Blog
Posts, case studies and news on closed-loop control in refining, chemicals and polymers, cement, and mining and minerals.
Process safety management and the role of industrial AI
How OSHA's 14 PSM elements depend on each other, where programs break down in a changing plant, and where AI helps with hazard detection, alarms and knowledge.
Imubit named a Leader in the ARC MarketMap 2026 for Autonomous Execution Agents
ARC Advisory Group positions Imubit as a Leader in the ARC MarketMap 2026: Autonomous Execution Agents, Continuous Process Edition, its study of AI that executes decisions in the plant rather than only recommending them.
How closed-loop optimization supports chemical process safety
How steadier control, earlier sight of drift and a recorded trail of moves support process safety in chemical plants, alongside the existing safeguards.
How Imubit differs from APC, and works with it
APC holds each unit to its targets within its limits. Imubit learns cause and effect across units and makes the key decisions on top of APC, inside limits set by plant engineers.
Causality and correlation in process control
Correlation is enough to predict. Acting on a running unit needs cause and effect. What the difference means for AI in process control.
Autonomy comes in steps, with experts in charge
Plants will not jump to full autonomy. It comes in steps, with domain experts, deterministic plant models and generative AI working together.
Why an agent that learned to be safe somewhere else isn't safe on your unit
Nadav Cohen's group at Tel Aviv University proved that safe execution of a task does not transfer the way execution does, so a general-purpose agent that was safe elsewhere is not, by that fact, safe on your unit.
Imubit CTO Nadav Cohen and team win the Israel AI Safety Research Prize
Imubit's co-founder and CTO Nadav Cohen and his Tel Aviv University students received the Israel AI Safety Research Prize for work published at NeurIPS 2025, selected as a Spotlight paper.
Long before AI kills us all, it will destroy the internet
Imubit's co-founder and CTO Nadav Cohen on TheMarker's podcast: why losing control of the internet is a nearer AI risk than doomsday scenarios, and why the physical world moves more carefully.
What LATAM is teaching us about AI adoption: lessons from Innovarpel 2025
Takeaways from Innovarpel 2025, where Latin American leaders treated AI adoption as a change in how teams decide and work together, and measured it that way.
What keeps industrial AI running: takeaways from the 2025 ARC Forum
Notes from the 2025 ARC Industry Forum: a refiner's closed-loop naphtha system, and a panel on data, people, safety, trust and keeping AI online.
Preem partners with Imubit to lower emissions at its Lysekil refinery
Preem and Imubit partner to run closed-loop optimization on the FCC unit at Preem's Lysekil refinery, with the aim of cutting emissions and fuel use.
Imubit opens an R&D center in Cluj, Romania
Imubit has opened an R&D center in Cluj, Romania, to develop the tools industrial teams use to build, evaluate and monitor their own process models.
ARC Advisory Group's 2024 market study covers closed-loop AI alongside APC
ARC Advisory Group's 2024 market study covers closed-loop AI next to APC and online optimization, as two U.S. refiners present closed-loop work at AFPM.
Answers to common questions about closed-loop AI control
Answers to the questions refining engineers ask most about DLPC®, from architecture and data flow to inferentials, retraining and operator trust.
Three questions to ask before you choose AI for your plant
Three questions for process industry leaders evaluating AI for plant optimization: how the model is built, who can use it, and whether it runs in closed loop.
Economic Engineering update: September 2021
Market commentary from September 2021 on record gas prices and rising EU carbon costs, and why energy now competes with yield and throughput in the plant.
Transforming safety culture from a program to an organizational principle
How Imubit's employee-led Safety Committee and short Safety Moments at the start of meetings make safety part of daily work, beyond a checklist or a rating.
July 2021 market update: the IEA oil outlook and peak gasoline
July 2021 market update: the IEA's Oil 2021 outlook saw gasoline demand peak in 2019, with growth moving to petrochemicals and to diesel outside the OECD.
June 2021 market update: RIN prices and the RVO-adjusted crack spread
June 2021 market update: record RIN prices raised the RVO, and the RVO-adjusted crack spread showed thinner refining margins than the headline crack.
April 2021 market update: an uneven recovery in U.S. fuel demand
Archive commentary from April 2021 on the EIA's demand outlook: gasoline and jet lagging 2019 levels while U.S. distillate demand recovers and grows.
March 2021 market update: the Texas freeze and the summer outlook for refining
March 2021 market update: vaccinations, the Texas freeze and heavy gasoline and diesel stock draws pointed to stronger refining margins for the summer.
January 2021 market update: FCC margins and the planning rhythms behind them
January 2021 market update: firm gasoline and diesel cracks lift FCC margins, and capturing them depends on monthly, weekly and hourly decisions.
December 2020 market update: distillate back in favor heading into winter
Archive commentary from December 2020 on the gasoline to distillate spread, the drivers behind the distillate rally, and why refiners had to stay nimble.
Complexities of applying AI in refining
Why applying AI to refining units is harder than standard analytics: skills gaps, historian compression, export limits, tag merging and model transparency.
Economic Engineering update: November 2020
Refining market commentary from November 2020 on the second COVID-19 wave, utilization, deferred turnarounds and where yield optimization still paid.
Your LP may be giving you a false sense of optimality
Why a refinery LP's optimal solution can mislead: one-off limits buried in case files, units linearized around one regime, and feed qualities out of date.
Economic Engineering update, October 2020
Archived October 2020 market commentary: distillates fall below gasoline, what that means for FCC feed and crude rates, and ethane cracking margins.
Welcome to the Imubit blog
The September 2020 note that opened the Imubit blog: why we started it, who writes for it and what we planned to share from work on operating plants.
September 2020 market update: distillate stocks build as jet demand collapses
Archive commentary from September 2020 on how the pandemic hit U.S. refining, why ULSD stocks built on the Gulf Coast, and what it meant for conversion.
You're ready for AI, but is your data?
Why imperfect plant data does not hold up the start of closed-loop optimization, and the data practices that help most over time, from historian compression and lab samples to base-layer control.
What the AFPM white paper says about adopting AI-based APC
A summary of the AFPM OPCAT white paper on adopting and maintaining AI-based APC, covering change management, explainability, AutoML and retraining.
What refining leaders told us at Transcend EU
Refining leaders at Transcend EU in Athens and Munich named their top concerns, from safety to reskilling, and what AI must do to earn their trust.
What ARC Advisory Group's 2024 study found on AI in process optimization and control
ARC Advisory Group's 2024 market study on why plants are moving to closed-loop AI for optimization and control, and how adoption differs by sector.
Tuning out the noise in the industrial AI landscape
Most industrial AI stops at advisory insight because closing the loop in a live plant is hard, and tools that depend on a person to act leave value unclaimed.
Three practices for running chemical plants through uncertainty
Three practices that help chemical plants hold margin through uncertainty: plant-wide optimization, real constraints, and models engineers can use.
The right AI solution: domain-driven or generic Industry 4.0?
Why generic Industry 4.0 AI vendors struggle in complex process plants, and how a partner that knows your industry works, from the demo to commissioning.
Renewable diesel optimization
How variable feeds, carbon-intensity credits and nonlinear yield trade-offs complicate renewable diesel units, and where closed-loop control fits.
Reflecting on Verdantix's 2025 Green Quadrant for industrial AI analytics
Imubit CEO Gilad Cohen on the Verdantix 2025 Green Quadrant for industrial AI analytics, and why closed-loop control reaches past analytics.
Oxbow runs petroleum coke calcining in closed loop
Oxbow runs its rotary kiln operations for petroleum coke calcining in closed loop with Imubit, raising yield and cutting natural gas use within specification.
Opportunities for positive ROI from AI in refinery processes
PTQ's Q1 2025 Q&A on where AI pays off in refining: distillate, FCC and coker results, and how hydrocrackers can swing yield between naphtha and diesel.
North America refining market outlook 2022
Archive commentary from 2022 on lost U.S. refining capacity, the return of fuels demand, and why a prolonged up-cycle in refining margins looked likely.
Myth busters: reinforcement learning in closed-loop control
Three common misconceptions about reinforcement learning in closed-loop process control, on first principles, prediction error and retraining, and support.
Minimizing jet fuel production in an existing refinery
Archive guide from 2020, when jet demand collapsed, on moving kerosene into the gasoline and diesel pools through crude unit, reformer and hydrocracker moves.
Keys to success with closed-loop AI in plant operations
Leaders from two U.S. refiners and Oxbow on what made closed-loop AI work in their plants, from choosing the right problem to multi-unit models.
Introducing the Imubit Foundation Process Model
Gil Cohen on the Foundation Process Model, one deep learning model of a plant that its engineers build, validate and reuse across planning, training and tuning.
Industrial AI's next phase will be won in operations
Takeaways from Transcend Houston 2026 on running industrial AI across teams, systems and shifts, and on what earns the trust of operators and engineers.
Industrial AI is maturing, and exposing the gap between executives and operations
Live polling at the AFPM Annual Meeting showed executives aligned on what AI promises, while operators judge its value at the moment of decision.
How a learned controller differs from real-time optimization
How a learned closed-loop controller differs from real-time optimization, from a model built on plant data to decisions made without a real-time solver.
How AI is changing the engineering workforce
Gil Cohen's Hydrocarbon Processing feature on the petrochemical skills gap, and five ways AI changes the work of plant engineers and operators.
Five strategic takeaways from Digitalization and AI in Mining North America 2025
Five themes from the 2025 Digitalization and AI in Mining North America event in Vancouver, and what they mean for operators turning plant data into throughput.
Energy strategy first, sustainability as a result
Energy is still one of the largest operating costs in process plants, and four tests help leaders choose energy projects with verifiable results.
Energy and carbon in the closed-loop optimization objective
How energy use and carbon intensity can enter a closed-loop optimization objective alongside margin, from a talk at a Madrid energy forum.
Conversion without transformation is the hidden risk in AI adoption
Adopting AI without changing how teams decide and work tends to give the same results faster. Why domain expertise and shared targets matter more.
Controllable AI and the future of plant modeling
Gil Cohen, writing in Control Global, on why process plants need AI models that engineers can inspect, adjust and ground in first principles.
Closed-loop control on a refinery debutanizer, FCC riser and diesel stripper
How closed-loop control worked three limits at a U.S. refinery: debutanizer flood, a static FCC riser temperature and a diesel flash point buffer.
Closed-loop AI and process control: why the public debate matters
Dennis Rohe on the trade-press debate over closed-loop AI, where it sits next to DCS and APC, and why trust and scale deserve open discussion.
AI, AFR and 1L at the 2025 IEEE-IAS/PCA Cement Conference
Takeaways from the 2025 IEEE-IAS/PCA Cement Conference on alternative fuels, limestone cement logistics, and where closed-loop AI fits at the kiln.
A world without step testing
A former APC engineer on what step tests cost operations and APC teams, and how a model learned from plant history gives a distribution of gains.
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.