How Hexion Cut Chemical Spill Response Time by 80% with AI-Powered Detection

Client Profile

Industry: Chemical Manufacturing — Thermoset Resins Size: 4,000+ employees globally Operations: 20+ countries across North America, Europe, and Asia-Pacific Products: Epoxy resins, phenolic resins, coating resins, and specialty adhesives

Hexion Inc. is a global leader in thermoset resins and specialty chemicals, supplying critical materials to industries including aerospace, automotive, construction, and electronics. Their production facilities handle a wide range of reactive chemical intermediates and finished resins, operating under strict environmental, health, and safety standards across every site.

The Challenge

Chemical spills in resin production environments carry consequences that extend well beyond the immediate safety risk. Epoxy and phenolic resin intermediates are hazardous to workers on contact, classified as environmental contaminants, and subject to strict EPA and OSHA reporting requirements. Even minor leaks from pipelines, vessel connections, or filling stations can escalate rapidly if not contained within minutes.

Hexion's existing approach relied on manual rounds by safety personnel and operator visual inspection — processes that left significant gaps, particularly during night shifts and in areas with limited foot traffic. Key pain points included:

• Slow detection on off-shifts: Small drips and seepage at pipe connections routinely went unnoticed for 30–90 minutes, allowing pooling to spread across production floors.
• Escalation risk: Undetected minor leaks frequently became reportable spill events requiring full hazmat cleanup, production shutdown, and regulatory notification.
• Environmental compliance pressure: EPA Tier II reporting obligations meant that even incidents categorized as "near-miss" generated significant administrative burden and increased regulatory scrutiny.
• High remediation costs: Average cleanup cost per containment breach exceeded $40,000 when factoring in labor, materials, downtime, and third-party waste disposal.

Hexion needed a detection approach that was continuous, consistent across all shifts, and capable of identifying the earliest signs of a spill before escalation.

The Solution

Hexion partnered with Visionify to deploy an AI-powered spill and leak detection system across their resin production facilities. The system uses computer vision models trained specifically on chemical manufacturing environments to identify:

• Liquid pooling on production floors and containment areas
• Drips and active seepage from overhead pipe runs and vessel connections
• Overflow conditions at filling and transfer stations
• Tank farm surface accumulation near bund walls

Detection events trigger real-time alerts delivered simultaneously to on-floor operators via mobile devices and to the central control room dashboard. Each alert includes a timestamped camera capture, location identifier, and severity classification, enabling rapid triage without requiring a physical walk to the affected area first.

All alert data feeds directly into Hexion's existing incident management system, creating an automatic audit trail for regulatory compliance and internal safety reporting.

Implementation

Visionify's deployment team completed site surveys across Hexion's primary resin production buildings, tank farms, and drum-filling areas, identifying 47 high-priority monitoring zones based on historical incident data and process risk assessments.

Cameras were positioned at:

• Pipe rack junctions and flange connections with known leak history
• Reactor vessel drain points and overflow channels
• Drum and IBC filling stations
• Tank farm bund perimeters
• Sump and secondary containment areas

Edge processing units installed at each production zone handle inference locally, eliminating network latency and ensuring alerts fire within seconds of a detected event — without routing video through cloud infrastructure. The full deployment, including camera installation, model configuration, and integration with the incident management platform, was completed in three weeks with no disruption to ongoing production.

The detection models were calibrated to Hexion's specific environment, accounting for steam venting near reactors, condensation on cold transfer lines, and the reflective surface characteristics of wet epoxy resin — conditions that would generate false positives in a generic vision system.

Results

Within the first six months of operation, Hexion's facilities recorded measurable improvements across every key safety and operational metric:

• 80% reduction in spill response time — from an average of 47 minutes to under 10 minutes, measured across all monitored zones
• 92% of detected spill events contained before escalation — prevented from becoming reportable incidents or requiring full hazmat-level remediation
• $1.2M in estimated annual savings — combining avoided cleanup costs, reduced regulatory penalties, lower insurance claim frequency, and production downtime prevention
• 340% ROI — achieved within the first year of deployment

Night-shift incident detection improved most significantly, with after-hours spill response times dropping by 87% compared to the pre-deployment baseline.

What's Next

Following the success of the spill and leak detection program, Hexion is expanding their use of Visionify's platform to additional environmental monitoring applications — including vapor release detection near solvent handling areas and effluent monitoring at wastewater discharge points. The incident data captured over the first deployment period is also being used to drive targeted infrastructure maintenance, with leak-prone flange connections and aging pipe segments flagged for priority inspection during planned shutdowns.