Motion stage precision prediction for photonic integrated circuit assembly

Abstract The consistently growing demand for robust automated Photonic Integrated Circuits assembly, testing and packaging, is increasingly oriented towards high volume and continuously sets newer challenges to overcome concerning throughput and cost effectiveness. Production processes’ intrinsic complexity, combined with short product life cycle and the necessity of quickly ramping up those to high volume, requires smarter solutions to guarantee high yield as well as low cycle time. Robust production demands for motion systems capable to realize repeated movements with precision and resolution in the range of tens of nanometers. These constraints on precision do not allow to operate the system at its highest overall speed; ideal working conditions are thereby preserved by slowing down motion, ending up trading cycle time for precision. Finally, the optimal trade-off between motion speed and repeatability is also expected to depend on hardware conditions and its optimization is therefore impossible without scheduling downtime and performing long evaluation processes. In this paper it is presented a solution for predicting linear stages’ motion inaccuracies from controller features by means of Machine Learning and Deep Learning modeling. The proposed formulation introduces a metric for calculating motion analytical imprecision that includes only the difference between successive position measurements, thus allowing a separation of short term repeatability from other error terms by removing the mean from the evaluation. Successive differences are interpreted as single-motions’ expected errors that can be aggregated into a repeatability estimate, serving as target distribution for the learning problem; predictions of single-motion metrics ensure the proposed approach to work in production scenarios when non identical movements are performed, opening up the possibility to realize advanced control paradigms and predictive maintenance for smart manufacturing.