Vision system for product quality control

Company description

The company operates an extensive park of lathes, milling machines and other machinery and equipment. Production encompasses steel and other metal processing as well as assembly into semi-finished and finished products. The company sells over 30,000 different items.

Problem

• Finished products and semi-finished goods consist of many different component parts — pins, bolts, nuts, cables, contactors, labels, seals, springs and others
• A great deal of time spent checking essential control points, e.g. presence and correct tightening of all nuts
• Verification of uniform coating and absence of scratches on surfaces
• No warranty and post-warranty documentation
• No archive of manufactured items with easy search by serial number
• No documentation of product assembly into cartons and packaging method before shipment to the end customer

Proposed solution

• An image analysis device using a neural network for pattern learning
• Operator login integrated with MS Active Directory or MS Entra
• Automatic switching to the selected product type based on the operator scanning a barcode from the plant register
• Automatic product entry into the dark chamber and measurement execution
• Verification of all control points:
  • Placement of labels and nameplates, and verification of print and character stamping correctness
  • Surface uniformity and colour
  • Presence of pins, springs, holes, bolts, nuts, microswitches and many other components
• Measurement archive with search functionality

System components

• Physical device — a dedicated construction for automated measurement
• Industrial automation and controller software
• Image analysis software — developed using open-source libraries
• Web application — handling all required processes and functionality via a browser, accessible without installation or licences on the required devices