A machine learning approach to transferable Loss-in-Weight feeder mass flow prediction from equipment configuration and material properties

Tools

Mendez Torrecillas, Carlota and Jolliffe, Hikaru G. and Elkes, Richard and Reynolds, Gavin and Aroniada, Magdalini and Shier, Andrew P. and Verrier, Hugh and Fathollahi, Sara and Robertson, John (2025) A machine learning approach to transferable Loss-in-Weight feeder mass flow prediction from equipment configuration and material properties. Powder Technology, 463. 121154. ISSN 0032-5910 (https://doi.org/10.1016/j.powtec.2025.121154)

[thumbnail of Torrecillas-etal-PT-2025-A-machine-learning-approach-to-transferable-loss-in-weight-feeder]
Preview
 Text. Filename: Torrecillas-etal-PT-2025-A-machine-learning-approach-to-transferable-loss-in-weight-feeder.pdf
Accepted Author Manuscript
License: Creative Commons Attribution 4.0 logo
Download (2MB)| Preview

Abstract

The present work presents a model structure to predict Loss-in-Weight feeder mass flow performance equation parameters from material and equipment properties via three Machine Learning (ML) models using an ensemble-of-trees approach; the dataset represents 50 materials and material grades and two feeders (with multiple screws per feeder). One ML model is used for feed factor (mass/screw revolution) magnitude prediction, another for the range of feed factor decay behaviour, and a final ML model for refinement of the range to a scalar value. Feed factor magnitude is accurately predicted (test R2 of 0.94, reducing to 0.84 when inputs e.g. material properties are missing) and decay behaviour range is predicted with good accuracy (weighted F1 score of 86.4 %, and 78.6 % with missing inputs), while decay scalar refinement is challenging due to inherent variability. The present approach can be used for equipment pre-selection to determine which feeder-screw combination will likely deliver the mass flow required.

ORCID iDs

Mendez Torrecillas, Carlota ORCID logoORCID: https://orcid.org/0000-0003-3139-9432, Jolliffe, Hikaru G. ORCID logoORCID: https://orcid.org/0000-0001-5847-0177, Elkes, Richard, Reynolds, Gavin, Aroniada, Magdalini, Shier, Andrew P., Verrier, Hugh, Fathollahi, Sara and Robertson, John ORCID logoORCID: https://orcid.org/0000-0002-2191-1319;
CORE (COnnecting REpositories)