Multiphase flow detection with photonic crystals and deep learning

Abstract

AbstractMultiphase flows are ubiquitous in industrial settings. It is often necessary to characterize these fluid mixtures in support of process optimization. Unfortunately, existing commercial technologies often fail to provide frequent, accurate, and cost-efficient data necessary to enable process optimization. Here we show a new physics-based concept and testing with lab and field prototypes leveraging photonic crystals for real-time characterization of multiphase flows. In particular, low power (~1 mW) microwave transmission through photonic crystals filled with fluid mixtures may be interrogated by deep learning analysis techniques to provide a fast and accurate characterization of phase fraction and flow morphology. Moreover when these flow characteristics are known, the flow rate is accurately inferred from the differential pressure necessary for the flow to pass through the photonic crystal. This insight provides a basis to develop a unique class of inexpensive, accurate, and convenient techniques to characterize multiphase flows.