Fluid–structure interaction (FSI), explained
Fluid–structure interaction is what happens when a flow loads and deforms a structure, and the deformation can change the flow in turn. This guide explains when one‑way coupling is enough and when you need two‑way, why the coupling can go unstable, how the mesh follows a moving interface, and three ways to run FSI in Simcenter.
What is fluid–structure interaction?
Fluid–structure interaction (FSI) covers problems in which one or more solid structures interact with a fluid flowing inside or around them. The flow loads and deforms the structure, and the deformation can change the flow in turn; when the deformation changes the flow too little to matter, a one-way analysis is enough.
FSI is one of the coupled-physics analyses in CAE simulation, next to combined thermal and mechanical loads and flexible bodies in multibody models. It joins two branches: computational fluid dynamics (CFD) for the flow and structural analysis for the structure. Simcenter, for example, combines finite volume CFD with finite element solid mechanics, the method behind finite element analysis (FEA).
Where FSI matters
- Boat hulls. A planing hull’s performance depends on how the water’s hydrodynamics interact with the hull’s structural dynamics.
- Hydrofoils. The fluid loads on a catamaran’s lifting foil deform it, and the deformation changes how the foil performs; Volupe’s one-way FSI post calls a hydrofoil a textbook example of FSI.
- Rotating blades. Helicopters, urban air mobility vehicles and wind turbines depend on predicting how aerodynamic forces and the blades’ structural response interact.
- Piping and flow-induced vibration. Flow-induced vibration from high-velocity multiphase flow is a common source of vibration in process piping and can lead to fatigue failures and leaks. In a Siemens webinar, Xodus uses Simcenter CFD to predict the forcing and finite element analysis to predict the piping’s response.
- Seals, gaskets, valves and nozzles. These are FSI problems that often also involve mechanical contact between solid parts.
- Heart valves. A prosthetic valve leaflet displaces about its own mass of blood and is very flexible, so blood and leaflet are strongly coupled.
Siemens describes a one-way workflow for a 5.5 m boat at 5 m/s in 1 m waves. Because the hull panels deflect little compared with the boat’s rigid-body motion, Siemens assumes that the deflection has no significant effect on the flow: the fluid pressure deflects the panels, and one-way coupling is enough. That reasoning is specific to the case; in general, compare the deflection with the lengths that shape the flow, as in the practical checks below.
Flutter: when the coupling feeds itself
Flutter is a self-excited FSI instability. Above a critical flow speed, the flutter speed, the flow forces caused by the structure’s own vibration reinforce that vibration, the amplitude grows, and the structure can fail; in aircraft, flutter can involve the coupling of bending and torsion modes. A Siemens article gives the Tacoma Narrows Bridge (1940) as an example. In a limit-cycle oscillation (LCO), nonlinear effects in the structure or the flow limit the amplitude, so the oscillation settles at a constant level; an LCO can grow out of an initially unstable flutter event. Both are two-way coupled FSI. Siemens lists aeroelasticity and flutter among the analyses of Simcenter Nastran. Two-way coupling is not always dynamic: in the HIRENSAD wing case under Verification and validation, the flow bends the wing into a static equilibrium.
This guide covers flow-driven FSI; coupled vibro-acoustics, which Simcenter 3D lists as a separate multiphysics type, is not covered here.
One‑way or two‑way FSI coupling?
The first modelling decision is which way data passes between the fluid and the structure.
| Aspect | One-way coupling | Two-way coupling |
|---|---|---|
| What passes | Fluid loads go to the structure. The deformation is assumed not to change the flow. | Loads go to the structure, the deformation goes back, and the flow domain is updated. |
| When it fits | The deflection is small compared with the lengths that shape the flow, and the structure’s vibration cannot feed back into the flow. Siemens assumes one-way coupling for the hull panels of a small boat in waves. | The deformation sets the flow passage or the flow forces, for example a rubber sleeve that acts as a valve and opens a flow path, or a flexible structure about as dense as the fluid around it; or the structure’s vibration feeds back into the flow, as in flutter. |
| Cost | The flow can be solved once and reused: change the structure without re-running the CFD. | Each solution depends on the other, so the problem is harder to solve. |
| Examples | Hull panels of a small boat in waves; a propeller’s stresses for three materials from one CFD run. | Flutter and limit-cycle oscillations; heart valves; a rubber-sleeve valve in a nozzle. |
A quick test. Ask three questions. First, does the deformation change the shape that sets the flow? In the Siemens nozzle below, the sleeve’s deformation opens the flow passage, and two-way coupling changed the answer. On the model propeller in Volupe’s Marine FSI post, the blades move at most about 0.04 mm, and one-way and two-way coupling differed by under 2 % in thrust coefficient and peak stress. Second, can the structure’s vibration feed back into the flow? Flutter grows from exactly that feedback, which a one-way analysis leaves out, so one-way coupling cannot predict it, however small the static deflection. To answer it, compute the structure’s natural frequencies and check whether the flow can drive those modes, either with a periodic forcing near one of them or, as in flutter, with forces that the vibration itself creates; in a liquid, include the added mass of the fluid, which lowers those frequencies. Third, is the structure a thin wall in a dense fluid such as water or blood, or about as dense as the fluid around it? Then the coupling is strong and, as the next section explains, also hard to solve.
One two‑way coupling step
- Flow solver
The flow solver computes the fluid traction (pressure and shear) on the interface and sends it to the structural model.
- Structural solver
The structural solver computes the resulting deformation and sends it back.
- Move the flow mesh
The flow mesh moves to the new shape, and the solvers go round again: once per time step, or several times within it. The diagram in the next section shows both.
At the interface two conditions must hold. The kinematic condition: the fluid velocity at the interface equals the velocity of the structure, so the fluid neither slips along the wall nor flows through it, and the interface moves with the structure. The dynamic condition: the traction, the force per unit area from pressure and viscous shear, balances across the interface (σf·n = σs·n).
What leaving out two‑way coupling can cost
about 10 %
Mass flow overpredicted when the rubber sleeve was deformed first and the flow solved after
Siemens, sensor-cleaning nozzle
under 2 %
Difference in thrust coefficient and maximum von Mises stress between one‑way and two‑way FSI (6 % in displacement magnitude)
Volupe, PPTC model propeller
In the Siemens sensor-cleaning nozzle, the sequential approach deformed the rubber sleeve under a fixed 1.0 bar load and then solved the flow; it gave a very different sleeve shape and flow field from two-way FSI. In a Siemens example of a prosthetic heart valve, leaving out two-way FSI overpredicted how fast the leaflets open and close, which could lead to a leaflet design that is too slow in real blood.
In Volupe’s Marine FSI post, Florian Vesting converts the Potsdam Propeller Test Case (PPTC) tutorial to two-way FSI to show the set-up; the figures above come from that one demonstration case.
In one-way FSI with file exchange, Florian notes that in some situations one-way coupling is sufficient. He runs the flow and the stress analysis in two separate Simcenter STAR-CCM+ files: the transient CFD run writes the surface pressure to CGNS files, and the finite element run maps it onto the solid mesh. One CFD run then served stress analyses of stainless steel, aluminium and naval brass propellers.
Coupling schemes and stability
Two-way FSI is solved either monolithically, as one system of equations, or with a partitioned scheme, as two solvers that exchange data.
| Aspect | Monolithic | Partitioned |
|---|---|---|
| How it solves | Fluid and structure as one system of equations, with one algorithm. | Fluid and structure separately, each with its own mesh and solver. |
| Interface conditions | Built into the single system of equations. | Passed as data between the two solvers. |
| Strength | Robust in strongly coupled cases: the interface conditions are solved together with both fields, so there are no fluid–structure sub-iterations between two solvers to converge and no added-mass instability of explicit partitioned schemes. | Reuses existing, validated fluid and structural codes. |
| Difficulty | Specialised codes need more resources and expertise to develop and maintain. | Strong coupling (the added-mass effect) can make explicit schemes unstable and implicit sub-iterations slow, and interface data must be mapped between non-matching meshes. |
With the same discretisation, the difference is mainly robustness and cost, not accuracy: in the analysis by Causin and co-authors, relaxed sub-iterations that converge reach the solution of the implicit coupled problem, the problem a monolithic solver solves in one go. Monolithic solvers also follow the moving interface: the Turek–Hron reference results were computed with a fully implicit, monolithic ALE method.
Explicit and implicit coupling
Partitioned schemes differ in how often the solvers exchange data. An explicit (loosely or weakly coupled) scheme solves the fluid and the structure once each per time step: each solver works with the other’s interface data from the previous exchange, so the interface conditions are met only with a time lag. An implicit (strongly coupled) scheme sub-iterates between the two solvers within each time step until both interface conditions are met at the new time level, to a set tolerance.
- Flow solver
Fluid traction (pressure and shear) on the interface
- Structural solver
Deformation of the structure
- Move the flow mesh
The flow domain takes the new shape
Implicit: repeat within the time step until the interface conditions are met (back to step 1, the flow solver)
Explicit, or once converged: next time step
“Strongly coupled” also describes the physics: when each side changes the other a lot, as with a flexible structure about as dense as the fluid around it, explicit schemes tend to go unstable, as the next part explains, so an implicit scheme is needed.
The added‑mass effect
A structure that moves in a fluid has to move some of the fluid with it, and that fluid acts like extra mass on the structure. This added-mass effect is a known source of numerical difficulty. Explicit schemes work in aeroelasticity, but in other applications they can become unstable when the fluid and structure densities are comparable or the domain is slender, whatever the time step. Implicit schemes then struggle to converge and need many sub-iterations per time step.
over 6,000 ×
Steel against air: explicit coupling usually works, as in aeroelasticity
Density ratio
under 8 ×
Steel against water: in a water-filled tube 10 cm long with a 1 cm radius, explicit coupling would need a steel wall about 13 mm thick, more than the radius
Density ratio
Causin and co-authors analyse an incompressible fluid and a thin elastic tube. The explicit scheme they study takes a leap-frog step for the structure and an implicit Euler step for the fluid, and the fluid sees the wall position from the previous step. This scheme goes unstable, whatever the time step, when the mass of fluid per unit area that the wall has to move outweighs the wall’s own mass per unit area (its density times its thickness). That fluid mass grows with the fluid density and with how slender the domain is. For an idealised artery 10 cm long with a 1 cm radius, filled with a fluid as dense as water, they found the limit at about 10.5 g/cm² (105 kg/m²). In steel that is a wall about 13 mm thick, thicker than the radius and far beyond the thin wall their model assumes: in that geometry, no thin steel wall is stable with explicit coupling. Shortened to 2 cm, the same tube has a limit of about 0.7 g/cm², under 1 mm of steel. So steel walls in water can be on the unstable side, and the shorter the domain, the thinner the wall that stays stable.
Not every coupling instability is the added-mass kind. If the structure is much denser than the fluid but very stiff, an explicit scheme can still go unstable, and then a smaller time step can stabilise it. When the sub-iterations use a relaxed fixed-point method, stronger under-relaxation is needed as the structure gets lighter or the domain gets longer and, in the same analysis, for a soft structure or a small time step. Aitken under-relaxation adjusts the relaxation factor automatically from one sub-iteration to the next, and quasi-Newton methods such as IQN-ILS (Anderson acceleration) go further; the documentation of the preCICE coupling library recommends them for most cases.
Moving meshes and data mapping
When the structure deforms, the flow solution has to follow the moving interface, and loads and displacements have to cross that interface, where the fluid and solid meshes generally do not match.
Following the moving interface
Conforming-mesh, or interface-tracking, methods keep the fluid mesh aligned with the moving interface and update it as the structure moves; they give partitioned coupling a convenient framework, and monolithic solvers use them too, as in the Turek–Hron reference results. The Arbitrary Lagrangian–Eulerian (ALE) method is one of them: the fluid mesh moves independently of the fluid, and the flow equations account for the mesh motion. On a moving mesh the flow solver must also satisfy the geometric conservation law: the change in each cell’s volume over a time step must equal the volume its faces sweep in that step, and violating it has been reported to give inaccurate flutter predictions. Non-conforming methods, such as immersed boundary methods, instead impose the interface as a constraint in the equations, so fluid and solid can use independent grids and no remeshing is needed.
At the interface, the meshes on both sides must deform in the same way, and both regions must use the same reference frame; Volupe’s Marine FSI post calls these the consistency conditions of an FSI set-up. To follow large deformations, Simcenter STAR-CCM+ uses mesh morphing, overset meshes and dynamic remeshing. For a flexible part that touches a wall, Christoffer Johansson’s step-by-step FSI set-up in Simcenter STAR-CCM+ moves the solid with a zero-gap overset mesh, which deactivates the cells in a closing gap and reactivates them as it opens.
Mapping loads and displacements
Because the interface meshes on the two sides generally differ, data must be mapped. Quantities that do not add up over the surface, such as pressure or temperature (intensive quantities), are mapped consistently, that is, interpolated from one mesh to the other. Forces and other quantities that add up to a total (extensive quantities) should be mapped conservatively, so that the total force is the same on both meshes. Hou and co-authors describe load-transfer methods that keep both the total force and the work done by the loads the same on the two meshes.
In Simcenter STAR-CCM+ the fluid traction (the surface load from the flow) is interpolated automatically from the fluid side of the interface to the solid side and applied as a load. The interpolation handles conformal interfaces, where the fluid and solid interface meshes match, and non-conformal ones, where they do not (a different sense from the conforming-mesh methods above); similar mesh resolution on both sides is advised, to avoid losing information. In the other direction, the solid’s motion drives the fluid mesh.
With gaps between the two meshes, the interface and the mesh quality can no longer be guaranteed. Simcenter STAR-CCM+ 2606 added a Surface from volume mesh operation that keeps the interface free of gaps, even for moving meshes.
FSI in Simcenter: three ways to couple flow and structure
Which route fits depends on where the structural model lives and how the structure is best described: as a mesh or as a system model.
| Route | Where the structure lives | How it couples | Example |
|---|---|---|---|
| In Simcenter STAR-CCM+ alone | The integrated solid stress solver in Simcenter STAR-CCM+ | Explicit, or iterative with several exchanges per time step; one‑way also by file exchange between two simulation files (CGNS) | Badminton shuttle Lienhart–Gomes benchmark PPTC propeller (one‑way) |
| With Simcenter 3D and Simcenter Nastran | A structural model in Simcenter 3D | One‑way: a Simcenter Data File (from 2206) Two‑way: co‑simulation, introduced in Simcenter STAR-CCM+ 2019.3; check the current release | HIRENSAD wing (Siemens) |
| With Simcenter Amesim | A 1D system model, such as a mass on a spring | Two‑way explicit co‑simulation, or FMI with co‑simulation FMUs | Check valve: rigid plate on a spring, no stresses Electric RIB: system-level FMI |
In one Simcenter STAR-CCM+ simulation
Simcenter STAR-CCM+ has an integrated solid stress solver, so flow and structure can be part of the same simulation, with no co-simulation needed. Siemens lists static, quasi-static and dynamic problems, nonlinear geometry, bonded and small-sliding contacts and, since 2022, mechanical contact between tessellated parts, as in the rubber-sleeve nozzle above.
One simulation does not mean a monolithic solver. The finite volume flow solver and the finite element solid stress solver stay separate and exchange interface data, fluid traction in one direction and solid motion in the other, so in the terms used above this is a partitioned scheme. Explicit coupling, with data exchanged once per set time interval (set in units of time, so it can span more than one time step), is the loosely coupled form. Iterative coupling exchanges data several times per time step; it becomes the implicit, strongly coupled form only when the sub-iterations run until the interface data stop changing. Coupling libraries such as preCICE test this with convergence measures on the exchanged displacements and forces; do the same, and monitor an interface residual within each time step.
The stabilisation settings answer the problems described above:
| Problem | Setting in Simcenter STAR-CCM+ | Use it when |
|---|---|---|
| The added‑mass effect (a thin wall in a dense fluid, or a structure about as dense as the fluid) | Dynamic: treats the fluid that moves with the wall as added mass and predicts how the fluid load changes as the interface moves. Simcenter STAR-CCM+ estimates the displaced volume, or you set it yourself. | A transient two-way case where the wall displaces a lot of fluid |
| Sub‑iterations that need stronger under‑relaxation | Constant Displacement Under‑relaxation: a fixed factor on how much the interface displacement may change from one iteration to the next | A two-way case in which the wall displaces little fluid |
| Convergence very hard to achieve: a dense fluid strongly coupled to a very flexible structure | Quasi-Newton (Anderson acceleration): uses the history of earlier iterations to speed up convergence, and reuses it in the next time step once a step has converged | An implicit unsteady, highly coupled case where the other two do not work; needs the Multiphysics Orchestrator model |
The table follows the Simcenter STAR-CCM+ 2606 user guide, which offers all three methods for transient runs and strongly advises one of them for two-way coupling. Volupe’s 2021 Marine FSI post recommends Boundary Interface Added Mass, the setting that did the added‑mass job in that release; in 2606 it is the Dynamic method. Setting names change between releases, so check the guide for the release you use.
With Simcenter 3D and Simcenter Nastran
When the structural model lives in Simcenter 3D, Simcenter STAR-CCM+ (from release 2206) can export results such as wall pressure to a Simcenter Data File (.scd5). Simcenter 3D imports it and interpolates it in space and time as a load for Simcenter Nastran: a one-way workflow. For two-way cases, Simcenter STAR-CCM+ runs a co-simulation with Simcenter Nastran: Simcenter STAR-CCM+ computes the flow and sends the fluid loads, Simcenter Nastran computes the structure and sends back the displacement, and the morpher moves the fluid mesh to follow. The coupling is explicit, so the two codes exchange data once per time step. Release 2606 supports Simcenter Nastran 2412 and 2506. The co-simulation arrived in release 2019.3; the flow engineer works in Simcenter STAR-CCM+ and the structural engineer in Simcenter 3D. Abaqus is also supported for co-simulation (2022 to 2025 in the 2606 release notes); for other FE codes, ask us.
A specialised co-simulation outside Simcenter 3D: for rotor blades, Simcenter STAR-CCM+ 2206 added a file-based, iterative co-simulation with rotorcraft structural dynamics tools such as RCAS, FLIGHTLAB, CAMRAD II and Dymore, which runs until flow and structure have converged.
With Simcenter Amesim: 1D–3D co-simulation
Sometimes the structure is better described as a system model than as a mesh. In Fabian Hasselby’s check-valve co-simulation with Simcenter Amesim, Simcenter STAR-CCM+ sends the flow force on the valve plate to Simcenter Amesim, which treats the plate as a mass on a spring (the plate moves, but no stresses are computed), computes its new position and sends it back; Simcenter STAR-CCM+ then moves the plate’s overset region. In the example, the built-in link runs a two-way explicit co-simulation over TCP/IP, with data exchanged every 0.005 s over a 0.05 s run.
Simcenter STAR-CCM+ can also couple through the Functional Mock-up Interface (FMI), a free standard for exchanging dynamic simulation models. It imports co-simulation FMUs, which carry their own solver; at the time of Fabian’s 2023 electric RIB post it did not import model-exchange FMUs, which rely on the importing tool’s solver. In that set-up, with FMI 1.0 or 2.0, only scalar values are exchanged, not spatially varying fields, so the link suits system-level quantities, such as boat speed, motor torque and propeller speed in his electric RIB case, rather than surface loads.
Flow and structure
- Flow, solid stress and FSI in one simulationSimcenter STAR-CCM+
- Structures in Simcenter 3D, loaded with Simcenter STAR-CCM+ resultsSimcenter 3D
One tool or two? Flow and structure in one Simcenter STAR-CCM+ simulation need no co‑simulation; a Simcenter Data File takes the flow loads to a structural model in Simcenter 3D.
Systems and training
- 1D system models, co-simulated with the flowSimcenter Amesim
- Courses run by VolupeSimcenter STAR-CCM+ training
Mesh or system model? A valve plate on a spring can be a lumped mass–spring model in Simcenter Amesim instead of a meshed solid.
Planning an FSI study?
Tell us about the structure and the flow, and a Volupe engineer will help you choose between one‑way and two‑way coupling and a Simcenter set-up.
Verification and validation for FSI
FSI codes are tested on benchmark cases with published data. Three examples:
- Turek–Hron. Laminar, incompressible channel flow past a fixed cylinder with an elastic bar attached behind it (the illustration at the top of this page), which, in its unsteady tests, the flow sets into self-induced oscillation, a flow-induced vibration. It was designed to compare coupling approaches, from partitioned, weakly coupled schemes to monolithic ones, and it has separate CFD, CSM (structural) and FSI tests, so each solver can be checked alone first. Of the three FSI tests, FSI1 reaches a steady state and FSI2 and FSI3 oscillate; the structure is as dense as the fluid in FSI1 and FSI3 and ten times denser in FSI2, so FSI3 also tests how a scheme copes with the added-mass effect.
- Lienhart and Gomes. A two-dimensional FSI reference experiment, published in 2006 and presented at ECCOMAS CFD 2006. Siemens ran it in Simcenter STAR-CCM+ 2210 with second-order time integration (the Generalized-α scheme), and Volupe’s 2021 Marine FSI post shows it run in one Simcenter STAR-CCM+ simulation.
- HIRENSAD wing. In 2019, Siemens validated its Simcenter Nastran to Simcenter STAR-CCM+ co-simulation on the HIRENSAD (High Reynolds Number Aerostructural Dynamics) wing from the 1st Aeroelastic Prediction Workshop, a static case in which the flow bends the wing into aeroelastic equilibrium. Siemens reports close agreement with the wind-tunnel data.
Practical checks
- Start one-way, then test the assumption. Compare the one-way deflection with the gaps, clearances and other lengths that shape the flow; if it changes them, go two-way. If the question is flutter or another vibration that the flow can reinforce, one-way coupling cannot answer it. One-way coupling also lets one flow solution serve several structural variants.
- Match the meshes. Use similar mesh resolution on both sides of the interface, to avoid losing information in the mapping.
- Map the right way. Map forces conservatively and pressures consistently, and check that the total force on the solid side equals the force on the fluid side.
- Verify each solver alone first. Check the flow solution and the structural solution separately before you couple them, as the Turek–Hron benchmark does with its separate CFD and CSM tests; an implicit coupling loop can only converge if each solver converges internally.
- Converge the interface. In iterative coupling, monitor an interface residual, such as the change in interface displacement and force between sub-iterations, and move to the next time step only when it meets its tolerance.
- Control the time step. Keep the change in displacement between two time steps within limits: a large jump forces too much fluid to accelerate in one step and makes the coupled solution hard to keep stable. For the fast snap-through of a heart-valve leaflet, Siemens shrinks the step with a dynamic time step control in Simcenter STAR-CCM+. Choose the step with the structure’s vibration modes in mind as well as the flow; from 2021.3, Simcenter STAR-CCM+ can extract mode shapes and frequencies, also during the run. A smaller step helps with fast, large motions and with the mesh, but it does not remove the added-mass instability, and with a relaxed fixed-point scheme it calls for stronger relaxation.
- Watch the mesh as it moves. The larger the mesh displacement, the more the cells distort, and morphing can lower the mesh quality; check the cell quality during the run, and remesh or move to an overset mesh before cells become invalid.
- Diagnose an unstable run. If an explicit coupling goes unstable, compare the structure’s mass per unit area with the mass of fluid per unit area it has to move: a thin, light wall in a dense fluid points to the added-mass instability, which a smaller time step does not cure; switch to iterative coupling with added-mass stabilisation. A smaller time step can cure the other kind, with a structure much denser than the fluid but very stiff.
- Use second order. For dynamic FSI, use second-order time integration where you can: first-order schemes dissipate a significant amount of energy, which can make results inaccurate (Siemens, 2210 release blog). The same blog notes that a small, well-controlled amount of energy dissipation often makes FSI runs more robust, which the Generalized-α scheme from 2210 provides. From Simcenter STAR-CCM+ 2402 the solid stress solver also has backward differentiation time integration, which keeps solid and fluid kinematically consistent when both run at second order.
- Budget the run time. In the transient shuttle example below, Christoffer notes that FSI needs both the finest time step and the finest mesh, so the simulation will need a lot of computing power.
Limits of this guide. These checks are general guidance drawn from the sources and examples above, not a validation procedure for your product. Settings and supported couplings change between Simcenter releases, and several statements here describe a specific release (2019.3, 2206, 2210, 2402 or 2606). Check the documentation of the release you use, and validate each model against test data or a benchmark before you rely on its results.
For verification and validation in general, see our guide to CAE simulation; for mesh convergence on three grids, the grid convergence index (GCI) calculator.
Worked example: FSI on a badminton shuttle in free flight
Christoffer Johansson, Volupe · 2025
A free-flying body that deforms
In a 2025 post, Christoffer Johansson adds FSI to a badminton-shuttle simulation in Simcenter STAR-CCM+. The shuttle flies freely and deforms (Christoffer calls it a deformable DFBI body, after the Dynamic Fluid Body Interaction model in Simcenter STAR-CCM+): a solid region makes it deform according to its material properties, and the fluid mesh morphs around it in every time step. The solid stresses and displacements are computed from the flow around the shuttle.
How the pieces fit together.
- The shuttle’s rigid-body motion has all six degrees of freedom free, so it can turn and fly on its own.
- Its solid region adds a solid displacement on top of that rigid-body motion.
- The overset region around it morphs with the solid at the contact interface, while its outer boundary floats with the body through the background mesh.
- The body’s centre of mass and moments of inertia are computed from the solid’s material properties.
In the first results, a snapshot early in the flight, the stresses are high in the rings that hold the feathers, and the flow is fast between the feathers where the gaps are small. The material is a placeholder for the whole body, cork and feathers alike, until material data for them are available: aluminium at a tenth of its density (270 kg/m³) but with aluminium’s stiffness (68 GPa). The shuttle then weighs 6.6 g instead of 5.2 g, and its moments of inertia come out a little higher, which makes it a bit harder to rotate. Christoffer notes that a lower Young’s modulus would give a more flexible shuttle and more accurate results, so the picture shows the method, not real shuttle stresses.
For a marine case with numbers, see the PPTC propeller above: one-way against two-way FSI, and one CFD run reused for three materials.
Read Christoffer’s post ‘Turning of badminton shuttles and flexible DFBI set up in Simcenter STAR-CCM+’.
FSI terms in brief
- Added mass. Fluid that moves with a moving structure and acts like extra mass on it; a known source of instability in partitioned FSI.
- ALE (Arbitrary Lagrangian–Eulerian). A method in which the fluid mesh moves independently of the fluid to follow the structure.
- Co-simulation. Two tools that run together and exchange data during the solution, for example Simcenter STAR-CCM+ with Simcenter Amesim.
- Conservative and consistent mapping. Transferring forces so that their total is kept, and interpolating pressures or temperatures, between non-matching meshes.
- Explicit and implicit coupling. One solve of each side per time step, with the interface conditions met only with a time lag, or sub‑iterations within each time step until they are met.
- Flutter and LCO. A self-excited vibration that grows above a critical flow speed, and an oscillation whose amplitude nonlinear effects limit to a constant level; both are two‑way coupled.
- FMI and FMU. The Functional Mock-up Interface, a free standard, and the Functional Mock-up Unit, the packaged model it exchanges.
- Monolithic and partitioned. One system of equations for fluid and structure, or two solvers that exchange interface data.
- One‑way and two‑way coupling. The deformation is assumed not to change the flow, or it is passed back and changes the flow.
Sources and changelog
Sources
G. Hou, J. Wang and A. Layton, Numerical methods for fluid–structure interaction: a review, Communications in Computational Physics 12 (2012) (open copy, PDF); P. Causin, J.-F. Gerbeau and F. Nobile, Added-mass effect in the design of partitioned algorithms for fluid–structure problems, Computer Methods in Applied Mechanics and Engineering 194 (2005) (ICES report 04-02, PDF).
S. Turek and J. Hron, Proposal for numerical benchmarking of fluid–structure interaction between an elastic object and laminar incompressible flow, and J. P. Gomes and H. Lienhart, Experimental study on a fluid–structure interaction reference test case, both in Fluid–structure interaction, Lecture Notes in Computational Science and Engineering 53 (2006); TU Dortmund, FSI benchmark pages.
W. A. Silva and S. Dunn, Higher-order spectral analysis of F-18 flight flutter data, AIAA paper 2005-2014 (NASA); L. Geronzi and co-authors, Advanced radial basis functions mesh morphing for high fidelity fluid–structure interaction with known movement of the walls: simulation of an aortic valve, Computational Science – ICCS 2020, Lecture Notes in Computer Science (2020); M. Benoit and S. Nadarajah, On the geometric conservation law for the non linear frequency domain and time-spectral methods, Computer Methods in Applied Mechanics and Engineering 355 (2019).
preCICE documentation, coupling schemes, data mapping and acceleration; Modelica Association, Functional Mock-up Interface.
Siemens blogs: Exciting days (flutter, co-simulation, HIRENSAD), CFD-FEA coupling in Simcenter, FSI and heart valves, FSI of a trileaflet heart valve, FSI with mechanical contact, rotor blade aeroelasticity, multiphysics examples, aeroelastic analysis of a full aircraft, Simcenter Femap 2506 aeroelastic solution enhancements and the Simcenter STAR-CCM+ release blogs for 2019.3, 2021.3, 2210 and 2402; Siemens webinar CFD for flow-induced vibration (with Xodus).
Siemens product pages: Simcenter STAR-CCM+, CFD simulation, Simcenter 3D, Simcenter Nastran, system simulation technologies and finite element analysis.
Worked examples: Volupe blog posts by Christoffer Johansson, Florian Vesting and Fabian Hasselby, linked in the text and below.
Images: Siemens Simcenter image library (planing hull) and Volupe’s own posts. The benchmark illustration and the coupling diagram are drawn by Volupe.
Changelog
October 2026: first version, written from Siemens web pages, blogs and a webinar page, published papers, benchmark pages, coupling-library documentation and the FMI standard’s site, and Volupe’s own worked examples. Revised in October 2026: interface conditions, the one-way test, the coupling schemes, the added-mass estimate and the practical checks. Release-specific statements are dated in the text; they describe the release named, not necessarily the current one.
FSI in practice: articles by Volupe engineers
All articlesTwo-way FSI
diaphragm valve
Simcenter STAR-CCM+
How to set up a FSI simulation in Simcenter STAR-CCM+
Christoffer Johansson, Volupe · 2021
Read the article →
PPTC propeller
one-way, two-way
One CFD run
three materials
Simcenter STAR-CCM+
One-way FSI for a rotating propeller using file exchange
Florian Vesting, Volupe · 2023
Read the article →
Check valve
1D–3D coupling
Simcenter Amesim
Co-simulation using Simcenter STAR-CCM+ and Simcenter Amesim
Fabian Hasselby, Volupe · 2023
Read the article →
Electric RIB
FMU in waves
Simcenter Amesim
Marine co-simulation using Simcenter STAR-CCM+ and Simcenter Amesim
Fabian Hasselby, Volupe · 2023
Read the article →
FMI and FMU
two FMU types
Simcenter Amesim
The functional mock‑up interface and FMUs for co-simulation
Fabian Hasselby, Volupe · 2024
Read the article →
Questions about fluid–structure interaction
Questions about your FSI case?
Volupe’s CFD and structures engineers
Simcenter STAR-CCM+, Simcenter 3D and Simcenter Amesim
Choose “Technical question” in the form below and describe your case.
Ask a technical question →
What is fluid–structure interaction (FSI)?
Fluid–structure interaction covers problems in which a solid structure interacts with a fluid flowing inside or around it. The flow loads and deforms the structure, and the deformation can change the flow in turn. Boat hulls, hydrofoils, rotor blades, valves and heart valves are typical cases.
What is the difference between one-way and two-way FSI?
In one-way FSI the fluid loads deform the structure, and the deformation is assumed not to change the flow. In two-way FSI the deformation is passed back and the flow is updated, so each solution depends on the other and the problem is harder to solve.
When is one-way coupling enough?
When the deformation is too small to change the flow and the structure’s vibration cannot feed back into it, as Siemens assumes for the hull panels of a small boat in waves. If the deformation changes the flow passage, as with a rubber-sleeve valve, or the question is a self-excited vibration such as flutter, it is not. One-way coupling also lets you reuse one flow solution for several structural variants, such as different materials.
What is the difference between CFD and FSI?
CFD solves the flow for a given geometry. FSI couples the flow with a structural analysis, so the flow loads pass to the structure and, in two-way FSI, the deformation passes back and changes the flow. In Simcenter STAR-CCM+ the structural side is finite element solid mechanics; with Simcenter Amesim it can be a system model, such as a mass on a spring.
What is the added-mass effect in FSI?
A moving structure has to move some of the surrounding fluid, which acts like extra mass on it. When the structure’s mass per unit area is smaller than, or comparable to, the mass of fluid per unit area it has to move (comparable densities, a thin wall or a slender domain; a heart-valve leaflet in blood is a strongly coupled example), explicit coupling can become unstable whatever the time step, and implicit coupling needs many sub-iterations.
What is the difference between partitioned and monolithic FSI?
A monolithic approach solves fluid and structure as one system of equations; it is robust for strongly coupled cases but needs specialised code. A partitioned approach solves them with separate solvers and meshes that exchange data at the interface, so existing, validated codes can be reused; iterated to convergence within each time step, it solves the same coupled problem.
Can Simcenter STAR-CCM+ solve FSI on its own?
Yes. Simcenter STAR-CCM+ has an integrated solid stress solver, so the flow and the structure can be one simulation, with separate flow and solid solvers coupled explicitly or iteratively and no co-simulation. For the set-up step by step, see Christoffer Johansson’s post ‘How to set up a FSI simulation in Simcenter STAR-CCM+’. To hand a coupled study over instead, see CFD and FEA consulting from our sister company FS Dynamics.
Can Simcenter STAR-CCM+ be coupled with Simcenter 3D?
Yes, for one-way cases: from release 2206, Simcenter STAR-CCM+ can export results such as wall pressure to a Simcenter Data File, which Simcenter 3D imports as a load for Simcenter Nastran. For two-way cases, Simcenter STAR-CCM+ 2019.3 introduced a co-simulation with Simcenter Nastran; confirm the co-simulation options in your current release, and ask us if you are unsure.
Can an FMU be used for FSI?
In the set-ups Volupe has described, only for system-level quantities: with FMI 1.0 and 2.0, only scalar values are exchanged, not spatially varying fields, so an FMU can carry a system model, such as a boat’s electric propulsion, while the surface loads stay in the CFD and structural models. Simcenter STAR-CCM+ imports co-simulation FMUs; Fabian Hasselby explains the standard in ‘The functional mock-up interface and FMUs for co-simulation’.
Talk to an engineer in your region
Every Volupe business developer holds an engineering degree. Find your regional contact below, or send us a message.
Ask us about FSI (fluid–structure interaction)
Use this form for licence quotes, demos, training and technical questions.
What happens next
- A reply within one day.
- A scoping call with one of our engineers.
- A quote, demo or other next step that fits your enquiry.
After purchase: support from Volupe simulation engineers.