Process Intelligence for Plant Operations

A physics-grounded AI agent platform that continuously sanitizes, maps, and analyzes Distributed Control System data to advise panel operators on energy-optimal targets and warn of process deviations hours before alarms trigger.

95% - 100% Diagnostic Accuracy
~35 Min Median Early Warning
LIC PIC FIC TIC VI SIC

The Industrial Data Challenge

Modern processing plants generate millions of telemetry records daily. Most of this data is only accessed to perform failure analysis after the fact. We utilize this readily available stream of information to detect faults early, advise operator action, and optimize plant conditions.

The Status Quo

  • Data Underutilization: 99% of captured DCS readings are stored in historians and only reviewed retroactively to analyze a failure.
  • Cognitive Strain: Panel operators manage massive workloads. Synthesizing dozens of parallel trends in real time strains human cognitive limits.
  • Delayed Action: Alarms are triggered only after operational limits are crossed, leaving tight windows to avoid emergency trips.
Unplanned Downtime Cost: Average of $1.2M - $20M+ per plant.

The AI Agent Approach

  • Continuous Synthesis: Automatically absorbs and standardizes client tags, standardizing sensor telemetry into standard units for evaluation.
  • Physics-Grounded Diagnostics: Merges machine learning with thermodynamic & hydraulic digital twin baselines to foresee deviations.
  • Early Warning Window: Flags emerging anomalies within ~35 minutes of inception, giving operators ample time to prevent trips.
Up to 85% reduction in escalations from minor process upsets.

Validated Process Units

The AI Agent is shipped with physics-first, plant-agnostic design basis configurations, enabling immediate integration on Low-Pressure Storage and Heat Exchange assets.

Storage & Transfer

Storage Tank & Pumps

Demineralized water vessel with blanket nitrogen gas padding, overpressure safety systems, and multiple duty/standby transfer pumps.

  • 9 Critical Faults: Loss of feed, overfill, padding valve stuck, padding gas loss, PVRV leak, heater loss, sensor freezes (LIC/PIC).
  • Pump Health: Monitors bearing wear (ISO 10816/API 670 vibration & temperatures) across up to 4 parallel pumps.
  • Staging Optimization: Advises on the lowest-energy staging configuration to hold the production target.
Thermal Systems

Air-Cooled Heat Exchangers

Process fin-fan cooling bay cooling a hot single-phase stream by drawing ambient air via multiple axial fans.

  • 8 Target Faults: Tube-side fouling, air-side fin fouling, tube plugging, fan belt slip, motor overload, motor cooling loss, blade imbalance, bearing wear.
  • Dimensionless Features: Normalizes ratios against design basis to ensure identical fault signatures on any size cooling bay.
  • VFD Speed Tuning: Identifies over-cooling conditions and recommends VFD throttling, saving fan horsepower.

Double-Blind Verification

To validate performance without model overfitting, the AI Agent prototype was tested blind against 30 days of synthetic DCS logs generated by an independent physics simulator.

95%

Diagnostic Accuracy

Correctly categorized fault families and rotating-equipment bearing wear modes across all blind runs.

35m

Median Detection Delta

Detected and classified anomalies in a median of 0.58 hours (~35 minutes) from initial fault event onset.

The Cross-Examination Protocol

1
Chaotic Data Generation

Simulator generated 30 days of 60s logs with injected sensor dropouts, random noise, and non-standard tagging.

2
Blind Analysis

The AI Agent analyzed the logs without knowing the locations, durations, or types of faults injected.

3
Automated Scoring

An independent Scorer compared the Agent’s reports against the hidden Ground Truth to calculate accuracy and speed.

Live Telemetry Transformation Pipeline

Observe raw telemetry logs stream from the plant simulator on the left, and watch the AI process core output standard, validated operator advisories on the right.

DCS Live Sensor Trend (Inputs)
Current Temp
Active Advisory Asset: HX_201 (Exchanger)
Processing

HX_201 Fan Belt Slip

The AI agent has detected a fan belt slip event on cooling fan bay HX_201. Fan rotational speed has dropped, motor stator current has unloaded, and the bundle cooling approach has risen.

Approach Temp
+14.2 °C
Motor Current
-18.4 A
Confidence
97.6%
Action Targets:
  • Schedule fan belt tension checks on motor bay HX_201.
  • Throttle standby fan speed targets to compensate for the cooling loss.

Designed for Air-Gapped Industrial Networks

Industrial facilities operate under strict data safety bounds. We bypass the traditional cloud compliance bottlenecks entirely.

On-Premises edge server deployment

The agent is installed on client-approved local servers or hypervisors within the facility's control network. Zero process data leaves your firewall.

No Remote Connection Required

Operational logic updates are packaged and updated locally. The agent operates in complete isolation, minimizing cyber-attack vectors.

Corporate Data Integrity

No external APIs, no public cloud endpoints, and no outbound tunnels. Total ownership of your plant data remains in-house.

Ahmed Sherief

Ahmed Sherief

Founder & CEO

Calgary, Alberta, Canada

Engineered by Operational Experience

"With process safety at the forefront, we designed Supreme Cortex to work from the control room, not from a remote cloud server. By combining on-the-ground operational experience with edge-computed physics, we empower panel operators and support engineers with real-time, actionable diagnostics to keep systems running within optimal operational boundaries."

20 Years in Process Operations

Led global plant commissioning, dynamic start-up, field operations, and control room (DCS) troubleshooting across downstream refineries, LNG plants, and midstream processes.

Optimization & Debottlenecking

Specialized in plant start-ups, control loop troubleshooting, equipment debottlenecking, and process performance optimization based on chemical engineering principles.

Request Technical Spec & Demo

Discuss deploying the AI Agent in "Shadow Mode" on your historians, or request detailed physics simulator reports.

Email Inquiries

info@supreme-cortex.ca

Headquarters

Calgary, Alberta, Canada

Demo Request Received

Thank you for your inquiry. A technical engineer from our team will contact you shortly at the provided work email with details regarding "Shadow Mode" evaluation configurations.