Why it matters
Without it, mis-timed fermentation and improper conditions lead to under- or over-fermented leaf, inconsistent colour, rejected lots, and repeated adjustments during audits due to unclear process data.
From design to safety
Design
AI informs design tolerances
AI defines target fermentation windows, bed layout, sensor placement, and data schemas from historical leaf specs, machine capacities, and factory line constraints to ensure consistent end-points.
- fermenting beds
- CFM machine
- humidity sensor
- oxygen sensor
Build
Data-gathering integration during build
Wiring to central data hub, calibration protocols, alarm hierarchies, and ID-tagged sensors established for real-time monitoring of humidity, oxygen, color metrics, and bed temperature.
- fermenting beds
- CFM machine
- humidity sensor
- oxygen sensor
Operation
Real-time control and adjustment
Streaming readings from beds, CFM, humidity, oxygen, and color metrics feed responsive time and temperature targets; AI issues work orders for manual checks and batch re-logging.
- fermenting beds
- CFM machine
- humidity sensor
- colorimeter
Maintenance
Preserve consistency through upkeep
Model recalibration using seasonal leaf profiles, drift correction for CFM speeds, recalibration of color units, and sensor hygiene audits to maintain thresholds.
- fermenting beds
- CFM machine
- humidity sensor
- oxygen sensor
Safety
Assure safety and compliance
AI flags unsafe humidity or oxygen deviations, color drift indicating fermentation errors, and maintains an auditable log for QA and regulatory audits.
- fermenting beds
- colorimeter
- oxygen sensor
- humidity sensor
Part by part
- 1
fermenting beds
- Watch for
- Deviations in leaf colour progress, hot spots, or bed temperature gradients during fermentation.
- AI action
- Reads bed colour and temperature trajectories; triggers time adjustment and batch tagging; stores batch-end colour units for QA traceability.
- 2
CFM machine
- Watch for
- Unstable air flow, unexpected pressure rise, or inconsistent exhaust temperatures leading to uneven fermentation.
- AI action
- Monitors airflow and exhaust metrics; adjusts fan speed, damper position, and predicts end-point through trend analysis.
- 3
humidity sensor
- Watch for
- Humidity drift outside target range affecting enzymatic activity or microbial balance.
- AI action
- Continuously records RH; recalibrates fermentation timelines and humidity setpoints; logs humidity excursions with timestamped batch IDs.
- 4
oxygen sensor
- Watch for
- O2 levels too high or low, risking uneven fermentation and colour drift.
- AI action
- Tracks O2; modifies aeration rate, bed turning schedule, and records end-point confidence for each batch.
- 5
colorimeter
- Watch for
- Color change lag or overshoot indicating under- or over-fermentation.
- AI action
- Measures color units; flags deviations, updates end-point estimate, and stores two-reference colour indices per batch.
- 6
exhaust damper
- Watch for
- Damper sticking or misalignment causing airflow imbalance.
- AI action
- Sends maintenance alert and records flow discrepancies; triggers preventive maintenance order.
Worked example
Catching a seal leak before a charter
Illustrative scenario with sample data.
Prevent a fermentation-season discipline failure by early detection of a humidity spike impacting end-point accuracy.
Signal in
estates_humidity_%RH:72, bed_temp_C:42, oxygen_ppm:14, color_unit:18, cfm_speed_rpm:320, batch_id: B1262
System out
alert: humidity spike detected; batch_B1262 flagged; auto-adjustment: reduce humidity target to 68% and extend monitoring window; maintenance requested for duct seal and damper check
Human sign-off: Factory supervisor approves adjustment, verifies color trajectory, and logs maintenance work order with batch traceability.
How to roll it out
Phase 1
Weeks 1 to 4: Baseline
Install sensors, connect data hub, define alarm thresholds, and run pilot batches to refine end-point targets and traceability rules.
Phase 2
Weeks 5 to 12: Tuning
Calibrate sensors against lab colour measurements; finalize aiDecision rules, end-point confidence, and alert routing across teams with batch IDs.
Phase 3
Weeks 13 and beyond: Stabilization
Full-scale operation with continuous learning, regular audits, and documented traceability for certification, including Rainforest Alliance or other schemes.
Where people stay in charge
- All final fermentation decisions must be signed off by the factory manager based on QA and safety constraints.
- Sensor data must be retained for each batch with per-batch IDs and audit trails.
- Only qualified personnel may override AI decisions during safety-critical adjustments.
Questions owners and crews ask
What data does AI require to control fermentation time and temperature?
AI requires real-time readings from fermenting beds, humidity sensors, oxygen sensors, CFM machine variables, and colorimetric measurements to determine optimal time and conditions, with per-batch identifiers for traceability.
Who reviews AI decisions during the fermentation phase?
The factory manager, QA supervisor, and senior tea tasters jointly review AI-issued adjustments and end-points before implementation, with sign-off on batch records.
How does AI handle safety during fermentation?
AI raises alarms for unsafe humidity, oxygen levels, or color drift and logs actions for audit and certification, ensuring containment and documented corrective actions.
Can AI improve consistency across lots?
Yes, by aligning sensor thresholds and end-point targets across batches, reducing variation in colour and degree of fermentation while maintaining traceability.