Neural network particle sizing in slurries by reflectance spectroscopy

Authors

  • Mingzhong Li,

    1. Centre for Molecular and Interface Engineering, Dept. of Mechanical and Chemical Engineering, Heriot-Watt University, Edinburgh EH 14 4AS, U.K.
    Search for more papers by this author
  • Derek Wilkinson,

    Corresponding author
    1. Centre for Molecular and Interface Engineering, Dept. of Mechanical and Chemical Engineering, Heriot-Watt University, Edinburgh EH 14 4AS, U.K.
    • Centre for Molecular and Interface Engineering, Dept. of Mechanical and Chemical Engineering, Heriot-Watt University, Edinburgh EH 14 4AS, U.K.
    Search for more papers by this author
  • Markus Schrödl

    1. Centre for Molecular and Interface Engineering, Dept. of Mechanical and Chemical Engineering, Heriot-Watt University, Edinburgh EH 14 4AS, U.K.
    Current affiliation:
    1. University of Applied Sciences, Amberg-Weiden, Germany
    Search for more papers by this author

Abstract

Measuring concentration and size of solids in suspension is important in many industries. Even though techniques based on optical transmission measurements have been well developed, they are not always successful in practical applications because low concentration suspensions are needed. A method developed determines particle-size distribution and concentration from reflection measurements in concentrated suspensions using neural networks with particle concentrations up to 10% volume fraction. Based on measured optical reflectance spectra of suspensions with known particle-size distributions and concentrations, a neural network was trained to identify particle-size distribution and volume fraction of suspensions. Training is a time-consuming process requiring presentation of many spectra and their corresponding particle-size distributions and volume fractions to the neural network, but once concluded satisfactorily, the neural network can be used to predict the particle-size distribution and volume fraction of high concentration suspensions rapidly in-situ.

Ancillary