Blending Two Sumps: Why the Average Dose Fails Both

The reasoning sounds unarguable. Sump A needs 4 percent, sump B needs 1.8 percent, so a 50/50 blend needs 2.9. Mix them in the tanker, save a trip, done. Then the blend arrives at the receiving facility with a paint filter failure and a free liquid layer, and nobody can explain why a mix of two compliant liquids is non-compliant.

It is not a chemistry mystery. It is what happens when a dose that is a nonlinear function of concentration gets averaged as if it were linear.

The curve is not a straight line

Absorption demand per cubic metre rises faster than strength, because at higher dissolved solids the gel has less osmotic driving force. A sump reading 8,000 mg/L COD might need 3.5 percent polymer; one at 16,000 might need 6.5, not 7.0 and certainly not double. Meanwhile a weak sump at 2,000 mg/L may need only 1.2 percent because the loads are low and the gel swells freely.

Average the two strengths and you get 9,000 mg/L, which might genuinely need 3.9 percent. But the average of the two doses is 2.35 percent. The arithmetic of averaging assumes you can dose each half correctly and then combine; blending first means the whole volume gets one dose, and that dose has to satisfy the harder half, not the mean.

Where the water goes

Think about what happens physically. In the strong half, gel particles hydrate slowly and incompletely; there is a lot of solute competing for the same water. In the weak half, gel hydrates fast and to a high ratio, and can end up holding water so loosely that it releases it under modest pressure. Put both in one tank and mix, and the free water released by the over-hydrated fraction migrates into the under-hydrated fraction, which then swells further and immobilises even more water. The blend can end up with a wet layer at the bottom and a stiff layer at the top, both of which fail for different reasons.

In one case, a site blended a 3,200 mg/L ammonia stream with a 240 mg/L stream at 60/40 and dosed the volume-weighted average. The resulting cake passed on the top third of the drum and failed at the base, which is exactly the stratification the model predicts.

Three ways to blend without lying to yourself

Blend for disposal, not for dosing. Mix the two streams in the tanker for transport purposes only, and set the dose on the blended sample after mixing. That means a sample, a jar test if you have time, and acceptance of a dose that will be 30 to 50 percent above the naive average. It works and it is honest.

Or dose sequentially. Put the strong stream in first, dose it at its own rate, then add the weak stream and top up. Two setpoints, one tank, one extra weighing step. The total polymer is usually lower than the blend-then-dose route because the weak fraction never gets over-dosed.

Or keep them apart. Two tanks, two batches, two tickets. This is the option plants resist because it looks like extra paperwork, and the option they return to after the first rejected load.

The sampling problem underneath

All three routes depend on knowing the strength of each stream, and that is where blending schemes usually fail before the chemistry does. A site blending by pump run time assumes the ratio is 50/50 because both pumps are the same model. They are not the same age, the impellers are worn differently, and one has a partly blocked strainer. The real ratio might be 65/35 and nobody knows.

Fit totalisers, or measure tank drawdown. Then calculate the blend ratio from volume, not from run time. It sounds basic. It is also the single most common reason a “blend” turns out to be one stream with a splash of the other.

A simple cross-check costs nothing: measure conductivity of each stream and of the blend. Conductivity mixes roughly linearly with volume, so a blend conductivity that sits closer to the strong stream tells you the ratio has drifted. It is not a laboratory method, but it catches a stuck valve before it becomes a rejected load.

The next blend you plan should carry three numbers on the ticket: volume of each stream, conductivity of each stream, and the single dose applied. If the dose was derived by averaging two other doses, write that down too, because it is the assumption most likely to be wrong. Where a site regularly blends, a grade with a flatter response across the strength range makes the arithmetic less punishing, and the absorption curves by conductivity show how much the curve bends for each product.