Multilinear modelling of particle size distributions

• K. Khodier ^[1] ; M. Lehner ^[1] ; R. Sarc ^[1]
  1. [1] Montanuniversität, Leoben, Austria
• Localización: Proceedings of the 8th International Workshop on Compositional Data Analysis (CoDaWork2019): Terrassa, 3-8 June, 2019 / Juan José Egozcue Rubí (aut.), Jan Graffelman (aut.), María Isabel Ortego Martínez (aut.), 2019, ISBN 978-84-947240-2-2, págs. 82-85
• Idioma: inglés
• Enlaces
  • Texto Completo Libro (pdf)
• Resumen

  • Mechanically treating mixed solid wastes, understanding how to influence particle size distributions through shredder parameters can be highly beneficial for concentrating certain materials and improving recovery rates of recyclable materials. Because of the inhomogeneity and variability of these wastes, multilinear empirical modelling is chosen as a practicable approach for describing these influences. In doing so, the compositional nature of particle size distributions needs to be considered, to obtain valid results when combining the predictions of the models for each dimension. Three potential methods for doing so were identified and analysed a priori for possible restrictions. The application of one of them (modelling the percentage of each particle size, subsequently applying the closure) was further examined with experimental data, using a linear model with two-factor interactions. It was empirically found, that distortion of the adaption to the calibration points is very high when applying model reduction on each dimension separately. Whereas, when using the same factors and interactions for each dimension, the closure becomes unnecessary, as the summation constraint is fulfilled automatically. The proof of this, as well as the calculation of confidence regions and the comparison with the other presented approaches is subject to further research.