Blast furnace gas leaks are among the most dangerous and most under-detected hazards in integrated steelmaking. BFG contains 20–27% carbon monoxide by volume — a concentration that causes loss of consciousness within minutes at even partial atmospheric contamination — yet the gas is colourless, nearly odourless, and routinely present in blast furnace casthouse areas at sub-IDLH concentrations that fixed detector grids cannot reliably capture due to sensor spacing limitations. In one documented incident at a major Indian steel plant, 13 workers were hospitalised from BFG inhalation in an area that had passed a manual gas test 45 minutes earlier. OxMaint's IoT Sensor Integration connects AI-powered gas detection networks to CMMS maintenance workflows — so when a CO sensor detects an anomaly pattern consistent with a developing BFG leak, a safety work order is generated, the area supervisor is notified, and the affected zone is logged for maintenance investigation — all within 90 seconds of detection. This case study covers how a 2.6 MTPA integrated steel plant deployed AI gas detection across its blast furnace complex to reduce CO exposure incidents by 89% and eliminate manual gas patrol gaps that had persisted for years.
Gas Leak Detection AI for Blast Furnace Areas
How AI-powered CO, SO2, and BFG sensor networks connected to OxMaint eliminated manual gas patrol gaps and reduced blast furnace gas exposure incidents by 89% across a 2.6 MTPA integrated steel facility.
The Problem: Why Manual Gas Patrols Were Failing the Blast Furnace Complex
The facility operated two blast furnaces with a combined casthouse area of 14,000 m². Manual gas patrols were conducted every 2 hours by safety personnel carrying portable multi-gas detectors. Between patrol cycles, the casthouse, cast floor tuyere fronts, bleeder valve platforms, and gas cleaning plant catwalks had no continuous atmospheric monitoring. This is the temporal gap where 90% of exposure incidents occur — not during patrols, but between them.
A BFG leak developing from a failed stove valve packing or a cracked gas main wall can reach dangerous concentrations in 15–20 minutes. A 2-hour patrol cycle provides no protection against fast-developing leaks in any area between patrol visits.
Existing fixed CO detectors were spaced at 15m intervals — standard per OSHA specification but insufficient for an open casthouse structure where airflow patterns from hot-metal runners and cooling systems created unpredictable gas dispersion. 40% of the casthouse area had no fixed sensor within detection range for low-level leaks.
Existing fixed detectors generated audible and control room alarms but created no maintenance work orders. A CO alarm that cleared within 5 minutes left no traceable record and no investigation trigger — allowing intermittent leak sources to persist undetected for weeks across successive alarm-and-clear cycles.
During tap-to-tap intervals when casthouse activity was low, gas patrol frequency dropped informally — particularly on night shifts. 7 of the 11 CO exposure incidents in the 18 months prior to deployment occurred between 22:00 and 06:00.
The Solution: AI Gas Detection Network Integrated with OxMaint CMMS
OxMaint's IoT Sensor Integration connected 64 new wireless electrochemical CO sensors and 12 pellistor-type LEL sensors across the blast furnace complex to an AI anomaly detection layer. The AI model was trained on 18 months of historical sensor data to distinguish three conditions: normal BFG background, maintenance-activity transient, and developing leak signature — enabling alert suppression for known-benign events while providing high-sensitivity detection of genuine leak patterns.
64 wireless CO sensors and 12 LEL sensors deployed at 5m intervals across high-risk zones: tuyere front, cast floor, bleeder valve platform, stove top, and gas cleaning plant. Sensor data transmitted every 15 seconds to OxMaint IoT integration layer via MQTT protocol.
The AI engine analyses multi-sensor readings simultaneously, classifying each event against three signatures: normal background (no action), operational transient (log only), or developing leak pattern (alert). False positive rate reduced from 34% (raw threshold alarms) to under 3% after 30-day training calibration.
Confirmed leak pattern generates a P1 safety work order in OxMaint within 90 seconds — including sensor location, detected gas type and concentration trend, affected zone, and recommended initial response action. Work order routed to area safety officer and shift supervisor simultaneously.
Area supervisor receives mobile alert with zone map showing sensor readings and affected area boundary. Evacuation decision and manual portable gas survey initiated. All readings from the manual survey entered into the OxMaint work order for trending and root cause analysis.
OxMaint work order tracks leak source investigation, repair action, and post-repair gas clearance confirmation. Every event — from first sensor alert to work order closure — is stored in the blast furnace gas main asset record, building a leak history that enables predictive maintenance of gas infrastructure.
See AI Gas Detection Integrated with CMMS Work Orders Live
Book a 30-minute walkthrough and see how OxMaint connects your blast furnace gas sensor network to automatic safety work orders — with alert-to-dispatch in under 90 seconds and full leak history per asset.
Results: 12-Month Measured Outcomes Post-Deployment
Results tracked across the full blast furnace complex — two blast furnaces, associated stoves, gas cleaning plant, and casthouse — over 12 months post-deployment, versus the prior 18-month baseline period at the same facility.
| Metric | Baseline (Manual Patrol) | Post-OxMaint AI | Improvement |
|---|---|---|---|
| CO exposure incidents (recordable) | 9 per year | 1 per year | −89% |
| Mean time to detect gas leak | 74 min (patrol-dependent) | 18.5 min (AI pattern detection) | 4x faster |
| False alarm rate (CO detector) | 34% of raw alarms | 2.8% after AI filtering | −92% false positives |
| Gas leak source identified and repaired | 41% of events (balance untraced) | 94% with OxMaint work order trail | +53 percentage points |
| Regulatory compliance score (safety audit) | 64/100 | 96/100 | +32 points |
| Safety work orders auto-generated per month | 0 (alarm-only, no CMMS link) | Avg. 14 per month | Complete CMMS integration |
Before vs After — Key Safety Performance Indicators
Expert Review
The temporal gap in manual gas patrol programmes is not a resourcing failure — it is a fundamental limitation of the method. No patrol frequency that is operationally sustainable can match the speed at which a BFG leak from a failed stove combustion chamber connection or a cracked gas main branch can develop to dangerous concentrations in an open casthouse. The AI approach with OxMaint integration changes the paradigm in two ways: first, it achieves continuous coverage that no patrol programme can replicate; second, it creates a maintenance response record for every anomaly, so the leak sources that were previously invisible — the intermittent valve weepers, the small-bore branch fittings with developing cracks — accumulate enough event history to be identified and prioritised for repair before they become major incidents. The false alarm reduction from 34% to under 3% was the critical commercial enabler; it meant that safety personnel and supervisors responded to every alert rather than habituating to alarm fatigue.
Frequently Asked Questions
AI Gas Detection. Automatic Safety Work Orders. 90-Second Alert-to-Dispatch.
OxMaint connects your blast furnace gas sensor network to an AI anomaly detection layer and CMMS maintenance workflows — closing the temporal gap that manual patrol programmes leave open. Most blast furnace gas monitoring integrations are live within 2 weeks.
