Deep Learning Advances Your Gear Inspection

Gears provide the drive mechanism in a wide variety of products and are therefore a key component in many applications. The functionality and reliability of entire products and systems often depend on them, and they must meet the highest standards. Ensuring quality is an equally significant challenge.

Together with our partners i-mation and Cognex, we have developed a fully automated inspection machine that provides you with greater reliability and a significant time advantage when inspecting your gears.

We help ensure that gears can be shipped free of defects and without any visible damage. Our machine detects any deviation from the standard and gets better every day—completely automatically!

To date, there have been two main problems with conventional optical inspection of tooth flanks.

• Strong contrasts are needed to detect errors
• Defects must be known and specified in advance in order to be detected and classified as defects and, for example, trained using boundary patterns.

Together, we developed a solution:

• Special lighting for the tooth flanks
• Image analysis using a deep learning approach to evaluate the appearance of the surface in order to detect deviations from trained “good” parts. The advantage of this approach is that the system is trained using “good” parts; defects do not need to be known in advance, as the deviation is detected.
• A manually loaded metrology machine or an automated in-process solution within the manufacturing facility, suitable for gears or gear shafts.

What does the machine recognize?

• Pressure marks
• Scratch
• Cavity
• Damage
• Grinding errors
• Discoloration [optional]

In cooperation with i-mation and Cognex, ROTH developed a modular cell concept that can be flexibly adapted to the needs of your quality assurance process.

The benefits at a glance:

• Interfaces to upstream and downstream systems
• Integration with ERP, MES, and other higher-level systems.
• Can be automated (robot, gantry, manual loading)
• Modern UI with an integrated, web-enabled dashboard