On a high-speed packaging line, scrap is made by the minute and discovered by the shift. Rejects drop into a bin, a count is written at handover and the plant learns about a seal problem or a coding fault hours after it began. By then the cost is not one bad pack but several hundred, plus the rework and the arguments about when it started. Real-time scrap detection moves discovery from the end of the shift to the first few minutes by capturing rejects at the line, classifying them at once and alerting on patterns. FMCG teams that want to see their own reject data handled this way can watch scrap events classified live on a working line feed in a short session.
Know About the Scrap While the Line Is Still Making It
iFactory AI captures every reject at the line, classifies it in seconds and alerts the supervisor when a pattern starts, so losses stop within minutes.
Four Ways to Capture Scrap at the Line
Detection is only as good as capture. Most plants combine two or three methods, because each sees something the others miss.
| Method | What It Captures | Strength | Weakness |
|---|---|---|---|
| Reject counters from the PLC | Count and time of each reject | Reliable, no extra hardware | No reason for the reject |
| Vision at the reject chute | Image and defect type | Classifies automatically | Needs lighting and set-up |
| Check-weigher data | Weight, over or under | Quantifies giveaway and shortfall | Sees weight defects only |
| Operator tap on a screen | Reason chosen from a short list | Adds human judgement | Depends on discipline |
A Taxonomy Short Enough to Use
Classification fails when the list is long. Four families, with a handful of types each, cover most FMCG lines.
Work Out Where Your Scrap Is Captured, and Where It Is Not
Share your reject points and categories. iFactory AI will map what each line captures today and what a reason-coded feed would add.
How Much Scrap Is Made Before Someone Knows
Detection time sets the size of a loss. The bars compare an illustrative line making 120 packs a minute with an extra 1.5 percent defect rate from the moment a fault begins.
Pattern Alerts Instead of Single Rejects
One reject is normal. A run of the same reject is a signal. The strip shows twelve ten-minute periods for one reject type, and teams can test alert rules on their own reject history before any alert goes live.
What a Useful Alert Says
An alert that only says scrap is high sends the supervisor looking. One that names the line, the type and the likely cause sends them to the fix.
A Composite Scenario: A Splice Problem Found in Minutes
A film splice began feeding badly after a reel change. The figures below are illustrative.
Where iFactory AI Fits
Reject counters, check-weighers, vision systems and operator taps each hold part of the picture. iFactory AI joins them by line and time.
Line-Side Capture
Rejects from PLCs, vision, weighers and operator taps arrive in one feed.
Instant Classification
Each reject is typed and linked to line state, product and material lot.
Pattern Alerts
Runs and rate jumps trigger alerts with the likely cause attached.
Loss Costing
Scrap is valued in material, time and rework, so priorities are clear.
Expert Review: Making Scrap Detection Stick
Detection technology is the easy part. This review sets out how experienced operations leaders judge whether a scrap programme will last.
Frequently Asked Questions
Do we need new cameras on every line?
Not necessarily. Many plants start with the reject counters and check-weighers they already have, and add vision only where a reason cannot be captured any other way. The first step is a review of what each line records today. Our team can run that review before any hardware decision is made.
How are false alerts avoided?
By comparing each line with its own normal for the same product and speed, and by requiring a pattern, not a single reject. Rules are tested on history, so the expected number of alerts is known before go-live. Alert quality is tracked, and rules that fire too often or too late are tuned.
What if operators will not enter reasons?
Keep it to one tap from a short list, and use automatic capture wherever possible. Reasons entered by operators are useful, but they should not be the only source. Sharing the results back with the crews helps, because they see their taps turn into fixes. See a one-tap capture screen in a short walkthrough.
Can it separate start-up waste from true defects?
Yes. Line state tells the system when a line is starting up, changing over or running steadily, so rejects can be grouped accordingly. That keeps expected start-up waste from hiding real problems. It also makes changeover waste visible, which is often an improvement opportunity in its own right.
How is the benefit measured?
By comparing scrap rate, time to detect and material cost per thousand packs before and after, normalised for product and speed. All three come from the same records. Ask for a sample benefit report that finance and operations can both use.
Stop Reading About Yesterday's Scrap
See rejects captured, classified and alerted on your own lines while the pattern is still small. Book a working session with an FMCG loss specialist.
