AI System for Classifying Production Scrap Records by Cause

The problem

When scrap, operator notes, alarms and product records are held in separate sources, their fields must be matched by hand before a cause class and review owner can be proposed.

The quality engineer must compare every proposed cause class and review owner with the source record; missing or conflicting evidence must not replace the final decision.

If overlapping material, setting and operator effects are not set aside for review, an incorrect cause could enter the permanent record; the record must not proceed without a human decision.

What we set out to improve

• Time taken for the draft cause class, review owner and supporting evidence to reach the quality engineer for review
• How records set aside for human review because material, setting and operator effects overlap are resolved
• Number of suggestions corrected after identifying a risk that an incorrect cause could enter the permanent record

How the system works

1. 01 Retrieve scrap, operator notes, alarms and product records from permitted sources using the record identifier
2. 02 Link source, date and version information to the shared work record
3. 03 Validate fields and units
4. 04 Prepare the proposed cause class and review owner with their supporting evidence
5. 05 Check for an exception: material, setting and operator effects overlap
6. 06 Have the quality engineer accept, revise or reject the suggestion
7. 07 Record the approved decision and its source in the processing log

Methods we used

• A source-identified common data schema for scrap, operator notes, alarms and product records
• A decision structure that separates deterministic rules from AI suggestions
• A review screen showing the proposed cause class and review owner alongside the underlying source evidence
• A mandatory human review queue when material, setting and operator effects overlap
• An event identifier that prevents duplicate processing

Where people stay involved

Data and security

Only the fields required for this decision should be processed from scrap, operator notes, alarms and product records. Access to the source systems must be authorised separately from the quality engineer role, and personal data, commercial information and technical secrets must not be passed into the model context unnecessarily. Reading, suggestions and human decisions must be logged separately.

Who this suits

Frequently asked questions