Home / Projects / Incomplete multipacks

Case study · Juices · 6-pack

Zero incomplete multipacks, with a single recipe for every reference

At a juice plant, the shrink wrapper occasionally let through shrink-wrapped 6-packs containing only 5 packs. The distributor was complaining. The challenge was not counting: it was doing so in a plant with so many references that maintaining a recipe for each one was almost impossible.

Complaints about incomplete multipacks
Eliminated
Recipes for all references
1
Tool
Inference (deep learning)
Camera screen counting five packs in a six-pack and flagging it as NG

The challenge

Too many references for one recipe per format

The usual approach is one recipe per format: what to look at, with which tool and with what tolerance. With this customer's number of references, across sizes, flavours and designs, replicating and maintaining a recipe for each one was not viable, and every product changeover would have been an opportunity for error.

  • Shrink-wrapped 6-packs at the shrink wrapper outlet
  • One missing pack cannot be seen with the naked eye in a shrink-wrapped multipack
  • Complaints from the distributor about multipacks with missing units

The solution

One recipe, one inference tool

Instead of one recipe per reference, a single recipe for all packs with a single deep learning inference tool, trained to recognise a pack whatever its design and to count how many there are in each multipack.

Counting

Every multipack, before it leaves

The camera counts the packs in each multipack after the shrink wrapper; any that does not have 6 is rejected with its photo.

References

No recipe per product

A new reference or a change of flavour does not require creating or loading another recipe.

Evidence

Photo of every reject

Every rejected multipack is stored with its image, so you can answer the customer with data.

The result

No complaints about incomplete multipacks

Since commissioning and to this day, the distributor's complaints about multipacks with missing units have disappeared completely.

More on this check: counting in cases and multipacks, deep learning inspection and machine vision for beverages.

Free remote feasibility test

Are you receiving complaints about incomplete multipacks?

Send us photos or videos of your multipacks. We will tell you how to count them on the line, even with many references.