We build and calibrate a physics-based model of your process, then test the change against every constraint. The answer comes back in euro per day, before the campaign rather than after it.
Three steps, and only one of them is yours.
We turn your existing process and lab data into a physics-based model. No software to install, no setup on your side.
As new data arrives, the model re-fits automatically, so it always tracks what your process is actually doing.
Ask in plain language. Get a quantified, model backed answer: in euro per day, in energy, in the next move to make.
Not a feature list. These are the questions plants actually argue about, and the shape of what comes back.
Ranks every move that earns today, each already checked against your real hydraulic and metallurgical limits, so nothing on the list is blocked by a constraint you would only find later. Here the top three moves are worth €21,000 a day together, and the wet gas compressor is what stops the fourth.
Projects the unit you will actually have at delivery, then tests every candidate mix against thousands of versions of that future rather than one base case. The mixes are ranked by risk-adjusted margin, and the cheapest crude on the list comes fourth once the constraint it triggers is priced in.
The model recalibrates as the plant runs, so drift shows up as a trend long before it becomes a date. At today's settings the run ends 10 weeks short of your planned turnaround, and you know that a year and a half out instead of a quarter out. It then ranks the levers that close the gap, and what holding each one costs per day for the rest of the cycle.
Names which limit binds first today, and what removing it would be worth, before anyone scopes a project around it. Here the preheat train fouling limit is worth €8,600 a day, and the next constraint behind it caps the benefit of any fix at about sixty per cent of that.
One week of planned production, run twice through the model: as scheduled, then with the transitions shaped instead of stepped. The difference is about 22% of the week's output, across three changeovers, and the single number that moves it most is the width of your specification band.
The model reads the instruments already on the unit and returns the property you sell on, continuously, instead of waiting on a sample. Across the same week that lead time is worth 249 t of product landing in the right silo, and it needs no new hardware.
Often the chemistry is not the limit at all. Pushed hard, pressure and heat removal bind first, and both go over together. Which of the two goes first is what decides whether a debottlenecking project is worth running and what it should target, and it is not the same answer at every rate.
The hottest material sits where no sensor reads, and the sensors that come closest are pulled toward the jacket, so they under-read. The control room shows almost two kelvin more headroom than exists. A well-mixed model shows none of this, because it has one temperature and no wall.
On a loop with no purge, the makeup rate sets the product mix, because at steady state makeup must equal net consumption. Cutting it about 6% walks the slate toward the higher-value products: less raw material bought, a richer mix sold.
A passivation layer on every particle is the rate limit. Mechanical work cuts its diffusion time from 776 hours to under two, and that alone takes one-hour conversion from 6% to 90%. The additive package accounts for 1.6% of the effect.
Every answer comes from a physics-based simulation of your actual process, continuously calibrated to your data. It advises. It never touches your controls.
And what it is not: a substitute for plant data. It is the thing that tells you which measurement is worth taking next.
Send us what you're working on. We build the model of your plant and keep it calibrated as it runs. You ask the questions.