Quality control goes deep
When the MEMS and micro-optic components at the heart of everyday systems fail or misbehave, the impact can be catastrophic, which makes abnormality detection a crucial task. Even automated visual inspection (AVI) machines cannot differentiate between critical and non-critical defects, and this time-sensitive task ultimately falls on the shoulders of experienced quality control managers. To optimize its inspection process and increase reliability, reproducibility, and speed by a factor of 10, our customer Axetris decided to enter the realm of deep learning, with CSEM at its side.
CSEM is experienced in using artificial intelligence, and the platform was developed very quickly compared to industrial solutions. The other surprise was its performance: certain steps of inspection and quality control can be completed up to ten times more quickly in comparison to our purely manual processes.
CSEM created a software solution that uses neural networks and machine learning to detect, diagnose, and categorize deviations. Then, deep learning enables the identification and automated assessment of arbitrarily trainable anomalies. CSEM also made it easy for Axetris to adjust the algorithms' decision criteria, adapting them to different parts and quality parameters. In use, the system has achieved near-perfect defect classification, provided insights for improving production processes, and resulted in increased customer satisfaction.