A computer vision sensor for the parallelization of actively regulated capillary slug flow microreactors

Abstract

Abstract In this work, a computer vision sensor for the extraction of slug length, slug velocity and phase ratio from capillary liquid–liquid slug flows from video feeds in real-time, including the necessary post-processing algorithms, is developed. The developed sensor is shown to be capable of simultaneously monitoring multiple capillaries and provides reasonable accuracy at less than 3.5% mean relative error. Subsequently, the sensor is used for the control of a parallelized and actively regulated dual-channel slug flow capillary microreactor setup. As a model reaction, the solvent-free epoxidation of methyl oleate with hydrogen peroxide and a phase-transfer catalyst based on tungstophosphoric acid and a quaternary ammonium salt to yield the product 9,10-epoxystearic acid methyl ester is conducted. A space–time yield of 0.679 kg L−1 h−1 is achieved.

Keywords

• Slug flow
• Flow sensor
• Machine learning
• Video imaging
• Computer vision
• Epoxidation