Deep Learning-Based Quality Assurance of Silicon Detectors in Compact Muon Solenoid Experiment

For the CMS experiment at CERN, Geneva, a substantial number of HGCAL sensor modules are manufactured at advanced laboratories across the globe. Each sensor module comprises around 675 checkpoints for visual inspection, making manual inspection practically unfeasible. In the industrial environment of manufacturing these sensor modules, this work is extremely challenging due to the various defect appearances, nonuniform, and ambiguous wire bond arrangements. Due to recent technological advances, there has been a rise in automated visual inspections and intelligent quality assurance systems in manufacturing. In order to simplify this, we propose a deep learning-based automated visual approach that processes and analyzes huge number of images for quality control and subsequent testing.