FSI Test Correlation & Validation — Wet Modes, Flow Loops & Coupled Response

Why FSI Validation Needs More Than a Final Response Match

A fluid–structure interaction model combines at least two physical models and an interface transformation. If the final measured displacement or strain agrees with prediction, that agreement does not prove that the fluid forcing, structural dynamics and coupling are each correct. A high pressure prediction combined with an over-stiff structural model can accidentally produce the right displacement. A low damping assumption can compensate for under-predicted forcing. The purpose of staged correlation is to prevent these error cancellations by validating the fluid, structure and interface separately before asking the complete coupled model to reproduce the final response.

Build a Validation Hierarchy

The strongest programme uses a hierarchy of evidence. First establish structural mass, stiffness, boundary conditions and dry modal properties. Then introduce the fluid and measure the change to wet natural frequencies and damping. Next validate the fluid field or pressure transient on a rigid or well-characterised structure where possible. Only after these component behaviours are credible should the fully coupled response be used as a validation target. This hierarchy makes discrepancies diagnosable: a wet-frequency mismatch points toward fluid inertia or structural stiffness; a correct wet mode with wrong response amplitude points more toward damping or forcing.

Dry Modal and Static Correlation

Dry testing provides the structural baseline. Modal frequencies, shapes and transfer functions constrain mass, stiffness, joints and restraints before fluid uncertainty is introduced. Static stiffness measurements can identify boundary-condition or joint compliance errors that may be hidden by modal tuning. Model updating should change physically uncertain parameters rather than arbitrary element properties. If the structural model is tuned using wet data before the dry structure is understood, fluid and structural errors become entangled and the resulting model may not extrapolate.

Wet Modal Testing

Wet modal tests measure the coupled shift in frequency, mode shape and damping caused by the fluid. Depending on scale and fluid state, excitation may use impact, shaker, acoustic or operational methods. The test configuration must reproduce the relevant fill level, immersion depth, surrounding boundaries and temperature because these influence the fluid contribution. Comparing both dry and wet results provides a direct estimate of added-mass effect and helps separate structural stiffness error from hydrodynamic error. Mode tracking should use shape correlation, not mode number alone, because fluid loading can reorder modes.

Flow and Pressure Correlation

For forced FSI, pressure or velocity measurements are needed to establish whether the fluid excitation is correct. In external flow this may use surface pressure taps, pressure-sensitive measurement, PIV or wake instrumentation. In internal flow it may use synchronised pressure transducers along the line. The comparison should include mean value, fluctuation spectrum, phase and spatial distribution as appropriate. A CFD field that gets the average pressure right but misses the dominant unsteady frequency cannot validate a vibration prediction.

Structural Response Correlation

Strain gauges, accelerometers, displacement sensors and load cells provide the structural response. Sensor placement should be selected using model sensitivity so that the measurements distinguish the modes and load paths of interest. Comparing only the sensor with the largest amplitude can hide a wrong mode shape. Frequency response functions, operational deflection shapes, strain ratios and phase between sensors are often more diagnostic than isolated maxima. For nonlinear FSI, correlation should span the amplitude or operating range over which the nonlinearity changes.

Synchronisation and Phase

FSI validation depends on phase because energy transfer across the interface is governed by the relationship between force and motion. Fluid and structural channels should therefore share an accurate time base. Unsynchronised acquisition can make a physically correct model look wrong or, worse, hide incorrect phase relationships when only amplitudes are compared. In transient tests, trigger definition, sensor latency and filtering delay should be documented. In periodic tests, cross-spectral phase and coherence provide useful measures of whether fluid forcing and structural response are dynamically linked as predicted.

Uncertainty and Repeatability

Test data contain uncertainty from sensors, mounting, environmental variation, operating condition and specimen-to-specimen differences. Repeat runs are valuable because FSI phenomena such as vortex lock-in, turbulent buffeting or contact can have inherent variability. Correlation tolerances should reflect measurement uncertainty and the intended engineering decision. A difference that is negligible for a screening load may be unacceptable when establishing a stability boundary. Uncertainty bands around both test and analysis are more informative than a single percent-error value.

Avoiding Compensating Calibration

Model updating should proceed from observable physics. Correct structural mass and stiffness using dry data; correct hydrodynamic added mass using wet modal shifts; correct forcing using pressure or flow data; and only then adjust damping using response amplitude or decay. Changing several parameters simultaneously to match one final curve destroys identifiability. The updated model may reproduce the calibration test but fail at another flow speed, fill state or configuration because the wrong physical parameters were altered.

A successful FSI correlation explains why each subsystem agrees. Matching the final response by tuning several coupled parameters at once is curve fitting, not validation.

Validation Matrix Across the Envelope

The validation case should not be only the nominal design point. Select cases that exercise different aspects of the model: low and high flow speed, multiple fill levels, different excitation frequencies, or configurations with altered restraint where practical. The goal is to demonstrate predictive capability over the intended use domain. If the model is validated only at one point, its extrapolation range should be stated explicitly and sensitivity or uncertainty allowances should cover the unvalidated region.

Evidence Package

• Dry structural evidence — Mass properties, static stiffness, dry modes and boundary conditions.
• Wet dynamic evidence — Frequency shift, wet mode shape and damping where fluid loading matters.
• Fluid forcing evidence — Pressure, velocity, spectrum or transient timing appropriate to the mechanism.
• Coupled response evidence — Synchronized strain, acceleration, displacement or load with phase information.
• Numerical verification — Mesh, time step, coupling tolerance and transfer conservation.
• Uncertainty statement — Measurement, model-form and parameter uncertainty tied to the engineering decision.
• Independent prediction case — Use at least one case not employed for parameter tuning where possible.

From Correlation to Substantiation

A validated FSI model becomes useful when its accepted domain and residual uncertainty are clear. The final substantiation should distinguish quantities directly correlated with test from those predicted by extrapolation, document any bias factors or conservative assumptions, and retain traceability from raw measurements through processed comparison metrics to the engineering limit. This makes the model an evidence-generating tool rather than a one-off demonstration. The result should answer not only “does the plot match?” but “what has been validated, over what range, with what uncertainty, and is that adequate for the decision?”

Key takeaways

• FSI validation should separate structural, fluid and interface evidence before judging the final coupled response.
• Dry-to-wet modal shifts are powerful diagnostics for added-mass and damping modelling.
• Synchronised pressure and structural measurements are essential when phase controls energy transfer.
• Model updating should change physically observable parameters sequentially to avoid compensating errors.