Simulations — Numerical simulation

Quick read

One chain, from meshing to shock response.

We started from the fan CAD, built and converged the OpenFOAM case, and extracted a first acoustic reading of the product.

The same stage then crosses three physics tracks — flow, acoustics and structure — with a nominal operating point read at 2,000 rpm, followed by a modal check that pushes the first 1X coincidence out to about 2,950 rpm.

< 37 dB(A) target sound level retained at the nominal operating point
2,000 rpm simulated operating point for the aeroacoustic readout
2,000-2,950 rpm range kept clear of mode 1 in 1X reading, or about 950 rpm of reserve
3 physics tracks crossed in the same stage: flow, acoustics, structure

Learn more — click a sub-step

01CAD & exports

The geometry was first simplified so it would remain meshable and numerically robust. We separated what had to feed OpenFOAM — surfaces and volumes useful for the flow calculation — from what would later be reused for the mechanical and vibration studies.

Starting point: FreeCAD view of the CAD before simplification, solver-oriented volume cleanup, and exports.

02Meshing

We assembled a dedicated OpenFOAM v12 case with a sliding AMI interface around the rotor. The final mesh uses a hexa-dominant snappyHexMesh workflow, locally refined around the blades, the shroud, and the wake, up to roughly 1.95 million cells on the exploratory aeroacoustic case.

Hexa-dominant mesh with local refinement around the rotor, the shroud, and the AMI interface.

03Convergence

Before talking about sound, we first had to converge the flow properly. This stage was used to stabilize the annular jet, the hub wake, the vortex structures, and the aerodynamic loads on the blades so we could work from a usable quasi-periodic regime.

This flow section helped us read the jet, the wake, and the recirculation zones before the acoustic interpretation.

04Aeroacoustic run

Once the flow was established, we continued with a URANS k-ω SST calculation on the sliding AMI mesh, then an acoustic extraction through the Ffowcs-Williams & Hawkings analogy, with several observers positioned 1 m around the fan.

Local reading of the blade pressure loading once the flow had converged.

05Results & reading

The post-processing highlights the tonal content of the fan: the 5-blade passing frequency around 126 Hz and its harmonics, along with their directivity across the observers. This first signature is used as a reference to guide the upcoming acoustic validations, instead of stopping at design intuition.

Acoustic post-processing: reading tonal components and tracking the blade passing frequency.

06Modal & shock

Alongside the CFD work, we returned to the CAD to produce dedicated STP exports for the mechanical studies: a modal analysis of the shroud — first mode around 49.2 Hz — then a shock simulation to identify sensitive areas, especially on the cylindrical shroud and its junctions.

Modal study: first shroud mode around 49.2 Hz.
Most stressed area during the oblique shock case: the junctions worth reinforcing.
Post-impact displacement field: the global dynamic amplitude across the shroud.

07Reading an eigenmode

Modal images amplify the deformation to make it readable: at real scale (×1), the mode is invisible; at ×80, the shroud seems to open up. It is a reading factor, not a real displacement — and that honesty of visualization is part of the deliverable.

The frequency map sums up what matters: seven computed modes against the motor's excitation lines at 2,000 rpm. The first mode (49.2 Hz) keeps a comfortable margin over the rotation (33.3 Hz); the third (131.9 Hz), sitting next to the 4X harmonic (133.3 Hz), is identified and kept under watch.

The same mode 1 at real scale (×1): the deformation is invisible to the naked eye.
Amplified ×80: the mode's shape becomes readable — a reading tool, not a real displacement.
Seven computed modes against the excitation lines at 2,000 rpm: clear margin for mode 1, vigilance on mode 3, next to the 4X.
✕