What happened
Honeywell announced an ongoing pilot with TotalEnergies at the Port Arthur Refinery (Texas) using Experion Operations Assistant, an AI-assisted control room solution that supports operators with short-term forecasts and decision support to optimise production and enhance autonomy over time. Early results at the Delayed Coking Unit show the system forecasted five potential events with ~12 minutes lead time before alarms, helping operators minimise downtime and flare-related emissions. November 11, 2025.
Autonomy in critical infrastructure is not a sudden switch. It’s AI helps human operators, raising situational awareness to stabilize production while retaining expert control.
Why this is a smart, strategic move
- Control the Core Intelligence. By embedding AI into the Experion DCS ecosystem, Honeywell positions itself as the control-layer leader for the shift from manual control to industrial autonomy, not just dashboards, but decisioning and workflow.
- Resilience as differentiation. Refining and process industries prize uptime, safety, and emissions performance. AI-assisted control rooms that detect anomalies earlier and recommend corrective actions create measurable resilience gains (fewer trips, more stable yields).
- Built for energy transition. Energy and process assets are getting more complex (new feeds, blending, co-processing). An AI layer that continuously learns plant behaviour helps operators run closer to constraints without crossing them, improving energy intensity and emissions profiles.
Inside the AI Control Center
At its core, an AI-assisted control room acts as a human-centred cockpit for complex assets. These systems give operators by prioritizing critical alarms, surfacing early indicators of abnormal conditions, and sometimes suggesting immediate next steps. The result is faster, more confident human action.
Key advantages:
- Better situational awareness. Fusing process data with context (units, assets, permits, work orders) yields an always-on operational narrative, not just alarm floods.
- Faster anomaly detection. Models surface subtle drifts and weak signals long before thresholds trip, cutting unplanned outages and flare events.
- Reduced cognitive load. AI Agents triage, summarise, and explain what matters; LLM-powered assistants turn tribal knowledge into searchable, actionable guidance.
- More stable production. Early interventions reduce oscillations, improve product quality, and stabilise energy use across heaters, compressors, and separation units.
Human-in-the-Loop Beats Both
- Humans stay in the loop for safety-related decisions, updates to procedures, and permits, ensuring guardrails that regulators and operations teams expect.
- AI handles the heavy lifting. Continuous monitoring, anomaly detection, and contextual recommendations so operators can focus on judgment, coordination, and risk.
- Defensive depth vs. single point of failure. Instead of over-reliance on full autonomy, the assisted model delivers redundant layers: control logic, operator experience, and AI-based early warnings.
A concise look at the technology
In many AI-assisted control-room solutions, you will see integration of DCS/historian data and sometimes SCADA/VMS feeds, process-signal anomaly-detection models, and even AI-agents or LLM-based assistants that help interpret results and support operator workflows. While the Honeywell/TotalEnergies pilot is described as forecasting events ~12 minutes ahead, the publicly available details do not list all of these features explicitly
Where AI shows ROI in energy
In refining and energy assets, AI shows value when it helps spot issues such as fouling, compressor surge precursors, column flooding, pump cavitation, or heat-exchanger degradation and reduces variability and downtime. One case is the Honeywell pilot at the TotalEnergies Port Arthur refinery’s Delayed Coking Unit (DCU) reported forecasting five potential events with around a 12-minute lead time before alarms.
Beyond process control, computer vision and intelligent automation extend the value chain by enabling hot-spot monitoring, intrusion detection, vehicle/people tracking, and integrating security/surveillance functions into the control-room workflow.
DXHub AI-Assisted Control
DXHub→ is a unified AI platform for video analytics, intelligent document processing, and edge computer vision that works with your existing cameras, VMS, and control systems.
- Runs custom ML models at the edge or in the cloud and streams results into dashboards, alerts, and Automation / AI Agent workflows with role-based access and encryption.
- Ingests PDFs, scans, and office files, using OCR and LLMs to extract structure, entities, and actions that can link directly into monitoring, workflows, and AI Agents.
- Earlier anomaly detection and clearer prioritised alerts have been proven in refineries, chemicals, power, and terminals, with support for mapping requirements and scaling safely.
It’s time to work smarter
Interested in a similar setup?
Let’s do a short consult and see what’s possible with what you already have.
