AI Fill Level Inspection for Beverage Lines: Cut Giveaway

Why Fill Targets Drift Upward

Variation, not the average, is what forces overfill.

Net content rules push plants to fill on the safe side. In the US, NIST Handbook 133 sets an average requirement: the average net quantity of a lot must at least equal the quantity on the label, and packages short by more than the maximum allowable variation count as unreasonable errors. Many other markets use similar average systems. The wider the spread of fills, the higher the target must sit to keep short fills rare. If a few valves run low, the whole filler is pushed up to compensate. The good valves then overfill on every cycle, and the plant pays for the bad ones many times over, every shift, for as long as nobody can see which valves are at fault. Our quality engineers can review how your target is set today.

How AI Fill Level Inspection Works

Measure, map to the valve, chart, act.

A camera after the filler or capper images every container. The AI finds the liquid line, even through foam, condensation or coloured glass, and converts it to a fill height or volume using a calibration for that container. Each container is tracked back to the filler valve that filled it, so every measurement lands on the right valve's chart. Because it measures at full line speed, there is no need to slow the line or pull containers for manual checks, and operators see results on a screen beside the filler. See it on a line like yours in a live demo.

Valve-Level SPC: Finding the Real Cause

One chart for the filler hides the valve that matters.

A rotary filler with dozens of valves can show a healthy overall average while one valve runs low and another runs high. Charting each valve separately shows which ones drift, how fast and since when. That turns a vague quality problem into a specific maintenance job. It also shows whether a rebuild worked, because the valve's chart should snap back into line with the others straight after the repair. Set up valve charts with our SPC specialists.

Where to Inspect on the Line

Camera position decides what you can see and fix.

Fill inspection is usually placed right after the filler and capper, before labelling, so containers can still be traced to their valve and rejected before more value is added. On some lines a second check after the labeller catches containers that leak or lose product later. Plan camera positions with our line engineers.

Records That Back Up Your Net Content Control

Every container measured means every container on record.

Auditors and customers increasingly ask how net content is controlled, not just whether a sample passed. Per-container data gives a full picture: averages by lot, rejects by valve and every adjustment made. Records are stored by lot and shift so they can be pulled up quickly when needed. Ask our quality team how records are structured.

A Worked Giveaway Calculation

Small millilitres, large money.

Take a 500 mL bottle line running 400 bottles per minute, so 192,000 bottles in an eight-hour shift. The plant sets its target two standard deviations above the label. With a spread of 3.0 mL, the target is 506 mL. After fixing the drifting valves, the spread falls to 1.5 mL and the target can move to 503 mL. Product costs $0.60 per litre. All figures are illustrative. Run your own numbers in a working session.

The two standard deviation target rule is an example only. Set your own target with your net content procedure and the rules of each market you sell into.

Giveaway by Product, Line and Shift

Know where the money goes, not just that it goes.

Once every container is measured, giveaway can be broken down any way that helps people act. Product costs differ, so a millilitre of premium juice is worth more than a millilitre of water. Shifts differ too, often because of how carefully the filler is set up after a changeover. Review the breakdown in a data session.

From Drifting Valve to Work Order

The fix is a maintenance job, so treat it as one.

When a valve drifts past its limit, the system drafts a work order with the valve number, its chart, the size of the drift and the likely cause. The planner approves it and schedules the rebuild into the next changeover or sanitation window. Until then, the system keeps rejecting under-fills from that valve, so no short containers reach customers while the job waits. Ask our support team how work orders reach your CMMS.

Fill Inspection Mistakes to Avoid

Measuring every container is only the start.

These patterns limit the value of fill inspection on many lines. A short line review can find them.

How iFactory Runs AI Fill Level Inspection

AI Vision QC, on your own network.

iFactory measures fill on every container, maps each one to its filler valve, runs SPC per valve, drafts work orders for drifting valves and reports giveaway in money per shift. It runs on an on-prem AI server beside your line. Fit questions go to our solutions team.

Results depend on your containers, products, line speed and target rules. We measure fill spread, giveaway and rejects on your own line during the pilot, rather than promising a general figure.

Turnkey AI: Delivered, Connected and Live in 6–12 Weeks

You do not build this. It arrives ready.

iFactory ships as a pre-configured NVIDIA AI server with the software pre-loaded, racked and ready. Rack it, plug in power and Ethernet, and the AI is live on your network. Our team handles cabling, network setup, PLC, SCADA, DCS and CMMS integration, filler and camera links, operator training and 24×7 remote monitoring. Data stays on your network. For a scope matched to your plant, request a turnkey quote.

Frequently Asked Questions