Quality Prediction

100% weld coverage while reducing manual verification efforts by more than 50%.

The Challenge

A major German automotive manufacturer produces thousands of resistance spot welds across its vehicles. With such high production volumes, quality assurance relies on statistical sampling rather than inspecting every weld. Verification involves manual inspection across multiple shifts, supported by periodic destructive testing, often hours after the weld was produced.

The manufacturer wanted to identify welds at elevated risk of quality issues using the process data already generated during production, while reducing the need for costly manual quality verification and increasing inspection coverage.

Our Approach

Fraunhofer USA CMA developed and implemented an AI-based system that learns the relationship between welding process data and weld quality.

By analyzing signals such as welding current and voltage, the system identifies welds with an elevated risk of quality issues and enables quality teams to focus inspections where they are most needed. The approach also provided insights that supported improved welding-tip maintenance and equipment efficiency.

The Impact

The AI-based solution enables 100% inspection coverage of spot welds while allowing quality teams to focus manual verification on welds with an elevated risk of quality issues. This targeted approach reduced the effort of manual weld quality verification by more than 50% and achieved an ROI of less than one year.

The project demonstrates how AI can transform existing manufacturing process data into actionable quality insights, enabling comprehensive inspection while reducing the effort and cost of quality assurance.