REFINERIES

AI technology for the most complex refinery processes

Deep Learning Process Control® can optimize the most complicated – and profitable – processes at your refinery

Are you looking for a competitive edge in today’s low-margin refining market? Whether your feedstock is relatively consistent or varies wildly, your economic and operation teams make weekly, daily and hourly adjustments to each unit in an effort to optimize each process. But there are some processes that are just too hard to optimize with the current tools available to you.

Today’s market presents so many ever-changing factors, from varying feedstock composition and equipment disturbances, to regulatory and environmental variations – it’s become impossible to capture all the dynamic relationships with your current models. That’s exactly why we built Imubit – the first AI technology developed specifically for oil refineries and other hydrocarbon processors.

Our process optimization artificial intelligence (AI) technology – helps refiners break through these optimization challenges. It’s not a generic AI solution targeting all industries, but next-generation artificial intelligence that focuses on the nuances of how oil refineries operate and can be run in closed-loop driven by planners with full operator controllability.

Featured deep learning application for refineries

Over the last few years, we have developed 6 types of strategic refining applications and field proven them with our tier-1 refining clients. These applications are uniquely tailored to our Deep Learning Process Control® technology, and we consider them highly proprietary.

Request access to our strategic DLPC refining application catalog >

Learn about how Imubit’s AI technology helps refineries discover, engineer and monetize new facility margin opportunities:

Generalized first-principle economic models

for key chemical processes like blending, fractionation, conversion and reforming. The models consider unit constraints, operation modes, feeds, products, and fundamental behaviors of process units.

Steady-state baseline models

that estimate in real time the potential benefit from engaging closed-loop DLPC control. Once DLPC is commissioned in closed-loop, we use the baseline models to assess the value it created over time as well as the potential loss if it were disengaged.

Performance dashboards

let you track the unique KPIs for each AI application, perform economic debottlenecking and analyze constraints. Our dashboards help you devise strategies to adapt to feedstock, equipment and global or regional economic changes.

Process modeling platform

that leverages your process understanding, historical and ongoing data to analyze your process and regulatory control and train your deep learning prediction models.

Deep learning process models

capture the hidden governing dynamics between variables in all process states and model the relationships between feed properties, key process variables, operational constraints, and economic objectives.

Dynamic relationships visualization

of Monte Carlo simulations on the trained models show the learnt relationships between process model variables as well as model prediction errors, so your process engineers can see and understand how the model works.

Pre-optimized truly dynamic controller

is designed, trained and continuously improving based on your manipulated variables, constraints, objective function, and real-time stream pricing.

Open-loop simulations

let your team visualize the dynamic controller moves and predictions in every desired historical period, run what-if simulations for various operating scenarios, and fully validate the controller’s behavior before commissioning the controller in closed-loop.

Process control network software

is installed, configured and continually supported to run your controller inside your air-gapped secure environment. Our on-premise software is fully compliant to process control network cybersecurity and reliability requirements and supports all major DCS vendors.

Control room application

gives operators the ability to directly update controller constraints and move limits, while engineers can update constraint operator ranges and control priorities.

Customizable dashboards

provide chart and matrix views for operators to understand controller status, projected moves, and prediction trajectories. Multi-unit control room dashboards show data across all controllers, all units and every process in one place.

Remote monitoring

via our cyber-secure proprietary network protocols enables those not in the control room to run safe cloud-based simulations and analyses on your production data. No data is ever transmitted into your control network.
Imubit’s team understands what we’re trying to do from an operating and an economics perspective. Their Deep Learning Process Control® technology and their approach have aligned everyone from planning and operations around a shared goal of maximizing production profits.”

-Process control and optimization manager
Tier-1 refinery

Ready to learn more about Deep Learning Process Control®?