One model of your plant.
One model every team can check. Advisory in the cloud, closed loop in the plant.
Map. Model. Control.
Read the transcript
Map is where Imubit starts. Every plant runs to a plan, built on economic and operational assumptions. If the plan lives in multiple places, Map pieces them together. Then Map looks at how the plant really runs, across the whole facility, and finds the gaps between the two.
It traces each gap to its causes, and to the recurring decisions behind it. It ties every objective, concern and constraint to those decisions, to the disturbances and the measured outcomes in the plant, and to the people who own each one. What looked like a monolith becomes a map you can act on: which gaps matter most, and which decisions to take on first.
Every gap, traced to a decision.
- The gaps between plan and plant, across the whole facility
- The recurring decisions behind each gap
- Who owns each decision, and where to start
Read the transcript
Model is the cause and effect behind the map. It shows how each decision moves each outcome, against its objective and its constraints. These relationships are nonlinear and dynamic. They change over time, and plant data is never complete, so data alone is not enough.
First principles matter, but on their own they fall short too. A first-principles or kinetic model is tuned on cleaned, averaged data, so it starts from an average plant, not the plant you run. And the averaging smooths away the dynamics, how the unit really responds minute by minute. Imubit's learning process models use both.
First principles are built in, and nothing sits in a black box. Every relationship is visible, causal rather than correlated, and deterministic. A Deep Learning Predictor learns each unit's real dynamics from years of its own data, selected by your engineers. And with DLPC® Builder, your process engineers build models of their own units, without Python or a machine-learning background.
Your engineers can make their own models and create their own competitive advantage.
What happens if we push?
- Cause and effect your engineers can check
- Models that hold where the data runs out
- Models your engineers build and own
Read the transcript
Control is where the decisions get executed. AI controllers take the key recurring decisions behind your gaps, and steer them around the clock, minute by minute, to maximize your objectives inside your constraints, through your existing APC or DCS. Reinforcement learning trains them across every regime the unit has run, millions of simulations and thousands of simulated years, before anything reaches the plant. What ships to the plant is a deterministic controller your engineers can certify and validate.
It does not hallucinate. Given the same inputs, it gives the same move, to every person, every time. Closed loop always runs inside the envelope your engineers set. And every gap you close is measured, continuously, against the plan.
Closed loop, inside your limits.
- The key recurring decisions executed, around the clock
- Deterministic controllers your engineers certify
- Always inside the envelope your engineers set
Each step builds on the one before. The gaps Map finds become the models your engineers build, and those same models are what the controller learns on. Nothing is thrown away, so the value adds up step by step.
A learning layer on top of what you run.
Each layer of plant automation does its job on its own clock. Imubit adds a learning layer that writes setpoints to your existing APC or DCS every minute. Everything else keeps its role.
Sets targets on a weekly or monthly cycle, on assumptions that age between cycles.
A learning layer. It learns each unit from its own data and writes setpoints to your APC or DCS every minute, inside the limits your engineers set.
Coordinates interacting variables against constraints, typically with linear models from commissioning step tests that drift as feed, fouling and catalyst activity change.
Holds setpoints second to second. It does not decide which setpoint is economically best right now.
Field devices, safety systems and operators all keep their roles.
Writes setpoints downReads historian data up
Why historian data is hard to learn from.
Pattern-finding misses four things. The Deep Learning Predictor learns a nonlinear, dynamic model of each unit from years of its own history, with mass and energy balances and the known physics shaping the model where the data runs thin.
Why the model moved.
How each input moved the predicted product yield over the test period. All input movements count toward each contribution.
Build a model of your own unit.
Import the data, auto train, validate. A process engineer does it without Python or a machine-learning background.
Ask your plant model in plain language.
Use the language models your teams already work with to reach your Imubit models through the Model Context Protocol (MCP). The language model asks. The deterministic model answers, the same way every time.
- Ask what a change would do before you make it.
- Analyze the results and see why the model moved.
- Build models of your own units, and your own analyses and workflows on them.
Part of the platform, not an add-on. Where a site does not use language models, everything works in the platform itself.
See a model built this wayYield rises with feed until the outlet temperature reaches its limit, then levels off. Past that point, more feed costs energy for no gain.
Outlet temperature limitLearn. Train. Run.
Deep Learning Predictor
Learns the unit from its own history.
Deep Learning Controller
Reinforcement learning, millions of simulations and thousands of simulated years.
Closed loop
Deterministic, inside limits you certify.
Engineers set the economic objectives and the safety and operating limits the controller has to respect. Scenarios are tested in DLPC® Sim App before go-live, and the same plant state produces the same move every time, so engineers can certify it and operators can anticipate it.
From open loop to closed loop.
Choose the units and handles behind the highest-value gaps.
Model each unit in DLPC® Builder and review it in DLPC® Explorer.
Run operating scenarios in DLPC® Sim App and see the result of a change before making it.
The model guides daily decisions. Teams compare predicted and observed behavior and adjust constraints as conditions change.
Once the model has earned operators' trust, the controller trained against it goes live, writing to the APC or DCS within explicit constraints.
DLPC® Performance Dashboard tracks model accuracy and plant performance. Closed gaps are measured against the plan, and engineers retrain models as the plant changes.
Going from open loop to closed loop needs no new tools and no new architecture. The controller is trained on the same model your engineers built and tested.
Runs inside your plant network.
Running in the plant
- Runs on-premise, inside the plant network, through the On-Premise Application.
- Enforces explicit, configured constraints.
- Scope can be limited to selected handles or a single unit.
- Operators keep full authority, with a direct path to take the controller out of service.
Built for production
- Native integration with historians and with DCS and APC environments.
- Version control and an audit trail for every model.
- Model and data governance, with role-based access.
- Performance visibility across assets.
What your engineers use.
Find out what your gap to plan is costing you.
One to two hours for your team to prepare and upload the data package
Results within a few working days of approval
No workshops
Send your data as it is. Gaps, outages and noisy tags are normal, and our engineers screen and clean them in the platform. If anything limits a use case, we tell you.
