Soft Sensor Development for Real-Time Process Monitoring of Multidimensional Fractionation in Tubular Centrifuges.

High centrifugal acceleration and throughput rates of tubular centrifuges enable the solid-liquid size separation and fractionation of nanoparticles on a bench scale. Nowadays, advantageous product properties are defined by precise specifications regarding particle size and material composition. Hence, there is a demand for innovative and efficient downstream processing of complex particle suspensions. With this type of centrifuge working in a semi-continuous mode, an online observation of the separation quality is needed for optimization purposes. To analyze the composition of fines downstream of the centrifuge, a UV/vis soft sensor is developed to monitor the sorting of polymer and metal oxide nanoparticles by their size and density. By spectroscopic multi-component analysis, a measured UV/vis signal is translated into a model based prediction of the relative solids volume fraction of the fines. High signal stability and an adaptive but mandatory calibration routine enable the presented setup to accurately predict the product's composition at variable operating conditions. It is outlined how this software-based UV/vis sensor can be utilized effectively for challenging real-time process analytics in multi-component suspension processing. The setup provides insight into the underlying process dynamics and assists in optimizing the outcome of separation tasks on the nanoscale.