Case Study: STIHL achieves 99.5% inspection accuracy and significant cost and time savings with Zebra's deep learning vision solution

Machine Learning Technology Helps STIHL Cut Cost, Time and Human Error

STIHL, a century‑old global leader in power tools and the world’s top‑selling chainsaw brand, needed to replace manual visual inspection of small, variable gasoline suction‑head components with a fully automated quality‑assurance process. Human operators struggled to consistently detect subtle defects on four small footbridges per part at production speeds, so STIHL sought a machine‑vision solution that would reduce slips (bad parts passed as good), cut costs and save time.

Working with Rauscher and Zebra, STIHL deployed Aurora Design Assistant/Copilot deep‑learning tools on a Zebra 4Sight GPm controller with PoE line‑scan cameras and lighting; Zebra’s vision experts trained a CNN on 8,000 labeled images to classify each footbridge automatically at 240 images per minute. The result: hit‑rate accuracy of 99.5%, significant cost and time savings, fewer false passes, and plans already underway to roll out a second system.

STIHL

Alexander Fromm

Engineer for Automation Systems