Fluid–structure interaction and biomechanical continua

Follow a pressure pulse through a flexible tube

Imagine two straight tubes filled with the same fluid. Their original dimensions are the same, but one wall is easier to expand. A driver sends a small pressure pulse into each tube. Why can the softer tube delay the pulse even though its radius changes by only a few micrometres? And does a fluid particle travel along the tube as fast as the pressure signal?

Pressure is a normal force per area. A pressure difference of 1.00 kPa acting across a flat 1.00 cm² patch gives a net normal force of 0.100 N: 1000 Pa × 1.00 × 10⁻⁴ m². At a curved wall, apply this relation locally to small patches whose normal directions differ. A pressure load can expand a tube, while the expansion changes how much fluid its interior can hold.

The fluid and the wall therefore share a mechanical problem. The pressure changes the wall, and the wall changes the fluid's space for motion. Fluid–structure interaction means retaining the relevant feedback between those two responses. A rigid-wall calculation suppresses the wall response; a prescribed-motion calculation supplies wall motion instead of solving for it.

Start with the pressure pulse, the interface diagram and the comparison tables. Units, standards, and physical constants and Newton’s laws and free-body diagrams supply useful preparation. Algebra, functions, graphs, and logarithms and Work, kinetic energy, potential energy, and power help with the numerical examples. The field-equation section is an optional return using The classical wave equation and Partial differential equations and Green functions.

All numerical specimens in this lesson are imagined mechanical models with assigned parameters. Their values are not measurements of living tissue. Pressure, deformation and relative fluid motion are defined here; the earlier porous-material lesson and the following fluid-statics lesson are connected topics, not prerequisites for this explanation.

Separate fluid motion from structural response

A fluid can transmit normal pressure and, when viscosity matters, additional normal and tangential stresses. Stress is force per area on an imagined cut through matter; tangential traction is the part of that force density along the cut. The structure responds through its own material law, supports and inertia. An elastic wall stores recoverable energy when it deforms. A viscous fluid dissipates mechanical energy through velocity gradients.

One direction of coupling is easy to see: fluid pressure and shear load the structure. The other direction can change the fluid domain, displace fluid, alter flow resistance or send a pressure disturbance back through the fluid. For a compliant conduit, extra volume storage changes pulse propagation. For an immersed moving body, accelerating surrounding fluid changes the inertial force. Neither effect requires the same material law or the same simplified equations.

A one-way approximation computes one response and uses it to drive the other without feeding the change back. It can be useful when the omitted feedback has a negligible effect on the chosen observable. Two-way coupling solves mutually compatible responses. Small displacement by itself is insufficient to justify one-way coupling: a small fractional area change can still dominate the volume storage that controls a pulse's speed.

Keep three velocities distinct. The local fluid velocity describes material motion. The structural velocity describes movement of the wall or solid. The wave speed describes propagation of a pressure pattern. Only suitable components of the first two must match at an interface; the wave speed is not an interface material velocity.

Match motion and force at an impermeable interface

At an impermeable boundary, fluid cannot pass through the moving wall. The normal fluid velocity at the wall must equal the wall's normal velocity. For the usual viscous no-slip condition, their tangential velocities also agree. These are local boundary statements. They do not say that the mean axial velocity in the tube equals its radial wall velocity or that every interior fluid particle moves with the wall.

Force transmission supplies the second condition. With no separate interfacial mass or surface-stress layer in the model, the traction exerted on one side is balanced by an equal and opposite traction on the other. Opposite outward normals explain the opposite signs. This matches the force per area across the interface; it does not make every component of the two materials' stress tensors identical.

The matching conditions also match mechanical power. A force acting through the same interface velocity supplies power to one material and removes the same power from the other. The interface transfers energy. It supplies no unexplained additional energy when the two contributions are added. External drivers, material dissipation and an active structure must be accounted for separately.

The simple pulse model below neglects fluid viscosity and replaces the full velocity profile by a mean axial velocity. It retains wall-normal motion through changing area and omits tangential shear. It does not impose the viscous no-slip profile while simultaneously pretending that the fluid is uniform right up to a stationary axial wall. If slip, permeability or a surface-stress layer is relevant, use its own boundary law.

Give the wall a small-perturbation law

Choose an equilibrated reference tube with area A₀ and fluid density ρ₀. The exterior pressure is fixed and uniform. Let p be a small change in internal pressure relative to that reference; because exterior pressure is fixed, it is also the change in pressure across the wall, called transmural pressure. Let u be the small mean axial fluid velocity, positive along the tube coordinate x.

Describe the wall by its fractional area compliance C_A = (1/A₀)(dA/dp) at the reference state. Compliance measures how much the area changes under pressure. Its units here are Pa⁻¹, and the linear law is (A − A₀)/A₀ = C_A p. An absolute area compliance dA/dp has different units, m²/Pa, and must be divided by A₀ before it is used as C_A.

Fluid compression supplies a second storage mechanism. Define the bulk modulus K by the small relation (ρ − ρ₀)/ρ₀ = p/K. A positive pressure increment slightly increases the density. Wall expansion instead increases the space available for fluid. The fractional mass storage per original tube volume is governed, to first order, by C_eff = 1/K + C_A.

Use the assigned inputs in the table. The wall is represented by an instantaneous, massless linear area response around this reference state. C_A is the response of the complete constrained tube: wall material, thickness, geometry and axial supports all matter. A generic tissue modulus or a thin-wall formula cannot be substituted without checking those conditions.

The pulse model is uniform, one-dimensional and lossless, with no mean background flow. Its axial disturbance scale is much larger than the tube diameter, and pressure, density and area changes are small. Wall inertia, wall memory, axial structural waves, fluid friction, collapse and reflections are omitted. These conditions define this particular model; they are not properties of all fluid–structure interactions.

Combine storage and inertia to find a wave speed

A pressure gradient accelerates the fluid, while wall expansion and fluid compression allow a segment to take up extra mass. Together these effects give c = 1/√[ρ₀(1/K + C_A)]. The units are m/s: ρ₀ C_eff has units s²/m². In the assigned tube C_eff = 2.0005 × 10⁻⁶ Pa⁻¹ and c ≈ 22.357885 m/s. The pressure signal therefore takes approximately 44.726949 ms to travel 1.00 m.

The comparison table holds fluid density, fluid bulk modulus and the imposed pressure amplitude fixed. Halving C_A makes the wall twice as stiff in this local area-response sense. The pulse travels faster, but the peak fluid velocity for that same prescribed pressure is smaller. Changing which quantity is driven would change the comparison. The rigid-wall case still has finite wave speed because the fluid remains compressible.

The original illustrated pulse is smooth: at the inlet p = (p_a/2)[1 − cos(2πt/T)] for 0 ≤ t ≤ T and zero outside that interval. A right-going pulse replaces t by t − x/c. For this single forward wave, u = p/(ρ₀c). Its peak fluid speed is approximately 44.726949 mm/s, far below the 22.357885 m/s propagation speed. The fluid's first-order net displacement during this pulse is p_a T/(2ρ₀c) ≈ 0.447269 mm, rather than the tube's 1.00 m length.

The aligned profiles at 25, 30 and 35 ms show the pressure pattern and its associated fractional area change. The pulse width is cT ≈ 0.447158 m, over forty times the 10.0 mm diameter. These times all precede the first arrival at the far end, so the plotted interval contains no returned end reflection. Wall area follows pressure instantly in this assigned model; a delayed area response would require another constitutive law.

Wall storage is 4000 times fluid compressive storage here, since K C_A = 4000. Making the fluid incompressible while retaining this compliant wall gives the finite limit c = 1/√(ρ₀C_A) ≈ 22.360680 m/s. Making both the fluid incompressible and the wall rigid removes both storage mechanisms; this reduced model then ceases to provide a finite propagation speed. It is not a prediction of an infinitely fast real signal.

Let the fluid change an oscillator’s inertia

A different coupling appears when a body accelerates in surrounding fluid. The body must set some of that fluid into motion. In an ideal incompressible, inviscid, irrotational fluid, its resulting force can oppose acceleration: F_fluid = −m_a d²y/dt². The positive coefficient m_a is called added mass. It represents fluid kinetic energy associated with the motion, not a fixed parcel of fluid glued to the body.

For a small horizontal translation of a sphere of radius r in otherwise stationary, unbounded fluid, m_a = (1/2)ρ₀(4πr³/3). Boundaries, nearby bodies, free surfaces and different shapes change this coefficient. This ideal potential-flow result omits viscous drag and acoustic radiation. The comparison is around an equilibrium with vertical loads balanced, so buoyancy does not supply a changing horizontal spring force.

Attach that sphere to a linear horizontal spring of stiffness k_s and give the solid mass m_s. Its small-motion equation becomes (m_s + m_a)d²y/dt² + k_s y = 0, giving natural angular frequency ω_n = √[k_s/(m_s + m_a)]. Fluid loading lowers this frequency while preserving the spring's assigned stiffness. With no damping in this model, oscillation does not decay: acceleration-dependent added mass is not a viscous damping force.

The kinetic energy is (1/2)(m_s + m_a)(dy/dt)² and the spring energy is (1/2)k_s y². They exchange during a free oscillation. This is an inertial reduction for a specified geometry, unlike the compliant-tube model in which fluid inertia and distributed volume storage produce propagation along a finite coordinate. An immersed flexible fin can need both structural deformation and fluid loading; its mode shape and force law must then be specified.

Treat biological material as a modelled continuum

A biomechanical continuum represents a biological material by fields such as displacement, stress and pressure at the scale of a chosen question. The averaging volume must be small relative to the variation being resolved but large enough for that description to be meaningful. A bulk wall model can answer a question about overall deformation while missing individual fibres or cells. If their arrangement controls the observable, it must enter the constitutive law or be resolved separately.

A passive elastic law relates deformation to recoverable energy; a viscoelastic law also retains rate or history. Nonlinear response means that the incremental stiffness can change with deformation. Anisotropy means that the response depends on direction, as when an aligned fibre family carries load differently along and across its direction. The local scalar C_A of our tube is a reduced pressure–area response under stated constraints, not a complete three-dimensional description of a tissue.

Large deformation requires a description that follows the change of geometry and distinguishes reference from current material directions. Initial loads and the chosen reference state also matter. Small-perturbation parameters are local to that state and loading range. A parameter fitted at one amplitude, frequency, orientation or temperature should not silently become a universal material constant.

Some biological materials are active: internal processes supply mechanical work, as in contractile material. A model may represent an active stress in addition to passive stress, or an active change of a material's preferred deformation. These are constitutive choices with their own constraints, rather than interchangeable names for an elastic spring. Active mechanical power requires an energy supply; a passive lossless energy balance cannot account for it by itself.

The useful application question is mechanical and specific: which part of a measured force, motion or pressure response is explained by the selected scale and mechanisms? A compliant wall and moving fluid can model conduit motion; a deforming porous skeleton and interstitial flow can model a hydrated specimen. Neither a good curve fit nor one measured modulus establishes every mechanism or validates an entire living system.

Distinguish boundary coupling from pore flow

In a fluid-filled porous solid, fluid can move relative to an internal solid skeleton throughout the specimen. This is different from an impermeable tube wall surrounding a fluid domain. The two materials have separate averaged velocities. Let φ be the fluid volume fraction, v_f the averaged pore-fluid velocity and v_s the skeleton velocity. Their relative volume flux per total area is q = φ(v_f − v_s). It can be nonzero even though each constituent is nearly incompressible.

For a slow isotropic Darcy reduction, q = −(κ/μ)∇p_p, where κ is intrinsic permeability in m², μ is fluid viscosity in Pa·s and p_p is pore pressure. The minus sign means that pressure-driven flow is toward lower pore pressure. The combination κ/μ has units m⁴/(N·s); some sources call this entire combination hydraulic permeability. Always check that convention before inserting a number.

Pressure supports part of the applied compressive load. As fluid redistributes or drains through permitted boundaries, the skeleton's deformation and load sharing can change. The outer boundary may allow drainage, restrict it or permit flow only through selected surfaces. Merely declaring the material saturated does not specify those conditions. Time dependence can arise from this coupled redistribution even when the skeleton itself is an elastic solid.

In a diffusion-dominated poroelastic reduction, assign an effective diffusivity D_p in m²/s. A characteristic drainage scale is τ_char = ℓ²/D_p, where ℓ is the drainage distance. This defines a scaling time, not a universal exponential relaxation time. Exact modes and prefactors depend on geometry and boundary conditions. For the table's assigned D_p = 1.00 × 10⁻⁸ m²/s, doubling the drainage distance multiplies this scale by four.

At observation times much shorter than the relevant drainage scale, little fluid redistribution may occur over that distance. At much longer times, permitted drainage can approach an equilibrated state. These comparisons do not prove that all mechanisms are poroelastic. Solid viscoelasticity and pore flow can coexist; vary size, drainage boundaries, loading mode and observation time to test what a proposed mechanism predicts.

Return to the field balances and interface power

This section is an optional derivation. The symbol ∂ indicates a partial derivative: vary time while holding position fixed, or position while holding time fixed. A short tube segment stores mass at a rate proportional to C_eff ∂p/∂t. More axial outflow than inflow reduces its stored mass. Linear mass conservation is C_eff ∂p/∂t + ∂u/∂x = 0. Pressure force imbalance supplies momentum conservation: ρ₀∂u/∂t = −∂p/∂x. A decreasing pressure along positive x therefore accelerates fluid in that direction.

Differentiate the mass equation with respect to time and the momentum equation with respect to position. Eliminate the mixed derivative of u to obtain ∂²p/∂t² = c²∂²p/∂x², with c² = 1/(ρ₀C_eff). Substituting any differentiable right-going function p = f(t − x/c) into the momentum equation gives u = p/(ρ₀c) for a wave with zero reference velocity; a left-going wave instead has u = −p/(ρ₀c). Superposed reflected waves require both contributions, so a total p/u ratio need not equal ρ₀c.

Multiply the mass equation by A₀p and the momentum equation by A₀u. Adding them gives ∂E_l/∂t + ∂P/∂x = 0, where E_l = (A₀/2)[ρ₀u² + C_eff p²] is energy per tube length in J/m and P = A₀pu is axial power in W. The fluid kinetic part and the combined elastic storage are both positive for positive ρ₀ and C_eff. For the single forward wave they are equal, and P = cE_l. Viscosity or a dissipative wall would add a loss term; an inlet driver supplies boundary power.

For the general interface, let σ_f and σ_s be the fluid and solid stress tensors, and n_f and n_s their outward unit normals, with n_s = −n_f. A traction vector is stress acting on a chosen normal. The conditions for the impermeable no-slip interface without its own stress or mass layer are v_f = v_s and σ_f n_f + σ_s n_s = 0. Taking the force terms' dot products with their shared velocity makes the summed interface power vanish. This explains force and energy consistency without asserting equality of whole stress tensors.

For an isolated passive system with viscous fluid and elastic solid, combined mechanical energy can decrease through viscous dissipation even though the interface powers cancel. An inviscid fluid and elastic structure can instead exchange energy conservatively under their stated boundary conditions. Solving separate numerical subproblems is compatible with two-way coupling if the interface conditions are satisfied consistently; using separate solvers does not by itself establish a one-way physical approximation.

Choose a coupled model before solving it

Begin with the observable and its scale: arrival time, wall displacement, force, oscillation frequency or drainage response. Then choose the retained motion, material laws and interface conditions. A pressure-driven expanding tube, a translating immersed sphere and a compressed porous specimen do not share a universal fluid–structure formula. Each reduced model retains a different route for storing, transferring or dissipating mechanical energy.

Check the neglected mechanisms against the loading and observation time. A slowly loaded conduit may permit a quasistatic wall response. Higher frequencies can make structural inertia and constitutive memory important. Viscous shear can change damping and the velocity profile. Background flow, bends and end conditions can alter propagation and reflections. Large deformation, buckling, collapse or contact invalidates a linear area law before a negative area predicted by that law becomes meaningful.

A quantitative coupling check should address the chosen observable. In our pulse model K C_A compares wall-area storage with fluid compression; in the oscillator m_a/m_s compares fluid and solid inertia. Their effects remain significant even for arbitrarily small linear amplitudes. For drainage, compare observation time with a geometry-appropriate flow scale and specify which surfaces exchange fluid. A model's amplitude range and its coupling strength answer different questions.

Finish with conservation, units and a prediction that can be checked. Does motion satisfy the boundary law? Are opposite interface forces and their powers consistent? Does a computed frequency change follow the retained mass and stiffness? Does changing drainage distance produce the predicted timescale change? These checks turn a fitted response into a mechanical explanation with a stated domain.

Worked examples

Try it yourself

1. At an impermeable viscous no-slip wall, which velocity must equal the wall velocity: the fluid velocity at the interface, the mean axial fluid velocity everywhere, or the pressure-wave speed?

   Show the answer and reasoning

   The local fluid velocity at the interface equals the wall velocity.

   Impermeability matches the normal components; no slip matches the tangential components. Interior mean velocity and propagation speed describe different quantities. An inviscid reduction that permits tangential slip retains normal matching but uses a different tangential boundary model.

2. Keep the assigned fluid and the 1.00 kPa imposed pulse, but halve C_A to 1.00 × 10⁻⁶ Pa⁻¹. Find the new wave speed and peak area change. Does the wave speed double?

   Show the answer and reasoning

   Wave speed approximately 31.614874 m/s; peak fractional area change 0.00100 = 0.100%. The wave speed does not double.

   C_eff becomes 1.0005 × 10⁻⁶ Pa⁻¹. The inverse square-root relation gives the new speed. With dominant wall storage, halving compliance increases speed by nearly √2; the small fluid-compression term makes the ratio slightly smaller than √2.

3. The original 20.0 ms pulse has peak pressure 1.00 kPa. Use its raised-cosine shape to find the first-order displacement of a fluid particle during passage. Has that particle travelled the pulse's 1.00 m observation distance?

   Show the answer and reasoning

   Approximately 0.447269 mm. It has not travelled 1.00 m.

   The pressure integral over the pulse is p_a T/2. Integrate u = p/(ρ₀c) to obtain p_a T/(2ρ₀c). This linear estimate evaluates the waveform at the reference particle position; the small actual displacement is consistent with the small-perturbation model. A pressure pattern advances much farther than an individual particle.

4. The ideal spring–sphere frequency decreases after fluid added mass is included, but the computed free oscillation does not decay. Is this a contradiction?

   Show the answer and reasoning

   No. Added mass stores fluid kinetic energy; it is not a dissipative force in this model.

   The coefficient multiplying acceleration increases while stiffness stays fixed, lowering the natural frequency. Total spring plus solid/fluid kinetic energy is conserved in the stated ideal model. Viscous drag, wave radiation or material losses would require additional terms before amplitude decay could be predicted.

5. A hydrated specimen relaxes after deformation. Someone concludes that its solid skeleton must be viscoelastic because its response depends on time. What mechanical alternative and comparison should be considered?

   Show the answer and reasoning

   Pore-fluid redistribution can produce time dependence even with an elastic skeleton. Compare drainage distance and boundary access as well as loading and time.

   Poroelastic flow couples pressure, skeleton deformation and relative fluid motion. In an appropriate diffusion regime its characteristic time scales with drainage distance squared. A time trace alone does not identify the mechanism; viscoelasticity and pore flow can also coexist.

What to carry forward

• Fluid loads change a structure; structural motion changes the fluid's domain, storage or inertia. State which feedback the model retains.
• At an impermeable viscous no-slip interface, local velocity and opposite outward tractions match. Their mechanical powers cancel when the two material balances are added.
• In the uniform small-pulse model, c = [ρ₀(1/K + C_A)]⁻¹/². Small wall deformation can coexist with strong coupling, and wave speed remains distinct from fluid and wall motion.
• Added mass changes inertia; pore-fluid drainage changes load sharing over a geometry-dependent time. Neither effect is interchangeable with a generic viscous damping or viscoelastic material term.
• Biomechanical continua need explicit scale, constitutive law, direction, reference state, activity, boundaries and observation time. A reduced model answers a bounded mechanical question.