Smart Surface Inspection: trevista® meets Deep Learning

Task:

  • Fully automated 100% surface inspection of complex metallic components with a wide range of parts
  • Highly flexible handling and automation approach
  • Possibility to use the inspection data for the purposes of Industry 4.0 or beyond quality assurance: documentation, monitoring, process optimization, production planning

Solution:

  • Smart Surface Inspection: Using trevista® in combination with innovative AI technology

Customer Benefits:

Superior image quality thanks to patented trevista® methods (Computational Imaging?) and shape from shading

  • Thanks to innovative shape-from-shading technology, trevista® images provide accurate information about surface structures and brightness.
  • High-quality result images are calculated from several input images (computational imaging)
  • Perfectly suited for glossy components up to diffusely scattering surfaces
  • Minimization of pseudo-rejects by reliable detection of errors down to the µm range
  • High part throughput due to highest inspection speed
  • trevista® domes can be flexibly positioned with regard to the distance, height, and rotational position of the inspection parts and can therefore be adapted to different types of components.
  • Universal inspection system enables comprehensive surface testing even with complex surface geometry, e.g. for gears (tooth flanks, internal toothing, helical toothing) for milling cutters, drills (cutting edges of lateral surfaces/end faces), or pistons (running surface, piston bottom).

Reliable error detection through Deep Learning

  • Neural networks can be used particularly well when a classical algorithm is difficult to find, e.g. with complex part geometry or high variance of surface properties (such as roughness).
  • By training a neural network, the inspection task can be adaptively adjusted to the current production situation at any time.

Variety of variants

  • Different component types can be taught in quickly.
  • Inspection parameters can be easily and quickly transferred to new part variants.

Best Match: trevista® and Deep Learning

  • Fast teaching in and adjustment of polygons and inspection parameters
  • Thanks to the excellent quality of the trevista® sensor data, an optimal evaluation can be carried out by the neural network.
  • Precise detection of defects on complexly shaped surfaces
  • The variety of different inspection tasks can be effectively and efficiently covered by the combination of trevista® and Deep Learning.
  • Communication interfaces are available for the extended use of inspection data, e.g. to support generic inspection plan generation, reporting, process optimization, etc.