Flow regime transition in Gas-Liquid-Solid columns identified by a Self-Organized Map applied to γ-densitometry time series
Julia Picabea1,2, Mauricio Maestri1,2,*, Miryan Cassanello1,2, Gabriel Salierno1.3,5,
Cataldo De Blasio3, María Angélica Cardona4,5,6, Daniel Hojman4,5, Héctor Somacal4,6
1 Departamento de Industrias, Facultad de Ciencias Exactas y Naturales, Universidad de Buenos Aires; Intendente Güiraldes 2160 - Ciudad Universitaria - C1428EGA - Buenos Aires, Argentina
2 CONICET-UBA–ITAPROQ; Ciudad Universitaria - C1428EGA - Buenos Aires, Argentina
3 Faculty of Science and Engineering, Åbo Akademi University, Rantakatu 2 - 65100 - Vaasa, Finland
4 Laboratorio de Diagnósticos por Radiaciones, Dep. Física Experimental, Comisión Nacional de Energía Atómica, BKNA1650 San Martín, Buenos Aires, Argentina
5 CONICET, C1033AAJ Buenos Aires, Argentina
6 Escuela de Ciencia y Tecnología, Universidad de San Martín, BKNA1650 San Martín, Argentina
Self-organizing maps are unsupervised neural networks that provide a visualization of the data in a lower dimension. In this work, the potential of this type of model as a tool to classify and identify a change in the flow regime of a three-phase bubble column is tested. Granulated carbon-air-water three-phase fluidized bed column is studied by γ-densitometry. Photon-count time series are analysed at different operative conditions. Statistical features are extracted as inputs to train the self-organizing maps. When each input data is presented to the map, a neuron is activated, giving a visual representation of the data. The γ-photon count time series can be analysed using a self-organizing map procedure to determine indexes related to the underlying flow regime. The resulting models show three different regions in the map for the homogenous, transition, and heterogeneous flow regimes. Once these regions are delimited, the map can be used for fast classification of the equipment operating conditions. The ability of the self-organizing maps to diagnose a flow transition is verified against visual observation and gas hold-up trends. The conclusions are tested for their sensitivity to alternative axial positions of the radiation source used for the densitometry. Best results were obtained with the source located at the second quartile of the column.
Keywords: Machine learning, Self-organizing maps, Gamma-densitometry, Bubble columns, Three-phase fluidization
Industrial Application: Flue gas treatment; fluidization; fermentation
Copyright © International Society for Industrial Process Tomography, 2021. All rights reserved.