Optimizing Terahertz Waveform Selection of a Pharmaceutical Film Coating Process Using Recurrent Network

In-line terahertz pulsed imaging has been utilized to measure the film coating thickness of individual tablets during process in a production-scale pan coater. A criteria-based waveform selection algorithm (WSA) was developed select signals reflected from surface and determine thickness. Since WSA uses many criteria thresholds waveforms sufficiently high quality, it could reject some potential candidate tablet that are close but do not reach threshold boundary. On premise availability large datasets, we aim improve efficiency with machine learning. This article presents recurrent neural network approach optimize selection. In comparison conventional method WSA, our allows more than double number be selected while maintaining excellent agreement offline measurements. Moreover, processing time decreases so can applied for real-time monitoring pharmaceutical industry, which leads advancements quality control coating.