The bed that weighs itself

Push air upwards through a bed of sand and the pressure drop rises with the flow, as the previous rung describes: linearly at first, then quadratically.

Then it stops. Not a bend, not a slower rise — it goes flat, and stays flat over a factor of several in velocity while the bed swells, begins to bubble, and takes on every mechanical property of a liquid. Objects denser than the bulk sink in it; objects lighter float; it finds a level; it can be poured out of a hole in the side of the vessel.

The velocity where that happens is where a correlation meets an exact statement, and it is worth being clear which is which.

The exact half #

Draw a control volume round the whole bed. What crosses its faces is a pressure at the bottom, a pressure at the top, and the weight of everything inside. In steady operation with the bed at rest, the difference between the pressures supports whatever fraction of the weight the walls and the grains resting on each other are not carrying.

Once the bed lifts, they carry nothing. So:

$$\Delta p = (1-\varepsilon)(\rho_s - \rho)\,g\,L$$

and that is the whole statement. No drag law. No correlation. No shape factor. Nothing about a grain except how much of it there is.

For the bed drawn here it comes to 7.35 kPa over half a metre — three quarters of a metre of water, from a pile of sand — and that number is exact for any bed of that porosity, density and depth in any fluid, at any flow rate above the threshold.

The half that is a correlation #

Knowing the ceiling does not say what velocity reaches it. That needs the resistance, and the resistance is Ergun’s fitted correlation.

Setting the two equal and solving gives the minimum fluidisation velocity, and the site does it by bisection on the difference. For 500-micron sand in air it comes out at 0.213 m/s; for the same sand in water, at 2.80 mm/s — a factor of seventy-six, because water is both denser and more viscous, and the two effects act on the same side.

That the slope changes from two to a half across the plot is the crossover of the previous rung read sideways. In the viscous regime the weight is balanced by a resistance linear in u, so u_mf ∝ d²; in the inertial regime by one quadratic in u, so u_mf ∝ √d. A single curve carries both, and the bend in it is at a pore Reynolds number of about ten to a hundred.

Comparing with a correlation without checking against it #

Wen and Yu’s fit for the minimum fluidisation velocity is the standard one, and this site draws it and does not assert against it.

The distinction matters and it is the same one the sphere-drag essays make. A check against Wen and Yu would be a check on somebody’s regression of somebody else’s beds; it would pass or fail for reasons that have nothing to do with whether the arithmetic here is right. What the site asserts is the force balance — that at the computed velocity the resistance really does equal the buoyant weight, to a part in a million — and what it reports is the comparison, which comes out 13 per cent apart for sand in air.

Thirteen per cent between two accounts of the same quantity, both of them fitted to real beds, is about the honest precision of this subject. Anybody quoting a fluidisation velocity to three significant figures is quoting a correlation’s arithmetic rather than a bed’s behaviour.

The plateau is the diagnostic #

The flat branch is what makes fluidisation unmistakable in a plant, and it is worth saying why the plateau rather than the corner is the useful signal.

A corner on a plot is hard to locate: two curves meeting at a shallow angle can be read differently by different people. A plateau is unambiguous. Once the bed is fluidised, increasing the flow by a factor of two changes the pressure drop by nothing at all — the bed expands instead, its porosity rises, its resistance per unit height falls, and the product stays exactly at the weight.

That is a self-regulating system with an exact set point, and it is the reason fluidised beds are used where uniformity matters: a bed that is fluidised anywhere is fluidised everywhere, because any region that tried to carry more pressure drop would expand until it did not.

Two beds that behave nothing alike #

The two ends of the size range are worth putting side by side, because a fluidised bed of powder and one of gravel are different machines with the same equation behind them.

Reading the balance as a design equation rather than as a physical one makes the point. The plateau is set by three things a designer chooses — the solid’s density, the packing, and the depth — and by nothing else. A deeper bed costs proportionally more pressure, forever. That is why industrial fluidised beds are shallow and wide rather than tall, and it is a conclusion from a single line of arithmetic that no amount of experience with the correlation would produce.

Why anybody wants this #

A fluidised bed is a solid that behaves like a liquid, and the properties that come with that are why the arrangement is used in oil refining, in power generation, in pharmaceutical drying and in ore processing.

It mixes. Solids in a fluidised bed circulate vigorously, so the temperature is uniform to within a degree across a vessel metres across. For an exothermic catalytic reaction that is the difference between a working plant and a runaway.

It transfers heat. A surface immersed in a fluidised bed sees heat-transfer coefficients an order of magnitude above a gas alone, because particles carry heat to the surface and away by physically moving — which is convection doing the carrying rather than conduction, the same distinction the convection ladder makes about a heated layer.

It can be pumped. Fluidised catalyst is circulated between a reactor and a regenerator through pipes, as a fluid, at tonnes per second. That is the fluid catalytic cracker, which is where most of the world’s petrol comes from.

Every one of those follows from the plateau: the bed cannot be over-pressured, cannot channel permanently, and cannot support a shear stress once it is fluidised.

What the model cannot say #

Bubbles. Above minimum fluidisation, gas-fluidised beds do not expand uniformly — most of the extra gas goes through as bubbles, which rise, coalesce and burst at the surface exactly as in a boiling liquid. Nothing here computes them, and they dominate the real behaviour of the bed: they set the mixing rate, the heat transfer and the fraction of gas that contacts the solid.

Slugging and spouting. In a narrow vessel the bubbles grow to the vessel’s width and the bed moves in slugs; with a jet at the bottom a spout forms instead of a bed. Both are outside a one-dimensional balance.

Entrainment. Above a much higher velocity — the terminal velocity of a particle — the fluid carries the grains away entirely, and the bed is no longer a bed. The competing quantity there is the same particle-tracking arithmetic the regimes field uses for a droplet in a flow, and where the two regimes meet decides whether a plant has a fluidised bed or a pneumatic conveyor.

A distributor. Everything above assumes the flow arrives uniformly at the bottom of the bed. Making it do so is a real design problem, and a poor distributor produces a bed that is fluidised in patches and dead elsewhere — the exact failure the plateau argument says cannot happen, arriving because the premise was false.

Nothing is drawn as a field. As in the two rungs before it, no figure in this essay resolves the flow between grains, and the bubbling that dominates a real bed’s behaviour is exactly the sort of unsteady structure this site declines to draw rather than compute.

Cohesion. Fine powders stick to each other, and below about thirty microns a bed cracks and channels instead of fluidising. The force balance says nothing about it, because a control volume cannot know whether the grains are holding hands.

The threshold that is a discriminant elsewhere #

There is a family of results in this collection whose content is a threshold — a velocity, a Reynolds number or a Stokes number below which one thing happens and above which another does — and it is worth putting this one among them.

The droplet that turns aside has a critical Stokes number of one eighth, from the discriminant of a quadratic. The drag crisis has a Reynolds band in which the drag force falls as the speed rises. A weir chokes at a critical depth found by minimising a specific energy.

Fluidisation is the simplest of the four, because its threshold is where two forces balance rather than where a solution changes character. Nothing becomes unstable at u_mf; a sum simply crosses zero. That is why it can be read off a pressure gauge, and why the other three need something computed before they can be seen at all.

The overshoot, and which direction the gauge was read in #

The plateau is exact and the corner is sharp, and a bed measured in a laboratory frequently shows neither. What it shows on the way up is a peak: the pressure drop rises past the theoretical plateau, overshoots it by ten or twenty per cent — much more for a fine or a long-settled powder — and then falls back onto it as the bed breaks open and expands.

That is not a failure of the control volume; it is the premise being false for a moment. The balance assumes the grains and the walls carry none of the weight, and in a bed that has been sitting undisturbed they carry some: the grains interlock, they arch against the walls, and friction holds the structure together. The flow has to supply an extra force to break that structure before the plateau’s assumption becomes true, and the overshoot is exactly the strength of the packing.

Run the velocity back down from a fluidised state and the peak is gone. The bed has been loosened, the grains are sitting in their loosest packing, the walls are carrying nothing, and the curve falls cleanly onto the packed-bed line at a well-defined corner.

So the measurement has a protocol, and the protocol is part of the number. The standard is to take the minimum fluidisation velocity from the decreasing branch, and a value read off the increasing one is high — sometimes by more than the thirteen per cent this essay records between two correlations. Reported values that disagree between laboratories frequently disagree about which direction the experiment was run in rather than about the bed.

There is a second, quieter version of the same effect in the balance itself. The porosity in the exact expression is the bed’s own, and a settled bed and a loosened one do not have the same one — so even the plateau’s height moves a little with the history, in the direction the packing does.

What kind of statement each half is #

This rung is the field’s method at its clearest, and it is worth naming the two halves once more.

The weight is a control volume: exact, unconditional, independent of the fluid, of the grain shape, of the flow rate and of what the bed is doing internally. It would still be true for a bed of cubes, of fibres, of anything.

The velocity is a correlation: Ergun’s two constants, fitted to other people’s beds, carrying an uncertainty of ten to twenty per cent and shading into a different form for a different packing.

A number is worth what the weaker of its inputs is worth. The fluidisation velocity of a bed is a correlation’s output however exact the balance behind it, and the plateau it leads to is exact however uncertain the velocity. Keeping the two apart in the same sentence is most of what an honest account of an applied problem consists of.

What happens above the plateau #

The flat branch has a right-hand end, and it is worth naming what the bed does between the two.

Just above u_mf a liquid-fluidised bed expands smoothly: the porosity rises with the velocity, the whole bed swells uniformly, and the arrangement is called particulate fluidisation. A gas-fluidised bed almost never does this. Instead the extra gas passes through as bubbles — voids that rise, coalesce and burst at the surface exactly as in a boiling liquid — while the dense phase between them stays at about the minimum-fluidisation porosity.

The distinction is set by the density ratio between solid and fluid, and it decides everything about how the bed is used: bubbles are what mixes a gas-fluidised bed so violently, and bubbles are also what lets gas bypass the solid without reacting with it.

At the far end, when the fluid velocity reaches the terminal velocity of a single particle, the grains are carried away and the bed ceases to exist. The ratio of that velocity to u_mf is typically ten to a hundred, and it is the operating window: below it, a packed bed; within it, a fluidised one; above it, a pneumatic conveyor with no bed at all.

The plateau computed in this essay spans that whole window, and it is flat across all of it — because the buoyant weight of the solid has not changed and the control volume does not care what shape the bed’s interior has taken.

Who found it, and when #

Fritz Winkler patented the fluidised-bed gasifier in 1922, in Germany, for coal — and the technique was in industrial use for two decades before anybody wrote down the pressure-drop balance that explains it. Ergun’s correlation arrived in 1952 and Wen and Yu’s fit for the minimum velocity in 1966.

The fluid catalytic cracker, developed in America during the Second World War and running by 1942, is the reason the technology matters: it is the process that made high-octane aviation fuel available in quantity, and it works because a bed of catalyst can be treated as a fluid and pumped between two vessels.

That the practice preceded the theory by twenty years is worth recording as the shape of this whole field. A control volume explains what a machine cannot exceed. It rarely explains how anybody thought of the machine.

What this rung inherits from the whole field #

It is the last of twenty-seven essays about machines, and the method has not changed once.

Every result in this field came from drawing a box, writing down what crosses its faces, and refusing to look inside. The disc that cannot take more than 16/27; the pipe whose sudden enlargement destroys an exactly computable head; the river whose jump conserves momentum and destroys energy; the rotor whose work is a change of swirl; the vessel whose jet is half its hole; the valve whose slam is ρaΔV; the ram whose ceiling is h/H; the tube in which mixing raises a pressure; and the bed that cannot resist more than it weighs.

Not one of those needed a solved flow field, and several of them are about machines nobody has yet built. That is what the field’s opening rule meant: a control volume does not need to know what is inside it, and the results are exact because of the refusal rather than in spite of it.

What the method cannot do is equally consistent. It never says how a machine will perform, only what it may not exceed; it never explains how anybody thought of the arrangement; and it goes silent the moment the question is about the inside — which is where the other eight fields live.

Where the field goes next #

That is the ninth field complete: fourteen anchors, from a disc in a stream to a bed that floats. What remains on this site is the field that comes after every other one — the explanations everybody has already been taught, tested against the machinery the other eight fields have built.