Challenge
In high-speed production lines, spotting defects or quality issues early on is crucial to avoid unnecessary waste. Therefore, you'll need a quality control system that can quickly make real-time decisions about quality and alert you if there's a problem. It would be even better if this system could gather detailed information about the defective parts, saving time and money on quality control.
Most current solutions focus on one-step processes that involve a lot of data collection, storage, and labeling—often done by experts. Unfortunately, these systems are (too) slow for fast-paced production processes.
So, with this project, we are working on a quality system that can instantly detect defects in a high-speed production process and allows us to improve the quality without a lot of manual labeling.
Project goal
The goal of this project is to develop a two-stage defect detection system that minimizes labelling effort and avoids complex data storage and management.
- The first part of our system works on the production line. Here, we use a smart detection system to differentiate normal and defective samples without the need for data labelling.
- The second part works off line. Here, we improve the defect detection by learning from the filtered data. The system focuses on understanding the type of defect and its location.
This two-step process allows for quick real-time decisions about product quality and further (offline) defect analysis to optimize the production process.
Our system (called RETINA) is edge-based and can be deployed on an embedded device, speeding up the process and eliminating the need for storing data in the cloud. Plus, it connects directly to the machines or process, giving us only the most important quality information.
Would you also like to participate in this project?
RETINA_IRVA is an industrial research project. We are looking for companies that want to be part of the user group and work with us to valorise the project.
Interested? Fill in the form below and we will contact you as soon as possible.
