Quality management

Why wrong material still gets through: three gaps in single-point checks

One scan, one feeder comparison, alarms that do not stop the machine — these three decide the ceiling of error-proofing.

Read 4 min Author: Baichuan · Reviewed by: Technical Director

Most plants scan at loading, and wrong material still happens. Not because the scan is missing, but because it happens once, while the errors occur at the second and third loading.

Three places where error-proofing fails

A loading scan answers "is the first loading correct". In production, from kitting to running out, a reel needs re-confirming at least three times:

  • A reel change during production, that is splicing
  • Switching to a substitute after the main part runs out
  • Loading again after a changeover

Without checks at these points, the earlier scan is a one-off inspection that cannot block any of the replacements made in the following eight hours.

Gap one: scanning only at loading

Splicing is the action most easily skipped. The operator sees the tape running out, takes a new reel of the same part number and joins it. The action is quick and the scan is left out, because "the part number is the same anyway".

The part number being the same does not mean the batch or the vendor is the same. Film thickness, viscosity and metal content differ between batches, and mixing batches shifts the optimum of the reflow profile; the same part number from different vendors can also differ electrically. Automotive electronics customers usually ask specifically about splicing records during audits.

Gap two: comparing part numbers but not substitution rules

Substitution is the hard knot in error-proofing.

Validate only the main part number and every substitute raises an alarm. After raising it a few times, operators learn to bypass: they scan the substitute as if it were the main part, or simply turn the check off. The validation becomes decoration.

Loosen the check and any part can be loaded, so error-proofing fails outright.

The workable approach makes substitution a data problem: a main part number carries a list of engineering-approved substitutes, each with an enabled state and the work orders it applies to. The scan then validates "is this reel on the allowed list for this feeder" rather than "is this reel the main part number".

Substitution rules also need an owner and a review record. When engineering changes a BOM, the substitute list changes with it, and the person who changed it leaves a name.

Gap three: an alarm that does not stop the machine

The system shows a red dialog and the operator clicks OK to carry on. This is the most common failure.

For an alarm to mean anything it has to be linked to the machine: if validation fails, the mounter cannot start, or the current work order cannot pass the station. That link needs a signal path between system and equipment, either machine IO or an interface the vendor exposes.

Without that link, the fallback is to move the check forward into kitting: an order cannot go live until material is kitted and verified. That at least moves the interception point from "already mounted" to "not yet mounted".

How the three layers fit together

By interception strength and implementation cost, error-proofing usually comes in three layers:

Layer What it does What it blocks Cost
First Loading scan validates feeder and part number Wrong feeder, wrong part Scanner or lit racks
Second Splicing, reel changes and substitutes checked the same way Wrong material swapped in mid-production A check point on every action
Third A failed check locks the machine or the order Ignored alarms Equipment interface or process authority

With all three, wrong material can be pushed close to zero; with only the first, a fraction is actually blocked. The problem in most plants is not that the first layer is missing, but that it is believed to be enough.

What a smart rack adds

The weak point of scanning is the person: a person can skip the scan, scan the wrong position, or carry on after an alarm. A rack constrains the process through physical action:

  • The feeder table is issued to the rack and bins light in feeder order; taking the wrong bin leaves the light on
  • One reel per bin, so taking a reel binds it to the feeder without an extra scan
  • Substitutes are marked on the rack rather than discovered at the machine

This shows the clearest effect on lines with frequent changeovers, where material search and verification happen most. For the hardware side see the smart rack; for the process side see the SMT solution.

Frequently asked questions

Do substitutes have to be managed

Any line that uses substitutes has to. Lines with no substitutes at all are rare, usually with a stable order book and a narrow material set. Once substitution exists, without a list the only options are "alarm on everything" or "let everything through".

Does scanning slow a changeover

First loading is slower because each reel is verified. But most of a changeover is spent finding and verifying material, while the scan itself takes seconds. With the feeder table and kitting order sorted out, the whole changeover usually gets faster; see quick changeover.

What to do when alarms are ignored

Look at the alarm volume first. Nobody reads a system that raises dozens of alarms a day. Push false alarms down first, by completing the substitution rules and generating the feeder table per work order, then talk about discipline. Once alarms drop to a few per shift, operators start looking at them.

Is error-proofing suitable for high-mix, low-volume work

More so. The more models, the higher the chance of a manual verification error and the larger the benefit. The condition is a feeder table generated automatically per work order; otherwise every changeover needs manual layout, and the changeover burden arrives before the error-proofing benefit.

FAQ

Loading is scanned, so why does wrong material still happen?
The scan only covers first loading. If reel changes, splicing and substitute use during production are not validated, one scan up front cannot block the later replacements, and most errors happen in those actions.
How should substitutes be validated?
Make substitution a data problem: a main part number carries an engineering-approved substitute list with an enabled state and the work orders it applies to. Validate whether this reel is on the allowed list for the feeder, not whether it equals the main part number.
What if the operator keeps producing after the alarm?
The alarm has to be linked to the machine or the work order so production cannot start until it passes. Without the link, move the check forward into kitting and let no order go live until material is verified.
Does error-proofing slow changeovers down?
First loading is slightly slower, but changeover time is mostly spent finding and verifying material. Once the feeder table is generated per work order and kitting is issued in feeder order, total changeover time usually falls.

Related reading

Next

Send us your line's problems

Describe your line's current state and the step that most needs fixing; we'll send a solution outline per process stage within 2 business days.

Address
Building 8, Room 303, Yilisheng Industrial Park, 602 Fengtang Avenue, Tangwei Community, Fuhai Street, Bao'an District, Shenzhen

Shenzhen Baichuan Zhike Technology Co., Ltd. · Since 2015 , serving electronics manufacturers across handsets, tablets, automotive electronics, FPC, WiFi modules, energy storage and more.

Used for this conversation only; never shared externally. By submitting you agree to be contacted about this request.

Questions?