System for visual optical inspection

Artificial Intelligence takes visual control in electronics factories to a new level. Precision, speed, savings

The interior of the factory, production line. Three women in white coats sitting in one row of the production line work above a tall device - a cuboid in a white housing

Project title

Development of an innovative automatic visual control system AOI 4.0 using artificial intelligence used to test electronic components dedicated to the automotive industry.

Name of Beneficiary/Beneficiaries

FITECH Sp. z o. o.

Name of programme

SMART GROWTH OPERATIONAL PROGRAMME

Competition

Sector programme INNOMOTO

Project value

PLN 15 445 009,50

Funding value

PLN 8 689 998,34

Project delivery period

1st Feb 2017 – 31st Aug 2019

View the results of our work

Five ready-made devices, equipped with cuboid monitors in a white housing, stand next to each other in one factory room. On the front, a blue inscription: FITECH AI solutions. Perpendicular to it, the inscription on the side of each device: Smart Optical Inspection. At the bottom of the machines, an inscription: SOI Euler 020. — Factory machines - ready to be delivered to customers

What problem is addressed by the project?

The project "Development of an innovative automatic visual control system AOI 4.0 using artificial intelligence used to test electronic components dedicated to the automotive industry" solves the problem of optical inspection of components assembled in electronics factories, where there is a very high variability of production.
In factories where high - mix/low volume products are produced, the product life cycle on the production line is short, and the number of products implemented often exceeds several hundred per year. In such a case, with the classical approach to optical inspection, the factory would have to delegate many engineers solely to the process of writing programs for new products that are checked by optical testers. Our solution uses artificial intelligence, which reduces the programming process to a minimum. After starting production, photos of the products on the line are collected. After collecting the appropriate number of photos, the data is prepared for training and sent to the server. There, a fully automatic network learning process takes place, after which the tester is ready to test a new product.

Who uses the project results?

Our solution is prepared for testing printed circuit boards with electronic components. We originally focused on the automotive sector, but our device can test products in any electronics factory, regardless of its size and the industry for which the products are manufactured. Of course, the biggest savings will come in factories with a large variety of products, but large-scale manufacturers will also appreciate our solution. Since it is very cost-effective, multiple AOI 4.0 machines can be inserted in one line. Thanks to this, the manufacturer can control the production at any stage, catch errors and make corrections at a very early stage, while it is still possible - especially in the case of serial errors. Additionally, thanks to the connection with the cloud, the factory receives on-going analytical information. It shows eg. the level of errors on the line or the level of line useage. This allows for early response and continuous improvement of production processes.

What was the greatest challenge during project implementation?

The biggest challenge turned out to be the automation of the process of adding new products. The considerable variety of components and the required high efficiency of error detection needed the use of advanced neural networks to be able to efficiently and quickly handle new products in the industry with very high dynamics.