Why it matters
Without it, inconsistent readings, drift in sensors, or delayed maintenance can lead to contaminated fills, unseen moisture, incorrect test reports, and safety incidents.
From design to safety
Design
Designing AI-guided air quality criteria
AI defines data requirements, sensor placements, and sampling frequencies to meet EN 12021 limits from the outset, aligning with lab report templates and audit checkpoints. It specifies six sensor channels, a 10-minute sampling cadence, a 60-second data latency target, and a standardized lab report header with batch_id, sample_time, and technician_id.
- Breathing air compressor
- Filtration system
- Oil mist filter
- Moisture separator
Build
Installing sensors and data pipelines
Infrastructure installed sensors, sampling lines, and a data bridge to the analytics platform are wired into lab report generation and retention workflows. It specifies shielded coaxial cabling, FTDI adapters for legacy sensors, MQTT publishing every 15 seconds, and redundant storage with 1-day rolling backups.
- Breathing air compressor
- Filtration system
- Air dryer
- CO2 sensor
Operation
Continuous monitoring during fills
Real-time readings feed dashboards, alerts, and scheduled reports; operators verify flags against lab paperwork for batch approvals, ensuring traceability from intake to release. It enforces a daily average check, max-per-batch review, and a 2-hour window for lab sample submission after batch release.
- Breathing air compressor
- Filtration system
- Oil mist filter
- CO2 sensor
Maintenance
Lifecycle data drives service scheduling
AI tracks sensor drift, filter loading, moisture separator performance, and desiccant status to plan replacements per EN 12021 compatibility and internal QA. It generates maintenance windows, logs calibration events, and tracks spare parts consumption with minimum stock alerts.
- Breathing air compressor
- Filtration system
- Air dryer
- CO2 sensor
Safety
Safety-critical validation and records
AI-generated reports feed emergency readiness, O2 kit checks, and validated lab reports signed by the qualified person, with archival to the QA file. It ensures document versioning, EAP alignment, and 7-year retention for critical test records.
- Breathing air compressor
- Filtration system
- Oil mist filter
- CO2 sensor
Part by part
- 1
Breathing air compressor
- Watch for
- Rising vibration, unusual temperature, or pressure fluctuation indicating wear or misalignment.
- AI action
- AI flags abnormal vibration and triggers service reminder to compressor technician.
- 2
Filtration system
- Watch for
- Filter differential pressure exceeding spec, breakthrough of contaminants, or oil carryover.
- AI action
- AI generates a filter change ticket and updates EN 12021 compliance log.
- 3
Oil mist filter
- Watch for
- Oil mist breakthrough or coating of downstream filters.
- AI action
- AI issues maintenance alert and schedules filter replacement.
- 4
Moisture separator
- Watch for
- Moisture saturation or condensate carryover into dried air.
- AI action
- AI alarms moisture breach and initiates condensate purge and dryer validation.
- 5
Air dryer
- Watch for
- Desiccant saturation, pressure drop, or temperature spike.
- AI action
- AI schedules desiccant replacement and logs dryer performance in lab reports.
- 6
CO2 sensor
- Watch for
- CO2 excursion beyond EN 12021 limits or sensor drift.
- AI action
- AI triggers immediate alert, cross-check with lab sample results, and flags batch for reanalysis.
Worked example
Catching a CO2 drift before a batch fill
Illustrative scenario with sample data.
An AI monitor detects gradual CO2 rise during a weekend fill session, risking non-compliance.
Signal in
timestamp: 2026-11-07 18:45, sensor_id: CO2-07, value_ppm: 550, threshold_ppm: 500, ambient_temp_C: 23.2
System out
alert:CO2_exceedance; action: initiate lab recheck; batch_id: B1123; recommended_followup: verify sensor calibration and run EN 12021 compliant sample.
Human sign-off: Dive centre supervisor reviews the alert, confirms sensor calibration, and approves sending a retest sample to the lab.
How to roll it out
Phase 1
Weeks 1 to 2
Baseline readings established, labels created, and data routes secured for all critical sensors.
Phase 2
Weeks 3 to 6
Automated alerts and lab report integration tested with sample fills and simulated deviations.
Phase 3
Weeks 7 to 8
Full deployment with training, maintenance schedules, and documented QA checks completed.
Where people stay in charge
- Decisions on oxygen compatibility, emergency action, and batch release must be signed off by the Diving Supervisor or a qualified person.
- AI-generated lab reports must be reviewed and archived by a trained Safety Officer before formal release.
- Sensor replacement intervals and calibration must be approved by the Equipment Technician after field tests.
Questions owners and crews ask
What data is required to start AI air quality analysis in a dive centre?
You need real-time readings from CO, CO2, oil mist, moisture, differential pressure across filters, and EN 12021 lab reports to anchor AI thresholds and audits.
Who signs off the lab reports generated by AI analysis?
A qualified person, typically the Dive Centre Manager or Safety Officer, validates lab reports and signs off on batch releases.
How does AI handle sensor drift or calibration issues?
AI flags drift, records calibration events, and schedules retraining or re-calibration with the lab to maintain traceability.
What maintenance triggers come from AI?
Alerts for filter changes, dryer validation, condensate purge, and compressor service are generated when readings exceed thresholds.