Standard of Excellence: Building Excellence From the Inside Out
Powering the Future: Why True Ceramic Circuits Are Not Just ‘Better PCBs’
Global Sourcing Spotlight: The Global Quality Gap—Why Consistency Wins the Contract
Artificial intelligence models work better with more data. While individual EMS companies can certainly create plenty of data over time, the broader the data set, the more insightful the AI results can be. Ben Rachinger, a research assistant at Friedrich-Alexander-Universität Erlangen-Nürnberg, received the NextGen Best Paper at IPC APEX EXPO 2025. His research asks: What if a model could be created that allowed industry-wide data in the model, while still protecting proprietary information?
Nolan Johnson: Ben, tell me about the award you received for your paper at IPC APEX EXPO.
Ben Rachinger: I received the NextGen Best Paper Award for my paper that focuses on verifying whether a collaborative machine learning approach called federated learning can be effectively applied for improving THD image classification. By using data from multiple companies and training a machine learning model, we can build an image classification model based on a broader knowledge base. In our use case, the input will be images, and the output is used to determine whether this image shows a good or defective solder joint.
Johnson: There's a lot of discussion about AI language models in a particular application. Are you specifically looking at image models?
Rachinger: Not necessarily. I am focusing on federated learning, a collaborative learning approach where multiple companies collaborate. It is applicable to every type of machine learning. It's potentially usable for classification on numerical data and on image data. It's also possible to do language models, of course.
To read this entire article, which appeared in the May 2025 issue of SMT007 Magazine, click here.
