Fractionation of ultrafine particles: Evaluation of separation efficiency by UV–vis spectroscopy

• Published in: Chemical engineering science 2020-02, Vol.213, p.115374, Article 115374
• Main Authors: Winkler, Marvin, Sonner, Heiko, Gleiss, Marco, Nirschl, Hermann
• Format: Article
• Language: English
• Subjects: Centrifugation; Colloidal particles; Fractionation; Machine learning; Solid-liquid separation; UV–vis spectroscopy
• ISSN: 0009-2509; 1873-4405
• DOI: 10.1016/j.ces.2019.115374

[Display omitted] •Tubular centrifuges enable bench scale fractionation of colloidal particles in single or multi component systems.•Machine learning algorithm predicts solids volume concentration based on spectroscopic sensor data.•New insights into the process monitoring of multidimensional property distributions are gained. Centrifugation is an established tool in solids process technology to handle classification of particulate products. However, dispersions with strict specifications for product-relevant properties often require a different approach. Fractionation is necessary when both geometric and material properties matter. Due to their high throughput and centrifugal acceleration, semi-continuous tubular centrifuges are suitable for the effective performance of such tasks. In order to monitor this mechanism for several size and density fractions, the separation efficiency is correlated with optical properties. Titanium dioxide and pigment particles serve as experimental products. Coarser size fractions are separated and overflow samples are analysed with analytical centrifugation and UV–vis spectroscopy. Each extinction spectrum is labeled with distinctive target values and a regression model is build. Finally, mixtures are processed and targets determined using the machine learning algorithm. It is discussed to what extent the advanced methodology can be applied to multidimensional unit operations such as density fractionation.