Tata Steel’s Jamshedpur facility is one of India’s oldest and largest integrated steel plants, running continuous conveyor belt systems around the clock across multiple production lines. These belts carry raw material and in-process steel across the plant floor and are critical to uninterrupted production. Any unplanned belt stoppage directly hits output, and damage to a belt itself requires hours of repair and replacement.
Prior to Vizibelt, the plant relied on periodic manual inspections and operator vigilance to catch foreign objects - metal scrap, debris, tools - that occasionally fell onto moving belts. The inspection process was reactive: by the time an object was noticed, the damage was often already done. There was no automated detection, no automatic conveyor stop, and no structured logging of incidents.
KGT Solutions became aware of a significant belt tear event that had caused serious production loss at the Jamshedpur facility. KGT approached Tata Steel with a proposal to engineer a purpose-built AI monitoring system that could prevent this class of incident going forward. The engagement began shortly after.
Foreign objects on conveyor belts — metal scrap, fallen tools, loose structural pieces — are an occupational hazard in steel plant environments. At Tata Steel Jamshedpur, the consequences of an undetected object were severe: belt tears, conveyor shutdowns, and substantial production losses. In the month preceding the Vizibelt engagement, a single belt damage event had cost the plant over $86,000 in lost production and repair costs (converted from INR at the time of reporting).
The core failure in the existing setup was detection lag. Without any automated system watching the belt continuously, foreign objects would only be noticed after the damage had already escalated — either by an operator doing a manual walkthrough or after the belt showed visible signs of tearing. By that point, the conveyor was down and the damage was done.
The secondary problem was traceability. Incident logs were maintained manually on paper — recording what appeared, which shift it occurred in, and how long the conveyor was stopped. But paper-based records made it nearly impossible to aggregate data, spot recurring patterns, or quantify the cumulative cost of these events. Without that analysis, the operations team had no structured way to build a business case for investing in prevention.
KGT designed and deployed Vizibelt - a real-time AI vision system that continuously monitors the conveyor belt through a fixed industrial camera, automatically detects any foreign object on the belt surface, triggers an immediate conveyor stop via PLC integration, and simultaneously sends multi-channel alerts to designated plant personnel. The system operates fully autonomously - no operator action is required to trigger detection or initiate a belt stop.
Mounted a high-resolution industrial camera above the active conveyor belt, configured to stream a continuous live feed of the belt surface into the Vizibelt AI detection pipeline.
Built and trained a computer vision model to identify foreign objects in real time against the steel plant's specific visual environment accounting for variable lighting, belt texture, and the range of object types likely to appear.
Integrated detection output with the plant's existing PLC system via Modbus Protocol, enabling automatic conveyor stop the moment a confirmed detection is made without requiring operator intervention.
Deployed an IIoT sensor layer providing live health telemetry (system, camera, and sensor status) with automatic red-flag alerts if any component goes offline.
Built the Vizibelt web application a control room dashboard deployed on the plant's secure internal network, accessible from any terminal on the floor. Operators get a live camera feed, detection snapshots, shift-based reports, and one-click Excel/PDF export.
Configured automated WhatsApp and SMS alerts sent instantly to designated contacts upon each detection event, plus a daily email summary of all detections from the past 24 hours.
Tech Stack:
Computer Vision (custom-trained model)
IIoT
Modbus Protocol
PLC Integration
Python
Web Dashboard
WhatsApp Business API
SMS Gateway
Excel / PDF Export
Timeline: 9–10 weeks from kickoff to live deployment. The timeline reflected approval and coordination requirements across safety, IT, and operations departments, as well as scheduled access to the active production floor for hardware installation.
The following data covers 3+ months of live operation since Vizibelt’s deployment at Tata Steel Jamshedpur on 20 February 2026.
Metric | Before KGT | After Deployment |
|---|---|---|
Unplanned Downtime Events | 3–4 per month | 0 since February 2026 |
Belt Damage Cost | $86K+ in a single month | $0 since February 2026 |
Foreign Object Incidents Detected | 0 (found only after damage) | 57+ auto-detected |
Reduction in Conveyor Downtime | — | 38% reduction |
Increase in Productivity | — | +18% improvement |
Alert Response Time | Manual inspection, hours | Automatic stop in seconds |
The following data covers 3+ months of live operation since Vizibelt's deployment at Tata Steel Jamshedpur on 20 February 2026.
Beyond the headline numbers, the operational shift has been significant. Plant operators now have a continuous, always-on eye on the belt — replacing what were previously periodic manual walkthroughs. Every detection event is automatically logged with a timestamped snapshot, conveyor stop duration, and shift information, giving plant management a full audit trail that did not previously exist.
The 57+ foreign object incidents detected since go-live represent 57+ potential belt damage events that were stopped before they escalated. At the pre-Vizibelt damage cost rate, the system has protected against a production loss exposure many times greater than the total project cost.
The next phase of the Vizibelt programme at Tata Steel Jamshedpur will extend coverage to additional conveyor belts across the facility. More significantly, KGT is developing a robotic intervention layer to complement the detection system: when a foreign object is identified on the belt, a robotic arm will automatically remove it and set it aside - eliminating the need for a floor operator to physically clear the belt before the conveyor can restart. This closes the last remaining manual step in the response loop and moves the system toward fully autonomous foreign object management.