Multiple‐Output Artificial Neural Network to Estimate Solid Cycle Times in Conical Spouted Beds

Knowledge of cycle times is essential to design spouted beds. Although experiments for determining cycle times are time‐consuming, there is hardly any mechanistic or empirical model for their prediction. Two types of artificial neural networks, namely, single‐ and multiple‐output ones, have been dev...

Full description

Saved in:
Bibliographic Details
Published in:Chemical engineering & technology 2021-03, Vol.44 (3), p.542-550
Main Authors: Estiati, Idoia, Atxutegi, Aitor, Altzibar, Haritz, Freire, Fabio B., Aguado, Roberto, Olazar, Martin
Format: Article
Language:English
Subjects:
Artificial neural networks
Conical spouted beds
Cycle time
Cycle times
Empirical analysis
Modeling
Neural networks
Parameter sensitivity
Spouted beds
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Knowledge of cycle times is essential to design spouted beds. Although experiments for determining cycle times are time‐consuming, there is hardly any mechanistic or empirical model for their prediction. Two types of artificial neural networks, namely, single‐ and multiple‐output ones, have been developed to estimate particle cycle times in conical spouted beds. They provide satisfactory cycle time predictions under different configurations. A comparison of single‐output neural networks with multiple‐output ones proves that, although the former fit slightly better the experimental data, the latter provide a reasonable prediction of all cycle times, and therefore, only one neural network is sufficient for their prediction. The maximum cycle is the one of highest sensitivity to all the parameters analyzed. Knowledge of the average cycle time is crucial when designing the reactor or improving its efficiency, whereas information on the maximum and minimum cycle times is required in catalytic reactions under fast deactivation. Promising results are provided for the estimation of the cycle times through artificial neural networks, allowing the progress in the scaling up of the conical spouted beds.
ISSN:0930-7516
1521-4125
DOI:10.1002/ceat.202000491