Ultrasonic spectrum for particle concentration measurement in multicomponent suspensions

Abstract

This paper studies the feasibility of applying the ultrasonic spectrum technique to the measurement of particle concentrations in multicomponent suspensions. A combination of the kernel partial least squares (KPLS) model and the interval selection methods is implemented to build the relationship between the ultrasonic spectra of the first reflected pulses and the particle concentrations. First of all, the interval selection methods are used to select optimal spectral interval(s) from full spectra. Then, the KPLS models with optimal spectral interval(s) are tuned, built and evaluated to obtain the optimal model. Finally, the optimal KPLS model is employed to measure the particle concentrations in the mixing process and its online prediction ability is evaluated. In comparison with the linear partial least squares (PLS) models, the optimal KPLS model shows the best performance. The results demonstrate that particle concentrations in multicomponent suspensions can be measured online by the ultrasonic spectrum technique, and the KPLS model with optimal spectral interval(s) shows the superiority in model calibration.