A Process Analytical Technology (PAT) for Smart Chemical Manufacture
: AI-Integrated Online Molecular Profiling System
Monitoring critical quality attributes (CQA) and process parameters such as the concentration of products, raw materials, media, metabolites, cell viability, product aggregations, and viral contaminations within the manufacturing line remains a significant challenge in chemical engineering. This task necessitates a new analytical system that is sensitive, selective, and capable of multiplexing for rapid on-site monitoring. Traditional analytics including spectroscopy, mass spectrometry, chromatography, and electrophoresis, have been employed for Process Analytical Technology (PAT) in the current chemical industry. However, these methods are limited by significant delays in analysis time and require substantial space due to their reliance on offline monitoring.
We are focused on developing innovative analytics for smart PAT using online/inline sensing systems within reactors and production lines. Our approach involves the creation of multi-array microfluidic or fiber optic systems based on label-free fluorescent transducer design techniques. We design and fabricate compact, integrated fiber optic nanosensor elements capable of measuring a broad spectrum of chemical process parameters, including drugs, decontaminants, and product qualifiers. Our lab-on-fiber design, featuring 3D-printed miniaturized sensor tips with high mechanical flexibility, enables at-line monitoring. These advancements collectively represent an effective form factor for nanosensor-based transducers, facilitating applications in industrial process monitoring and aiming to maximize production efficiency.
In addition to advanced PAT hardware, we are also focused on developing innovative software for smart PAT, specifically AI-based multivariate nanosensor signal analysis. We are creating high-performance, deep learning-based real-time concentration analysis programs. These programs are designed for precision molecular mapping, high-throughput single-cell image analysis, hidden signal detection, and process parameter tracing within the baseline noise of the sensor.