Exploring the Challenges of Manufacturing with Learning Machine Vision

Learning-based Machine Vision Tackled the Challenge

The challenge of ABB consisted of inspection of a one specific product on the automatic assembly line. They needed a system to make sure the product is assembled correctly. To make it easier, the factory already had a camera on the assembly line’s final inspection station. That camera collected the testing material needed to pilot the deep learning algorithm. Algorithms based on neuro-networks are one of the most recent trends on the field, but already familiar to Mika Saarinen:

”The development process was amazingly easy. This is a technical challenge we have been trying to solve with traditional rule-based algorithms for years, but it is not going to happen that way. The novelty of algorithms based on deep neuro-networks is that we can now solve problems that were previously very difficult.”

The solution’s Proof of Concept (PoC) is ready, waiting only for the paperwork to be completed until it can be utilised in production.

The project was called Cognex Vidi Deep Learning Machine Vision PoC. We tested learning machine vision instead of traditional machine vision in one automation cell. The co-operation with AP Vision went well and the results are promising. The learning machine vision was able to perform the inspection in this cell more reliably than previously used rule-based application. Programming the system was faster and easier than it was with the old method. We are considering expanding the utilisation of the new technology into new implementations after this project is completed. -Mikko Jukkanen, ABB

Seamless Co-operation

Reboot IoT Factory has been establishing ecosystem that creates easy co-operation among researchers, SMEs and factories. The companies have fresh networks, and chances to introduce their products and technology to the new audiences.

The project work has fitted seamlessly to AP Vision’s everyday business and there have been fruitful consequences. The company got another case, a chance to work with GE Healthcare Finland in a project that aimed to implement Cognex’s neuro-network AI camera into inspecting the coupling cables’ solders of an oxygen measurement sensor. This solution is ready and currently being used in production.