derivatives occur only to the first power and are not multiplied together. For example,

is linear, whereas u_t = u u_x is nonlinear. The coefficient α² is the thermal diffusivity in a heat equation and has units m²/s.

A linear second-order PDE in two variables can be written as

where the coefficients may depend on x and y. Its classification is determined by the discriminant

provided A, B, and C are evaluated for the second-order terms.

For equations written with variables x and t, the same principle applies after identifying the coefficients of the second derivatives. Classification is local if coefficients vary with position.

1.2 Forming PDEs by eliminating arbitrary functions

A PDE may be formed by differentiating a relation containing arbitrary functions and eliminating them. For example, let

where f and g are arbitrary differentiable functions. Differentiation gives

Adding and differentiating appropriately yields

Hence the required PDE is

The equation is hyperbolic because its second-order coefficients are A=1, B=0, and C=−1, so Δ=4>0.

1.3 First-order PDEs

A first-order PDE often has the form

For a linear transport equation,

where v is a constant velocity, the solution is constant along characteristics x−vt=constant. Thus

for an arbitrary function f. This represents a profile translating without changing shape. If v>0, the profile moves in the positive x direction.

For a first-order quasilinear PDE, the method of characteristics converts the PDE into ordinary differential equations along curves. For

the characteristic equations are

where s is a parameter. A classical solution requires sufficient differentiability and compatible initial data; characteristic curves must not intersect if a single-valued classical solution is to remain valid.

2. Boundary and Initial Conditions

A boundary-value problem specifies conditions on the boundary of a spatial domain. Common one-dimensional conditions at x=0 and x=L are:

• Dirichlet: u(0,t)=T₀, prescribing the value;
• Neumann: u_x(0,t)=q₀, prescribing a derivative or flux;
• Robin (mixed): a u+b u_x=r, combining value and derivative.

For heat conduction, Fourier’s law is

where q_x is heat flux in W/m², k is thermal conductivity in W/(m·K), and T is temperature in K or °C when only differences are used. A Neumann condition therefore often specifies heat flux rather than temperature.

An initial condition, such as

specifies the state at t=0. Boundary and initial data must be compatible at corner points when a classical solution is required. For example, if u(0,t)=0 and u(x,0)=f(x), then normally f(0)=0.

3. Separation of Variables and Eigenfunctions

Separation of variables assumes that a solution can be built from products of single-variable functions. For a homogeneous problem, write

Substitution often produces an expression in which a function of x equals a function of t. Since both variables are independent, both must equal a constant, conventionally denoted −λ or λ. The spatial equation becomes an eigenvalue problem, and its nontrivial solutions are eigenfunctions.

For fixed-end conditions X(0)=X(L)=0, the spatial problem

has nontrivial solutions only when

These eigenfunctions are orthogonal:

and

Consequently, a suitable initial function can be expanded as

with coefficients

The expansion is interpreted in the usual Fourier sense: convergence is pointwise under standard smoothness and piecewise-continuity assumptions, with special care at discontinuities.

4. The One-Dimensional Heat Equation

4.1 Model and assumptions

For a slender homogeneous rod with no internal heat generation, constant thermal properties, and negligible lateral heat loss, the temperature satisfies

where α²=k/(ρ c_p), k is thermal conductivity, ρ is density, and c_p is specific heat capacity. The units of α² are m²/s. The model assumes one-dimensional conduction and temperature-independent properties.

4.2 Homogeneous end temperatures

Suppose

Set T=X(x)G(t). Substitution gives

and therefore

The spatial boundary conditions yield the eigenvalues and eigenfunctions above. The time equation is

so

Thus the solution is

where b_n are the Fourier sine coefficients of f(x). Every mode decays, and higher spatial frequencies decay faster. This explains why diffusion smooths temperature gradients.

Worked Example 1: initially triangular temperature

A rod of length L has zero temperature at both ends and initial temperature

where T_m is the peak initial temperature. The sine coefficient is

Using symmetry about L/2, the result is

Therefore even terms vanish and

At long times the first mode dominates:

This approximation is useful for estimating thermal settling time. Its characteristic first-mode time scale is τ=L²/(α²π²).

Nonhomogeneous boundary conditions

If the ends are held at nonzero constant temperatures, write

where w(x) is the steady solution satisfying the boundary data and u has homogeneous boundary conditions. For end temperatures T(0,t)=T_a and T(L,t)=T_b,

The transient u=T−w then satisfies the homogeneous heat equation and can be expanded in a sine series.

5. The Wave Equation

5.1 Physical model

For a taut, uniform string with small transverse displacement, constant tension 𝒯, and linear density μ, the wave equation is

Here c is wave speed in m/s, 𝒯 is tension in N, and μ is mass per unit length in kg/m. The assumptions include small slopes, uniform tension, negligible damping, and no external forcing.

For a string fixed at both ends,

with initial displacement and velocity

Separation gives the same spatial eigenfunctions sin(nπx/L). The modal time equation is

where ω_n is angular frequency in rad/s. The complete solution is

The initial displacement determines

and the initial velocity determines

Unlike the heat equation, wave modes do not decay in this ideal model. Energy transfers between kinetic and potential forms. Real systems require damping, which changes the modal time factors and causes eventual decay.

Interpretation

The wave equation is hyperbolic: disturbances propagate at finite speed c, rather than influencing the entire domain instantaneously. The natural frequencies are

where f_n is in hertz. This result is used in vibration analysis, actuator design, robotic link modeling, and condition monitoring.

6. Laplace’s Equation and Steady Fields

Laplace’s equation in two dimensions is

It is elliptic and models steady temperature without internal generation, incompressible potential flow, and electrostatic potential in charge-free regions. Elliptic problems are generally governed by boundary data throughout the boundary rather than by an initial condition.

Consider a rectangular plate 0<x<a, 0<y<b with three sides held at zero potential and the side x=0 prescribed as f(y). Seek u=X(x)Y(y). Substitution gives

The conditions at y=0 and y=b give

The x equation is

To satisfy u(a,y)=0, choose

After normalization at x=0, the solution is

where

The hyperbolic sine appears because the separated spatial equation has a positive coefficient. In electrostatics, u may represent voltage; in steady conduction, it may represent temperature after suitable scaling.

7. Energy, Uniqueness, and Interpretation

For a homogeneous heat equation with zero boundary temperature, a useful energy-like quantity is

Differentiation and integration by parts give

assuming sufficient smoothness and T(0,t)=T(L,t)=0. Thus the squared temperature norm cannot increase. This mathematical result expresses diffusion and supports uniqueness: under appropriate regularity, two solutions with the same data must coincide.

For the undamped wave equation, the physical energy is proportional to

With fixed ends and no forcing or damping, E(t) is constant. The first term is kinetic-energy density and the second is strain-energy density.

A solution must satisfy more than the differential equation. Boundary conditions, initial data, material parameters, geometry, and units must be checked. Fourier series may converge slowly for discontinuous data, and Gibbs oscillations can occur near jumps. Numerical methods are preferable for irregular geometries, nonlinear materials, variable coefficients, or complicated boundary conditions.

8. Illustrative Mechatronics Case: Thermal Settling of a Servo Motor Housing

Background

Consider a motor housing approximated as a one-dimensional rod of length L=0.20 m. The housing is initially warmer near its center because of a localized winding loss. Both ends are connected to heat sinks maintained at the reference temperature. Assume α²=1.0×10⁻⁵ m²/s.

Problem

Estimate the dominant thermal settling time using the first heat-equation mode.

Analysis and approach

For zero end temperatures, the first-mode decay time is

Substitution gives

The first mode is reduced by a factor e^−t/τ. To reduce it to approximately 5% of its initial modal amplitude, solve e^−t/τ=0.05, giving

Outcome and engineering meaning

The estimate indicates that the housing requires roughly 20 minutes for the dominant transient to fall to 5%, under the stated idealizations. A control engineer could use this time scale when selecting a thermal protection delay or designing temperature-observer gains. The estimate is not a guarantee of exact temperature because real housings have three-dimensional conduction, convection, radiation, internal heat generation, and temperature-dependent properties.

Learning points

1. Geometry and diffusivity determine the dominant time scale.
2. Higher modes decay faster and mainly describe early spatial detail.
3. Boundary conditions represent the thermal interface with the environment.
4. Model validation requires comparison with sensor data.
5. Safety decisions should use measured uncertainty and worst-case conditions, not an ideal PDE alone.

9. Computational Practice and Contemporary Tools

Analytical eigenfunction solutions are valuable for insight, verification, and reduced-order models. Finite-difference, finite-element, and spectral methods extend PDE analysis to realistic mechatronic devices. A finite-difference approximation to the heat equation on a grid with spacing ∆x and time step ∆t is

For the explicit one-dimensional scheme, stability commonly requires r≤1/2. Here i indexes space and n indexes time. This is a numerical restriction, not a property of the continuous PDE.

Finite-element software is appropriate for complex geometry and mixed boundary conditions, but mesh quality, time-step selection, material data, convergence, and boundary implementation must be checked. AI-based surrogate models can accelerate repeated simulation or parameter estimation, but they do not remove the need for physical constraints, validation, uncertainty analysis, and safe engineering judgment. In safety-critical applications, model predictions should remain traceable and independently tested.

Challenges and Limitations

The standard equations may fail when assumptions are violated. Important limitations include nonlinear conductivity, large deformation, damping, convection, radiation, internal sources, discontinuous coefficients, and multidimensional geometry. Boundary data may be uncertain or measured only indirectly. Ill-posed inverse problems—such as inferring an internal heat source from boundary temperatures—can amplify measurement noise. Regularization, sensor placement, and uncertainty quantification are then essential.

Boundary conditions also have physical meaning. Imposing both temperature and heat flux at the same boundary can overconstrain a problem, while specifying too little data can make it nonunique. Correct formulation is therefore as important as algebraic solution.

Best Practices

• Identify the dependent and independent variables and their units.
• Classify the PDE before selecting a solution method.
• State geometry, material assumptions, and neglected effects.
• Make boundary and initial conditions explicit and check compatibility.
• Convert nonhomogeneous boundary data into homogeneous data when convenient.
• Use orthogonality to calculate Fourier coefficients rather than guessing them.
• Check limiting behavior: long-time heat solutions should approach steady state, while undamped wave solutions should conserve energy.
• Verify analytical solutions with dimensional analysis, numerical computation, or experimental data.
• Report approximations and uncertainty, especially in thermal and vibration safety decisions.

Chapter Summary

Partial differential equations model distributed physical systems. First-order PDEs can often be interpreted through characteristics, while second-order PDEs are classified as elliptic, parabolic, or hyperbolic using the discriminant of their second-order terms. Boundary conditions prescribe values, fluxes, or mixed behavior at spatial boundaries; initial conditions prescribe the state at a starting time.

Separation of variables converts standard heat, wave, and Laplace problems into spatial eigenvalue problems and ordinary differential equations. Orthogonal eigenfunctions permit arbitrary initial or boundary data to be represented by Fourier series. The heat equation produces exponential modal decay, the ideal wave equation produces persistent oscillatory modes, and Laplace’s equation describes steady spatial fields. These distinctions correspond respectively to diffusion, finite-speed propagation, and equilibrium.

The methods provide analytical benchmarks for mechatronic design, but real systems may require numerical methods, parameter identification, damping models, and experimental validation. Sound engineering practice connects mathematical solutions to physical assumptions, units, uncertainty, and safety.

Key Terms

PDE: An equation involving partial derivatives of an unknown function.

Order: The highest derivative order appearing in a differential equation.

Linear PDE: A PDE in which the unknown function and derivatives occur linearly.

Characteristic: A curve along which a PDE can reduce to an ordinary differential equation.

Boundary condition: A condition imposed on the spatial boundary of a domain.

Initial condition: A condition specifying the state at an initial time.

Dirichlet condition: A prescribed value of the dependent variable.

Neumann condition: A prescribed normal derivative or flux.

Robin condition: A linear combination of value and derivative.

Elliptic PDE: A PDE associated with equilibrium or steady-field behavior.

Parabolic PDE: A PDE associated with diffusion and smoothing.

Hyperbolic PDE: A PDE associated with waves and finite propagation speed.

Eigenvalue: A parameter for which a boundary-value problem has a nontrivial solution.

Eigenfunction: The corresponding nontrivial spatial function.

Orthogonality: A zero inner product between distinct eigenfunctions.

Fourier series: An expansion of a function in orthogonal trigonometric modes.

Thermal diffusivity: The ratio α²=k/(ρc_p) governing heat-spread rate.

Wave speed: The propagation speed c=√(𝒯/μ) in the ideal string model.

Well-posed problem: A problem with existence, uniqueness, and continuous dependence on data.

Review Questions

Short Answer Questions

1. What distinguishes a PDE from an ordinary differential equation?
2. State the discriminant test for classifying a second-order PDE.
3. What physical quantity is commonly represented by a Neumann condition in heat conduction?
4. Why do sine functions arise for a string fixed at both ends?
5. What is the principal physical difference between the heat and ideal wave equations?

Long Answer Questions

1. Derive the separated solution of the one-dimensional heat equation for a rod with zero temperature at both ends.
2. Explain how Fourier coefficients are obtained using orthogonality, including the assumptions needed for a useful series representation.
3. Derive the natural frequencies of a fixed string and interpret the dependence on length, tension, and linear density.
4. Formulate and solve the rectangular-plate Laplace problem with three zero-potential sides and prescribed data on the fourth side.
5. Discuss the mathematical and physical consequences of using incompatible or insufficient boundary conditions.

Analytical / Application Questions

1. A rod has L=0.5 m and α²=2×10⁻⁵ m²/s. Calculate its first-mode thermal time constant and explain what the result means physically.
2. A string has tension 100 N, linear density 0.01 kg/m, and length 1 m. Calculate its fundamental frequency using the ideal fixed-string model.
3. Design a modeling workflow for estimating the temperature of a motor housing from sparse sensors. Identify the PDE model, boundary data, numerical or analytical method, validation procedure, and major sources of uncertainty.

Multiple Choice Questions with Answers

Question 1

Which classification applies to u_t=α²u_xx?

A. Elliptic

B. Parabolic

C. Hyperbolic

D. Algebraic

Correct Answer: B. Parabolic

Question 2

For a string fixed at both ends, which spatial eigenfunction satisfies the endpoint conditions?

A. cos(nπx/L)

B. exp(nπx/L)

C. sin(nπx/L)

D. sinh(nπx/L)

Correct Answer: C. sin(nπx/L)

Question 3

What does a Dirichlet boundary condition prescribe?

A. The value of the dependent variable

B. The time derivative only

C. The material density

D. The eigenvalue only

Correct Answer: A. The value of the dependent variable

Question 4

For the ideal string equation, the wave speed is

A. c=𝒯μ

B. c=√(μ/𝒯)

C. c=𝒯/μ

D. c=√(𝒯/μ)

Correct Answer: D. c=√(𝒯/μ)

Question 5

In the heat equation solution, why do higher modes decay faster?

A. Their exponential decay rate is proportional to n²

B. They have zero amplitude by definition

C. They travel at higher wave speed

D. They violate the boundary conditions

Correct Answer: A. Their exponential decay rate is proportional to n²

Question 6

For L=0.2 m and α²=1.0×10⁻⁵ m²/s, the first heat-mode time constant is approximately

A. 4.05 s

B. 40.5 s

C. 405 s

D. 4050 s

Correct Answer: C. 405 s

Question 7

Laplace’s equation is most directly associated with

A. Steady equilibrium fields

B. Finite-speed shock waves only

C. Pure temporal oscillation

D. Unconstrained random motion

Correct Answer: A. Steady equilibrium fields

Question 8

For the explicit finite-difference heat scheme, r=α²∆t/(∆x)² is commonly required to satisfy

A. r≥2

B. r≤1/2

C. r=π

D. r<0

Correct Answer: B. r≤1/2

Question 9

The discriminant for A u_xx+B u_xy+C u_yy is

A. A²−4BC

B. B²−4AC

C. C²−4AB

D. 4AC−B²

Correct Answer: B. B²−4AC

Question 10

For an undamped fixed string with no forcing, the ideal wave-equation energy is

A. Always increasing

B. Always zero

C. Constant in time

D. Equal to the heat diffusivity

Correct Answer: C. Constant in time