Interactive model
The BWRX-300 has no recirculation pumps. Water rises through the core because the steam it picks up makes it lighter than the water around it. Pick a reactor power and the model finds the core flow where that push is balanced by friction and the other losses around the loop.
Same equations as my calc report, JW-TH-001 Rev A, solved live in your browser. Read the project write-up →
Reactor power
Core flow
Buoyancy push
Chimney water speed
Steam model
A heavy column of water on one side, a light column of steam and water on the other. The difference in weight pushes the loop around, and the flow grows until friction uses up all of that push.
Water enters the core at 270.0 °C, a little below boiling at 7.17 MPa. The fuel heats it and it starts to boil 0.73 m up the 3.8 m core.
The core and the 8.64 m chimney above it hold a steam and water mix averaging –. The downcomer outside holds water at –. Over the 12.44 m height, that difference is the push: –.
Friction in the fuel bundles, losses at the core inlet and the steam separators, and speeding the mix up as it boils all take pressure. They grow with the square of the flow, and grow faster when there is more steam.
The model searches for the one flow where the push equals the losses. Right now that is –.
Heavy column against light column
Density along the 12.44 m riser height, now
Where the push goes
Pressure, kPa, at the current operating point
The balance the model solves at every power (JW-TH-001 Eq. 6-13). The form losses are tuned so that 100% power gives the published 1,827 kg/s.
More power makes more steam, so the push grows. But the losses on a steamy mix grow faster. Past about 55% power, adding heat slows the water down.
Core flow against reactor power
With the base model, flow rises to a peak of 1,907 kg/s at 55% power and is already falling at 100%. That shape is typical of natural-circulation boiling water reactors, and it is the real prediction here.
The single value at 100% is not a prediction. The real loss coefficients are proprietary, so I tuned them to make the model give the published 1,827 kg/s. Both steam models pass through that point by construction.
The drift-flux model puts less steam in the chimney at low power, so its flow is lower there and its peak moves to 1,855 kg/s at 70%.
Power above 100% is past the plant's rating. It is shown only to finish the curve.
The only experiment that checks this kind of model at plant conditions is a full-size chimney test. The base model goes past what that test covered.
GE Vernova Hitachi and Hitachi tested a large-diameter chimney at the reactor's real pressure and temperature (reference R3). Water speeds in those tests ran from 0.09 to 0.31 m/s. Here, water speed means the water flow divided by the whole chimney area.
At 100% power the model gives 0.261 m/s, inside the tested range. But from 20 to 60% power the homogeneous model goes above anything tested, peaking at 0.328 m/s at 35%.
Switch to drift-flux and the peak drops to 0.308 m/s, just inside. So the excess comes from the homogeneous model's assumption that steam and water travel at the same speed, not from the plant. I kept the homogeneous model as the base case, as planned before running the cases, and report this as a limitation.
Chimney water speed against reactor power
The flow does not jump straight to its new value. Fuel takes about 5 s to pass on a change in heat, steam takes about 4 s to rise up the chimney, and water takes about 36 s to come back down the downcomer.
Start a run and watch the vessel above or the traces here. On a cut to 50%, flow first jumps up, because the steam losses drop before the chimney empties of steam. It then sags while the downcomer is still delivering water that is too cold for the lower power, and recovers once warmer water arrives.
On a trip to 6%, boiling stops for about 30 seconds and the flow falls to about a quarter of normal before it recovers.
Runs play at 10× speed.
The steady results are what the calc report verifies. This time-dependent part is my extension, built on the same equations to show how the loop moves between steady states. It has not been checked against plant data.
Last run
Seconds since the change
Some inputs are not published, so the report tests how much each one matters. Move one and the panel beside it compares the result with the report's base case.
Real loss coefficients are proprietary, so they are tuned to the published flow. This shows how far untuned values could put the answer.
How much of the loss acts like a water-only loss versus one that grows with steam. Re-tuned so 100% stays at 1,827 kg/s.
Derived from the published core temperatures, not given directly.
Not published. Estimated from a scaled drawing of the vessel, ±10% tested.
Even heating, or a cosine shape that peaks mid-core at 1.4 times the average.
Off keeps the homogeneous model's loss coefficients, which isolates the effect of the steam model alone. Only matters with drift-flux selected.
How fast heat leaves the fuel. Affects power-change runs only.
A one-dimensional steady model of the loop: one average fuel channel split into 20 heights, a chimney and a downcomer. Water and steam properties come from CoolProp (IAPWS-95) at 7.17 MPa. It is the same set of equations as the Python code at tag v1.0, rewritten in JavaScript so it can run here.
Plant data come from public sources only: GE Vernova Hitachi's BWRX-300 General Description, a public fuel licensing report and published test papers.
Each time this page loads, it recomputes 12 results from the calc report and compares them digit for digit. Checking…
| Result | Report | This page |
|---|
The project page covers the calc report: inputs and their sources, assumptions, the verification tests, a hand calculation and every sensitivity case.