Duhamel's Integral¶

Compute the zero-initial-condition response of a linear time-invariant vibration system to an arbitrary load history by causally convolving that load with the system's impulse response.

Version: v2 · 2026-08-30 · History
Domain-specific #: 1726
Origin domain: structural dynamics
Subdomain: linear vibration analysis
Aliases: Duhamel integral, Duhamel convolution integral, Response integral

Core Idea¶

Duhamel's integral is the causal time-domain response formula that turns an arbitrary load history into the forced response of a linear time-invariant vibration system. It treats the load as a continuum of infinitesimal impulses. Each impulse excites the system's known impulse response; linear superposition adds all delayed contributions that can have reached the observation time.

For a quiescent single-degree-of-freedom system

\[ m\ddot{x}(t)+c\dot{x}(t)+kx(t)=p(t), \]

with zero initial displacement and velocity, the response is

\[ x(t)=\int_0^t p(\tau)h(t-\tau)\,d\tau, \]

where \(h\) is the causal displacement response to a unit force impulse. For the underdamped viscous case,

\[ h(s)=\frac{1}{m\omega_d}e^{-\zeta\omega_n s}\sin(\omega_d s),\qquad s\ge 0, \]

so

\[ x(t)=\frac{1}{m\omega_d}\int_0^t p(\tau)e^{-\zeta\omega_n(t-\tau)}\sin\!\bigl(\omega_d(t-\tau)\bigr)\,d\tau. \]

Here \(\omega_n=\sqrt{k/m}\), \(\zeta=c/(2m\omega_n)\), and \(\omega_d=\omega_n\sqrt{1-\zeta^2}\). The named abstraction is not merely “integrate something.” It is the engineering realization of causal convolution in which the kernel is the system's impulse response and the input is a dynamic excitation.^[1]^[2]

Structural Signature¶

The method requires:

Linear time-invariant dynamic system: fixed mass, damping, stiffness, and boundary conditions over the analyzed interval.
Quiescent forced-response component: zero initial displacement and velocity for the convolution term.
Load history: an external force, base-acceleration equivalent, or generalized modal load indexed by time.
Unit impulse: the limiting short-duration input whose finite area changes system momentum.
Causal impulse-response kernel: \(h(s)=0\) for \(s<0\), with form determined by the system and desired output.
Delay variable: an impulse applied at \(\tau\) contributes through \(h(t-\tau)\).
Superposition: scaled impulse responses add without interaction because the governing equations are linear.
Causal integration: only \(0\le\tau\le t\) contributes to response at \(t\).
Response quantity: displacement, velocity, acceleration, or another output obtained with the corresponding impulse-response kernel or derivative.

The invariant is arbitrary excitation decomposed into impulses + one fixed causal impulse response + linear superposition = convolutional time history. If the kernel changes with the excitation, state, or absolute time, the standard Duhamel integral no longer applies unchanged.

What It Is Not¶

Duhamel's integral is not every convolution. Convolution is the substrate-independent operation; Duhamel's integral specializes it for dynamic response, with a causal impulse-response kernel, a load history, structural parameters, and an observation-time upper limit.

It is not a Daniell integral, Riemann integral, or Lebesgue integral as a foundational integration construction. Those define or extend integration itself. Duhamel's integral names a response representation built with an ordinary definite integral.

It is not the Fourier transform or a frequency-domain response calculation. Fourier methods turn convolution into spectral multiplication. Duhamel's integral remains in time and can be evaluated directly from the load history.

It is not Duhamel's principle in its broad PDE sense, although both use superposition of homogeneous evolutions driven by increments of the forcing. The structural-dynamics integral is a specific finite-dimensional response formula.

It is not a general solution for nonlinear, time-varying, hysteretic, or state-dependent systems. Such systems lack one fixed impulse response that scales and shifts for every load increment. Equivalent linearization or Volterra-series approximations add assumptions and are not the exact standard method.^[3]

It is not the whole response when initial conditions are nonzero. The homogeneous free-vibration response must be added separately.

Scope of Application¶

The formula is standard in structural dynamics, mechanical vibration, earthquake engineering, aerospace dynamics, acoustics, and linear controls. It calculates response to pulses, blast and impact histories, wind or machinery loads, measured ground acceleration, arbitrary laboratory inputs, and other nonperiodic excitations.

For base excitation, the equation is usually written in relative coordinates and the effective load is proportional to ground acceleration. For multi-degree-of-freedom systems, modal decomposition can reduce a classically damped linear system to independent modal equations; each modal coordinate receives its own Duhamel integral and the physical response is reconstructed from the modes. Matrix impulse-response formulations provide an equivalent direct statement.

The integral is also a foundation for response spectra: a recorded ground-motion history is applied to a family of linear oscillators with varying natural periods and a specified damping ratio, and peak responses are retained. Numerical implementations approximate the load within each time step or use recursive filters and fast convolution. Dempsey and Irvine derive formulas for unequal time steps and piecewise-linear forcing, illustrating that numerical quadrature is an implementation layer around the analytic identity.^[4]

Clarity¶

The abstraction separates system memory from input history. The kernel \(h\) contains the mass–damping–stiffness dynamics; \(p\) contains the event. Once \(h\) is known, changing the load does not require solving the differential equation symbolically from scratch. Conversely, changing the system changes the kernel even when the input remains fixed.

It also clarifies causality. The integrand at \(\tau\) means: take the small impulse \(p(\tau)d\tau\), delay the unit response by \(\tau\), and evaluate its surviving contribution at \(t\). Values of the load after \(t\) cannot influence the current response, and an underdamped kernel shows how older contributions decay and oscillate rather than vanish immediately.

A useful diagnostic is: Could one experimentally or analytically obtain a single unit-impulse response and predict every scaled, shifted, summed excitation? If yes, Duhamel's representation is available. If impulse shape or amplitude changes the later kernel, linear time invariance has failed.

Manages Complexity¶

An arbitrary load history is infinite-dimensional, and direct differential-equation reasoning can appear to require a new solution for every waveform. Duhamel's integral compresses all system-specific dynamics into one reusable kernel. The event then enters only as weights in a causal accumulation.

This decomposition makes sensitivity legible. Increasing damping changes the kernel's decay; changing stiffness shifts its oscillation frequency; changing mass scales the response. The same earthquake record can therefore be evaluated against many candidate oscillators by changing kernels, while the same oscillator can be tested against many records by changing inputs.

The method also organizes computational choices. Closed-form integration is possible for simple pulses. Numerical quadrature handles tabulated input. Recurrence relations reuse the state between samples. FFT convolution accelerates long stationary records but introduces padding and discretization considerations. Each is an evaluation strategy for the same response identity rather than a different physical theory.

Abstract Reasoning¶

The formula licenses several reusable inferences:

Impulse-response sufficiency: for an LTI system, knowing \(h\) determines the zero-state response to every admissible load.
Shift rule: delaying the load delays the response by the same amount because the kernel depends on \(t-\tau\), not absolute time.
Scaling rule: multiplying the load by a constant multiplies response by that constant.
Addition rule: response to \(p_1+p_2\) equals the sum of separate responses.
Memory diagnosis: the support and decay of \(h\) state how long past inputs matter.
Resonance reasoning: a load whose time structure aligns repeatedly with the oscillatory kernel can accumulate a large response.
Initial-condition decomposition: total response equals zero-input free response plus the Duhamel zero-state response.
Modal extension: when linear equations decouple into modes, apply the same convolution mode by mode and recombine.
Failure diagnosis: amplitude-dependent frequency, path-dependent hysteresis, stiffness change, or time-varying coefficients invalidate one fixed kernel.

Knowledge Transfer¶

Reuse across linear structural, mechanical, acoustic, electrical, and control systems is literal: each retains a time-indexed input, fixed causal impulse response, superposition, and convolutional output. Only the physical interpretation of input, output, and kernel changes.

The wider transferable structure belongs to prime:convolution, which covers filtering, probability, imaging, and distributed-lag models. Calling every convolution a Duhamel integral would erase the engineering commitments. Conversely, recognizing a Duhamel integral as convolution allows frequency-domain multiplication, deconvolution diagnostics, kernel identification, and fast-convolution algorithms to transfer into structural dynamics without confusing them with the named response formula.

Examples¶

Rectangular pulse. A constant force acts from \(0\) to \(T\). Before \(T\), integrate the kernel over the elapsed portion of the pulse; after \(T\), subtract the contribution of an equal delayed step. The result displays the system's transient build-up and free decay after unloading.

Earthquake ground motion. For an SDOF oscillator in relative coordinates, measured ground acceleration supplies an effective forcing history. Convolving it with the oscillator's impulse response produces relative displacement; differentiating or using appropriate kernels provides velocity and acceleration. Repeating across periods yields response-spectrum ordinates.^[5]

Modal structural response. A building model with classically damped modes receives a wind-load vector. Project the load into each modal coordinate, evaluate a Duhamel integral using that mode's frequency and damping, then sum modal shapes weighted by modal responses.

Negative case. A yielding frame loses stiffness as excursions grow and retains hysteretic memory. The unit impulse response measured near the origin cannot be shifted and scaled to predict a large inelastic event. Standard Duhamel integration is not exact.

Structural Tensions¶

T1: Analytic identity versus numerical evaluation. The continuous formula is exact under its assumptions, while real loads are sampled and kernels are discretized. Accuracy depends on interpolation, time step, truncation, and arithmetic even though the representation is exact.

T2: Kernel reuse versus model change. Reusing one impulse response is the method's power. Damage, contact, control switching, temperature drift, or time-varying boundaries can make that reuse invalid.

T3: Time-domain transparency versus frequency-domain speed. Direct integration preserves event timing and causality. Spectral multiplication may be faster for long records but requires transform conventions, padding, and care with transients.

T4: Modal reduction versus coupled fidelity. Mode-by-mode Duhamel integration is efficient when modes decouple. Nonclassical damping or strong coupling may require matrix kernels or direct state-space treatment.

T5: Complete history versus peak summaries. The integral returns a time history; engineering workflows often retain only peaks or spectra. Compression supports design comparison but discards phase and sequence information relevant to cumulative or nonlinear effects.

Structural–Framed Character¶

Duhamel's integral is structural. Its defining roles—load, causal kernel, delay, superposition, and response—are mathematical and mechanically testable. The relation does not depend on a legal, cultural, or evaluative frame.

Its name and the conventional structural-dynamics formula provide domain framing, especially the common SDOF underdamped presentation. The structural core is still exact and portable across LTI physical systems. The entry remains domain-specific because the live Convolution prime already owns the substrate-independent operation, while this node retains the forced-vibration interpretation and scope controls.

Structural Core vs. Domain Accent¶

The structural core is a causal convolution of input with a fixed impulse-response kernel. That lifts directly to prime:convolution.

The domain accent consists of the equation of motion, mass–damping–stiffness parameters, force or ground-motion input, displacement/velocity/acceleration output, zero-state convention, modal extension, response-spectrum use, and engineering failure boundaries. Remove these and the formula becomes generic convolution or Green-function evolution rather than Duhamel's integral as used in structural dynamics.

The minimal prospective placement is a strict specialization of prime:convolution. Duhamel's integral inherits the fixed-kernel slide–weight–integrate structure and adds causality, an LTI dynamic system, an impulse-response kernel, engineering input/output roles, and zero-state response.

prime:superposition and linearity explain why impulse contributions add, but they are constitutive properties already entailed by the convolutional LTI representation and need not become redundant parents. domain_specific:fourier_transform offers an alternative evaluation domain, not a genus. Direct time integration, response spectra, and modal analysis are applications or companion methods.

Relationships to Other Abstractions¶

Current abstraction Duhamel's Integral Domain-specific

Parents (1) — more general patterns this builds on

Duhamel's Integral is a kind of Convolution Prime

The minimal prospective placement is a strict specialization of prime:convolution.

Hierarchy path (1) — routes to 1 parentless root

Duhamel's Integral → Convolution → Function (Mapping)

Neighborhood in Abstraction Space¶

Duhamel's Integral sits in a sparse region of the domain-specific corpus (89^th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Unclustered & Miscellaneous (1565 abstractions)

Nearest neighbors

Computed from structural-signature embeddings · 2026-09-08

Not to Be Confused With¶

Convolution: the general operation. Duhamel's integral is its causal structural-dynamics specialization.

Impulse response: the kernel used by the integral, not the integral or response history itself.

Duhamel's principle: the broader superposition representation for inhomogeneous evolution equations, especially PDEs.

Green's function: a response kernel for a differential operator with specified boundary/initial conditions; a Green-function integral can generalize the same idea in space and time.

Fourier response analysis: solves an LTI problem through spectral decomposition and multiplication. It is mathematically linked by the convolution theorem but operationally distinct.

Direct integration methods: Newmark, central difference, Runge–Kutta, and related schemes advance states stepwise and can handle classes of problems beyond one fixed convolution kernel.

Volterra series: represents nonlinear systems with a hierarchy of multidimensional kernels; the first-order term resembles linear convolution but higher-order interactions are absent from standard Duhamel integration.^[3]

References¶

[1] Hallauer, W. L. Jr. (2016). Introduction to Linear Time-Invariant Dynamic Systems for Students of Engineering, §8.10, “The Convolution Integral as a Superposition of Ideal Impulse Responses.” Virginia Tech Libraries / Engineering LibreTexts. https://eng.libretexts.org/Bookshelves/Electrical_Engineering/Signal_Processing_and_Modeling/Introduction_to_Linear_Time-Invariant_Dynamic_Systems_for_Students_of_Engineering_(Hallauer)/08:_Pulse_Inputs_Dirac_Delta_Function_Impulse_Response_Initial_Value_Theorem_Convolution_Sum/8.10:_The_Convolution_Integral_as_a_Superposition_of_Ideal_Impulse_Responses. registry ↩

[2] Chopra, A. K. (2017). Dynamics of Structures: Theory and Applications to Earthquake Engineering (5^th ed.). Pearson. Standard structural-dynamics derivation of response to general dynamic loading by superposition. registry ↩

[3] Cheng, C. M., Peng, Z. K., Zhang, W. M., & Meng, G. (2017). Volterra-series-based nonlinear system modeling and its engineering applications: a state-of-the-art review. Mechanical Systems and Signal Processing, 87, 340–364. https://doi.org/10.1016/j.ymssp.2016.10.029. registry ↩a ↩b

[4] Dempsey, K. M., & Irvine, H. M. (1978). A note on the numerical evaluation of Duhamel's integral. Earthquake Engineering & Structural Dynamics, 6, 511–515. https://doi.org/10.1002/eqe.4290060507. registry ↩

[5] Indian Institute of Technology / NPTEL. “Dynamics of Earthquake Analysis,” Chapter 3. Government-supported engineering course material deriving the convolution or Duhamel integral from independent impulse responses and superposition. https://archive.nptel.ac.in/content/storage2/courses/105101004/chapters/chapter3/1.10.html. registry ↩