Automating Serial Number Recognition for Cold Electronics in the DUNE Far Detectors Using AI-Driven OCR Tools

The Deep Underground Neutrino Experiment (DUNE) is a cutting-edge, large-scale project aiming to answer fundamental questions about neutrino properties and the universe's evolution. One critical component of DUNE's infrastructure is its cold electronics system, which relies on a vast number of components operating at cryogenic temperatures. Accurate tracking of these, which includes knowing their serial numbers and test results, is essential not only for data integrity and system reliability but also to ensure precise identification and traceability throughout DUNE's entire lifetime.
In this presentation, we will explore how AI-based Optical Character Recognition (OCR) tools, such as GPT-4o, Google Cloud Vision and MiniCPM, are being used to automate serial number extraction from DUNE's cold electronics, particularly their application-specific chips. Given the scale of DUNE, automating this process is essential to improve efficiency, reduce human error, and guarantee reliable chip tracking for such large-scale deployment. This talk will demonstrate how these tools address challenges like image quality variability and potential inconsistent serial number readings, and how they enhance the management of DUNE's cold electronics inventory.