Why it matters
Without AI, inconsistent berry selection, hidden MOG, and calibration drift cause rejected lots, longer sorting times, and variable juice quality across vintages.
From design to safety
Design
AI-informed layout and QC plan
AI maps sorting line capacity, sensor placements, and calibration routines from berry profiles, harvest forecasts, equipment specs, and compliance checkpoints to guide layout, workflow, and QC checklists.
- vibrating sorting table
- optical sorter
- conveyor belt
- camera sensor array
Build
Install sensors and calibrations
Data streams from tone/vibration sensors, color imaging, weight, and size measurements, plus reference berries with known Brix and acidity, establish baseline thresholds and maintenance triggers in SOPs.
- vibrating sorting table
- optical sorter
- conveyor belt
- calibration weights
Operation
Real-time sorting and QC monitoring
AI reads berry metrics, flags out-of-spec fruit, logs MOG events, and triggers automatic rejects to the rejects chute while updating yield, RAs, and traceability across the batch.
- optical sorter
- conveyor belt
- rejects chute
- camera sensor array
Maintenance
Continuous calibration and cleaning
Periodic re-calibration against reference berries, vibration checks, camera cleaning, lens decontamination, and drift audits to prevent performance loss and ensure sorting accuracy across vintages.
- camera sensor array
- optical sorter
- vibrating sorting table
- rejection bin
Safety
Traceability and safety auditing
All sorting decisions linked to batch records with timestamps, operator IDs, calibration logs, and QA notes to satisfy HACCP, appellation controls, and regulator inspections.
- calibration weights
- rejects chute
- conveyor belt
- camera sensor array
Part by part
- 1
vibrating sorting table
- Watch for
- Wear or misalignment of vibration pad causing inconsistent berry lift and mis-sorting.
- AI action
- Detect vibration anomaly and alert maintenance; pause line if threshold exceeded.
- 2
optical sorter
- Watch for
- Color sensor drift or glare causing false positives in berry rejection.
- AI action
- Trigger recalibration or halt line until camera gains are restored.
- 3
conveyor belt
- Watch for
- Slippage or belt wear reducing berry conveyance velocity and picking accuracy.
- AI action
- Adjust speed and log wear for preventive maintenance.
- 4
camera sensor array
- Watch for
- Dust or condensation reducing image clarity and MOG detection accuracy.
- AI action
- Run cleaning cycle and re-validate focus matrix.
- 5
rejects chute
- Watch for
- Blockage or misrouting leading to berry loss or cross-contamination.
- AI action
- Redirect flow and alert operator to clear chute.
- 6
calibration weights
- Watch for
- Drift in reference standard mass or temperature affecting thresholds.
- AI action
- Trigger recalibration and update threshold library.
Worked example
Catching a shielded berry leak before a batch moves
Illustrative scenario with sample data.
A subtle shift in berry color distribution indicates potential underripe fruit entering the optical sorter queue.
Signal in
berry_mass_g: 1.8-2.4, color_index: 0.65, size_mm: 6.2-7.1, MOG_proportion: 0.03, temperature_C: 18.5, brix: 21.4, YAN: 0.95, harvest_block: A3, lot_id: L20261108-09, timestamp: 2026-11-08T09:15Z
System out
pass/fail decision per berry, with total rejects 5.2% of queue; color drift flagged; lens contamination alert; line speed adjustment recommended; QC notes generated and batch disposition recorded for traceability.
Human sign-off: Winemaker and QC lead review drift report, approve recalibration, confirm calibration interval, and authorize a temporary line speed reduction pending full audit.
How to roll it out
Phase 1
Weeks 1 to 2
Baseline data collection, install sensors, and validate pass/fail criteria across typical harvest lots.
Phase 2
Weeks 3 to 6
Tune AI thresholds, calibrate cameras, and establish standard operating procedures for sorting and rejects handling.
Phase 3
Weeks 7 to 12
Run concurrent audits, generate traceability records, and finalize dataset for end-to-end quality checks.
Where people stay in charge
- Critical decisions about pass/fail and lot disposition must be signed by the Winemaker or QC Manager with batch records.
- Calibration and threshold changes require approval by Vineyard Manager and Compliance Officer to protect appellation integrity.
- All data retention and privacy policies must be documented and reviewed by Estate Legal Counsel before deployment.
Questions owners and crews ask
What signals does the AI watch on the sorting table and optical sorter?
The AI monitors berry size, color indices, MOG presence, weight, and surface defects, then decides pass/fail and logs actions for traceability across the full batch lifecycle.
Who approves changes to sorting thresholds and calibration?
The Winemaker and QC Manager sign off on threshold updates, calibration changes, and batch disposition after reviewing the audit trail and SOP records.
How does this system handle recalls or segregation of batches?
All movements are tagged with batch IDs; if a flag triggers, the system redirects fruit, archives logs, and notifies the cellar team for immediate containment and traceability.
What records are created for compliance and tasting room audits?
Batch records include sensor readings, timestamps, operator IDs, calibration history, disposition actions, and QA notes to satisfy HACCP and appellation rules, with traceable vintner signatures.