Flow Visualization Experiments

01 — How it works

You cannot see inside

A reactor core is packed with thousands of fuel rods and support structures. They are metal, so no light gets through and nothing about what happens inside is visible from outside. Push a probe in and the probe itself disturbs the flow, so what you measure is no longer what you wanted to measure.

02 — How it works

Build a transparent scale model

So we rebuild the same geometry in something transparent. Acrylic served for a long time, but a great many shapes cannot be machined at all: a wire wound in a spiral, a pebble bed, a gyroid. This is where 3D printing comes in: printed in a light-transmitting resin, even the interior that no cutter can reach comes out as designed. And then we fill it with water and the view closes up again, because water and the model bend light by different amounts. At every boundary the light refracts, so the structures behind appear displaced or vanish altogether.

Water n ≈ 1.33, acrylic n ≈ 1.49, an SLA print n ≈ 1.55. That gap is what distorts the image.

03 — How it works

Match the refractive index

The fix is to bring the fluid to the structure. Two liquids are mixed so that the refractive index of the blend equals that of the model. Once the two media match, light does not bend at the boundary at all, and the structure turns as transparent as if it were gone. The innermost region becomes visible without distortion.

MIR — matched index of refraction. Benzyl benzoate and isopropyl myristate are blended to n = 1.546, with temperature as a fine adjustment.

04 — How it works

Light one plane with a laser

Now we have to choose where to look. Passing a laser beam through a cylindrical lens turns it into a sheet of light as thin as paper. Only what lies in that plane is lit; everything else stays dark. In effect we slice one two-dimensional plane of interest out of the three-dimensional space.

The sheet is typically about 1 mm thick.

05 — How it works

Seed particles and take two frames

The fluid itself is transparent, so its motion is invisible. We mix in fine particles of nearly the same density as the fluid, light enough to follow it faithfully. Two photographs are taken a very short interval apart; the frame is divided into a grid, and for each cell we find statistically how far and in which direction that cluster of particles moved. Divide the displacement by the time interval and you have the velocity at that point.

PIV — particle image velocimetry. The interval Δt between frames is on the order of tens of microseconds.

06 — How it works

The same light also reads temperature and concentration

The laser sheet is already in place, so the same light can be made to show one more thing. Dissolve a little fluorescent dye in the fluid and, where the laser strikes it, it emits light at a different colour from the beam. How bright that emission is depends on how much dye is there and how hot that spot is, so reading the intensity gives a concentration or temperature field rather than a velocity. Particles tell you how fast; fluorescence tells you how well mixed and how hot, on the same plane at the same time.

PLIF — planar laser-induced fluorescence. The emission differs in colour from the laser, so a filter separates the two and each camera takes one.

07 — How it works

Overlay it on the simulation

The measured velocity field is compared, point for point, against a CFD run at the same conditions. Where the two agree, the simulation can be trusted and used in design. Where they disagree is the interesting part: that is exactly the region today's turbulence models fail to explain, and where the next piece of research begins.

Cases compared this way: a 19-pin wire-wrapped bundle against RANS (2020), and helical cruciform fuel against CFD (2025).