Why two batches of the same grade cost different money
The recipe is fixed. The equipment is the same. And yet the yield, the cycle time and the steam bill move batch to batch, and the reason usually arrives after the batch has shipped.
Variability with nothing to compare against live
You have the historian trend and you have the batch sheet. What you don't have is this batch laid over the best forty batches of the same grade, phase by phase, while it is still running.
The lab result arrives after the decision
Sampling, queueing and analysis take hours. By the time the assay comes back, the batch has moved to the next phase or the next vessel.
Utilities are your second-largest cost and nobody costs them per batch
Steam, chilled water, nitrogen and compressed air are tracked at the plant meter. Almost never at the batch, the grade or the equipment train that consumed them.
Cleaning and changeover quietly eat capacity
Grade sequencing decides how many CIP cycles you run in a month. Most plants sequence by order date and pay for it in cleaning time, solvent and lost reactor hours.
Genealogy is rebuilt by hand
Tracing a finished lot back through intermediates to a raw material lot means pulling paper records across departments — and it always happens under pressure.
Years of historian data, no decisions from it
The tags have been logged faithfully since commissioning. Nobody has ever turned that archive into a model of what a good batch looks like.
What we connect to in your plant
We read from your existing DCS and historian. Connection is passive — nothing is written back to the control system unless you explicitly ask for it and your controls team signs it off.
DCS and control systems
Historians and SCADA
Business and quality systems
The signals we read, and what each one tells you
A tag on its own is noise. What matters is the question each one answers for the person standing at the machine.
| Signal | What it tells you |
|---|---|
| Internal and jacket temperature with ramp rate | Exotherm control and whether this phase is tracking the profile that produced your best yields. |
| Agitator motor power draw | Viscosity change, crystallisation onset, and impeller or seal trouble — long before anyone opens the manhole. |
| Dosing flow and totaliser per charge | Addition-rate deviations, which is where a surprising share of yield variance originates. |
| Pressure, vacuum and reflux ratio | Separation efficiency and the early shape of a column upset. |
| pH, conductivity and turbidity | In-process quality indication hours before the lab confirms it. |
| ISA-88 phase timestamps | Where cycle time actually goes — charging, reaction, workup, transfer, cleaning — instead of one number for the batch. |
| Steam, chilled water, nitrogen and power per batch | True conversion cost per kilogram, by grade and by equipment train. |
| CIP cycle count and duration | What each grade-to-grade transition really costs in cleaning time and solvent. |
| Effluent flow and load | Environmental cost attributed back to the batch and grade that generated it. |
Plant, quality and finance looking at the same batch
One record per batch that production, QA, EHS and costing all trust, because it comes from the instruments rather than from three separate spreadsheets.
Batch cycle time by phase
hours, against best-in-class
Golden-batch conformance
%, per phase
First-pass yield
%, by grade
Off-spec and rework rate
%, by grade
Steam per tonne
kg per tonne of product
Power per tonne
kWh per tonne
Solvent recovery
% recovered
Reactor and dryer occupancy
% of available hours
Changeover and CIP time
hours per transition
Cost per kilogram
currency, by grade
Effluent load per tonne
COD / BOD per tonne
Batch record closure time
hours from end of batch
AI trained on your process. Not a generic factory.
Models built on batch process behaviour — phases, recipes, campaigns and shared equipment — not on a generic time series.
Golden-batch fingerprint and live deviation
Builds the reference profile from your own best historical runs, then alerts phase by phase when the current batch drifts from it — with the correction still available to the operator rather than a post-mortem for the QA meeting.
Soft sensors for end-point prediction
Predicts assay, moisture or conversion from the live process signature, so the desk can decide whether to hold, extend or move on without waiting on the lab queue.
Specific-energy optimiser
Learns steam and power per tonne by grade, season and equipment train, then surfaces the setpoint combinations behind your cheapest good batches. Utilities usually hold the fastest payback in a specialty plant.
Campaign sequencer
Orders grades across shared reactors, dryers and centrifuges to cut cleaning cycles and changeover loss, while respecting the commitments you have already made to customers.
Your first thirty days
Read-only, one train, no DCS changes. The historian does most of the work in the first fortnight.
Tap the historian and one train
Read-only connection to the DCS or historian for a single reactor train. Tag mapping and cleanup happen here — this is usually the honest bottleneck, not the technology.Â
Â
2
Rebuild the last twelve months
Every historical batch of the chosen grade is reconstructed phase by phase and ranked. The golden profile emerges from your own data, not from a textbook.
3
Add lab and utilities
LIMS results and utility meters join the same timeline, so a batch can finally be costed end to end in currency per kilogram.
4
Run live deviation alerting
One grade goes live with phase-by-phase alerting. The desk sees a deviation while the correction is still possible.
What plants aim for in the first quarter
5–12%
batch cycle time reduction on the connected train
2–5 pts
first-pass yield improvement on the modelled grade
6–15%
reduction in steam and energy per tonne
30–50%
faster batch record closure
Target ranges for scoping a pilot. Replace with your own validated customer results before this page goes live.
