Advisors Prof. Pratim Biswas, Member, US National Academy of Engineering; former Dean of Engineering, University of Miami Prof. Janat Shah, founding Director, IIM Udaipur
YIELDDYNAMICS

Thermo-Cognitive Advanced Process Control (TC-APC) models, simulates and optimizes your process in real time to improve yield, conversion and throughput, and reduce energy use, with a quick payback.

Deployment

It writes setpoints only after you have watched it be right.

You are being asked to let a supplier you have not worked with move setpoints on a plant you are responsible for.

So the order of the work is built around that, not around our convenience.

The deployment sequence, showing when the system writes to the plant Five stages. Through the data audit, the offsite tuning, shadow mode and validation against the laboratory, the system writes nothing to the plant. Only at the final stage does it write setpoints, in clamped increments. nothing is written to your plant writes, inside your limits 000102 0304 data audit tuned offsite shadow mode validated clamped increments Stage 02 lasts as long as you want it to. We have no interest in shortening it, because the whole argument depends on you having watched it. The deployment sequence, showing when the system writes to the plant Through the data audit, the offsite tuning, shadow mode and validation against the laboratory, nothing is written to the plant. Only at the final stage does it write setpoints, in clamped increments. 00 data audit 01 tuned offsite 02 shadow mode 03 validated 04 clamped increments nothing written so far writes, inside your limits

The five stages, and what you are agreeing to at each.

00

The data audit

You send operating data from one unit with the laboratory results beside it. We tell you how wide the margin you are holding actually is and what it is costing, with the uncertainty stated openly. Nothing is connected to anything, there is no charge, and the answer may be that there is nothing here worth pursuing.

01

Tuned offsite

TC-APC's model, simulation and optimization engines are built in. At our end, before any hardware arrives, they are customized and tuned against a simulator matched to your unit. What you are buying at this point is engineering, and you can see the model rather than just its output.

02

Shadow mode

TC-APC is installed and reads live plant data across as many batches or cycles as you want. It writes nothing. You compare its estimate against your own laboratory results for as long as it takes you to believe them or stop believing them.

03

Validated against your laboratory

Live writing is enabled only once the estimate tracks your own results inside a tolerance band. You set the tolerance, you decide when it has been met, and if it is not met the honest outcome is that it does not go live.

04

Clamped increments

Setpoints begin to move, in small steps, inside limits held in your own DCS rather than in TC-APC. Your operators see every move and the reason given for it. The fallback at every moment is the system you are running today, and going back to it is a single action.

Guardrails

The limits live in your system, not in ours.

This is the part worth being pedantic about. A supervisory layer that polices itself is asking you to trust the thing you have not yet learned to trust. The rate-of-change limits, the watchdog and the fallback setpoints are configured in your own DCS block, by your own people, and TC-APC cannot alter them.

If this happensWhat happens next
The link to TC-APC drops The heartbeat watchdog in your DCS times out and control reverts to your predefined fallback setpoints, with bumpless transfer.
The TC-APC system loses power, or is switched off Identical to the above. It stops writing, the watchdog notices, and the plant carries on under the control it had before we arrived.
The estimate goes implausible The output is blocked before it reaches your system, the condition is alarmed, and control reverts. An estimator that knows it is lost is worth more than one that does not.
A setpoint move is larger than agreed The rate-of-change clamp in your DCS truncates it. The clamp is yours and it does not ask our permission.
An instrument drifts or fails The model and the plant begin to disagree, which is itself the alarm. The affected estimate is marked unreliable rather than quietly used.
Your laboratory disagrees with the estimate The running bias is already on the operator's screen. Past your agreed tolerance, the system stops writing and waits.
An operator simply wants it off One action at the panel, or the DCS block goes to manual. No explanation required.

What arrives on site

A system beside the one you have, and nothing rewired.

The TC-APC system mounted in a rack, alongside a field HMI panel in a stainless enclosure displaying a schematic with live readings.
TC-APC system and field HMI.
  • SystemIndustrial fanless unit, rack or panel mounted, running on its own machine alongside your controls. Your existing logic is not modified.
  • Operator HMIField panel showing the live estimate, the last laboratory result, the current bias, and the reason for every move. Enclosure to suit the area.
  • InterfaceOPC‑UA or Modbus. Reads process variables, writes setpoints. Read-only throughout stages 02 and 03.
  • InstallationMinimal wiring and no shutdown. In most plants the physical work is a day, and the schedule is set by your shadow-mode period rather than by the installation.
  • Not touchedBurner management, interlocks, trips and any other safety logic. The supervisory layer writes nothing into them and cannot.
  • What stays yoursThe validated model. It remains useful for debottlenecking, operator training, what‑if studies and turnaround planning, whether or not TC-APC is writing.

What we need from you, and what we will tell you if it is missing.

Historical data

Operating data from the unit with the laboratory results that go with it. Three months is usually plenty. Without the lab results beside it, the data cannot be used to validate anything.

Instruments that are honest

Not new ones, in most cases. The measurements a unit already has are usually enough, but they have to be calibrated and they have to be logged. Where something is genuinely missing we will say so and say what it costs.

Your engineers in the room

This works as an engineering partnership with your process team, not as a delivery. If nobody on your side is going to look at the model, the result will be a box nobody trusts, and that is a bad outcome for both of us.

And the answer you might not want

If the data audit shows the unit is already being run close to its limit, we will tell you that and there is nothing to buy. That has to be a real possibility for the audit to be worth anything to you.

The same applies to a missing precondition. Where the problem is a data gap, a mechanical limitation or an upstream work-up rather than a control policy, we would rather say so than sell TC-APC to paper over it.

The mark

Find the optimum. Hold it while it moves.

The curve is a yield or selectivity response. The amber point is the optimum, and the dashed drop is where it sits on the operating axis.

The second half of the line is the part that matters. Finding an optimum is a study, and studies end up in drawers. Holding one while the feed, the fouling, the weather and the wear keep moving it is a control problem, and that is the work.

Stage 00 starts here

Send operating data from one unit

With the laboratory results beside it. We will tell you how wide the margin is and what it is costing, with the uncertainty stated openly. No charge, no commitment, and nothing connected to anything.

projects@yielddynamics.in Talk to IIT Bombay engineers Call +91 94230 07004Monday to Saturday, 9:00 to 19:00 IST

Or check us first

One deployment, named, with the man who signed for it

Fourteen carbonation towers, validated over six months of plant operation. The signed letter is published and the General Manager who coordinated it still takes reference calls.

Our track record

WhatsApp