Skip to main content

Leveraging state-of-the-art AI technologies aiming to increase flexibility in automated quality inspection systems

In the previous article related to the Philips pilot in the STAR project; Capturing the different perspectives during an early stage of innovation development, we talked about the first stages of developing use-cases where we focused on capturing the different perspective for systems and components to be developed during the project. In this article, we are diving into one of the specific use-cases worked on in the Philips pilot during the STAR project. Specifically, the second use-case, human supervised learning for quality inspections.

Visual quality inspections are critical in the production process to ensure the delivery of correct products. A large portion of (potential) non-conformities within the Philips factory in Drachten are related to the visual appearance of parts and products. Due to the complexity (e.g. short cycle times, complex part handling, high gloss multi curved products) and related costs for (partial) automation, as well as the reconfiguration of related quality inspection systems, a lot of visual quality inspections at the Philips factory in Drachten are still a manual task.

The relatively high labor costs for these manual inspections, and the results of rather subjective quality inspections performed by the different quality inspection professionals are not ideal inputs for autonomous process control of our manufacturing processes.

Therefore, solutions are being explored within the STAR project in collaboration with the technical partners aiming to implement a system that will make setting up automated quality inspections easier & faster by applying techniques like active learning. By doing this, the goal is to develop a system that can be implemented within the factory, and that can be used to easily set-up an automated quality control for a new product. The system can learn from the quality inspection professionals on the job, and after receiving enough information about the quality of the products, the system will be able to take over the quality inspection. This way, manual inspections are only performed during the learning phase of the system.

By doing this, collaboration between human and machine is explored aiming to leverage the best of both machine and human where the human provides the machine with the required information after which the machine can take over and the human can focus its efforts on other topics.


By: Jelle Keizer, Philips