While many manual tasks in industrial production are made easier by digitalization and robotics, the visual quality control of products is still often carried out by human workers - a tiring task for employees. It is therefore not surprising that defective products sometimes 'slip through' during visual inspection. If the deficiencies of a product then come to light at a later stage, this can lead to costly failures and result in time-consuming rework. At the same time, there is high competitive pressure on the market. This is why alternative production and inspection options are in vogue, such as the 'Kitov One' system developed by Israeli manufacturer Kitov for automated visual inspection. This robotic system combines inspection and image processing technology, robotics and artificial intelligence in one device and inspects, for example, plastics, 1D and 2D barcodes, labels (OCR, OCV), screws, connectors and connections. The system checks whether components such as (THT or through-hole technology) parts are present and have been installed correctly. It also performs a simple surface inspection of all visible surfaces of a product. A product to be inspected can be moved to the 'Kitov One' using mobile or stationary robotics.

The 'Kitov One' solves inspection tasks independently. The picture shows the 'target' - 'actual' comparison of a missing screw.

© ATEcare

The test results and images generated with the inspection system can be integrated into reports, exported and transferred to long-term storage using various archiving solutions. It is also possible to merge 2D images created during production at other workstations.

Can be integrated into existing production lines

Difficult for the human eye to recognize: The 'target' - 'actual' comparison of contacts.

© ATEcare

The inspection system, which is based on standard components, can be used wherever complete products are manufactured. It can be integrated into existing production processes and adapted to the requirements of different applications - both for intermediate inspection during assembly and for the final inspection of a finished product. The device, which is designed for a range of 100 μm, detects tiny structures that are difficult or impossible for the human eye to see. This makes the 3D system suitable for monotonous quality checks that previously had to be carried out manually.

Self-learning technology

As soon as programming has been completed using a default product, the device can automatically inspect a wide variety of products. The device currently inspects products with a maximum height of 80 cm, a maximum circumference of 80 cm and a maximum weight of 40 kg. A version that can inspect larger items is also planned.

Olaf Römer is Managing Director of ATEcare Service in Aichach.

© ATEcare

In order for the system to recognize the ideal distances of all side views and the top view, the operator must enter the external dimensions or the 3D CAD data of a product in the menu.

Based on this information, a 3D model is created and the product to be inspected is checked using AI-based data. Initially, the system shows the operator any deviations. The operator then decides whether an irregularity is okay, acceptable or definitely a defect. At the same time, the robot learns how to classify the displayed differences. Once the learning phase is complete, the robot imitates the human classification and categorizes deviations as independently as possible.