Gas induction in agitated vessels with turbine impellers can be modelled accurately by means of radial basis function neural nets. The results obtained with the radial basis neural net were significantly better than those obtained by multivariate regression models or standard back propagation neural nets. Moreover, by using the radial basis function neural net model, it was possible to conduct a sensitivity analysis of the variables affecting aeration. Increased medium density showed a strong adverse effect, while variation of the viscosity can cause an increase or a decrease in the rate of aeration, depending on the prevailing process conditions.
L'induction de gaz dans des reservoirs agites avec des turbines peut ětre modelisee avec precision au moyen de reseaux neuronaux a fonction de base radiale. Les resultats obtenus avec le reseau neuronal de base radiale sont significativement meilleurs que ceux obtenus au moyen des modeles de regression multivaries ou par les reseaux neuronaux a retropropagation standards. En outre, grǎce au modele de reseau neuronal a fonction de base radiale, il a ete possible de mener une analyse de sensibilite des variables qui influent sur l'aeration. On note un fort effet adverse quand la densite du milieu augmente, tandis que la variation de la viscosite peut entraǐner une augmentation ou une diminution de la vitesse d'aeration, selon les conditions de procede qui s'appliquent.
[1]
John Howard Perry,et al.
Chemical Engineers' Handbook
,
1934
.
[2]
R. H. Myers,et al.
Probability and Statistics for Engineers and Scientists
,
1978
.
[3]
R. C. Weast.
CRC Handbook of Chemistry and Physics
,
1973
.
[4]
D. A. White,et al.
Rates of induced aeration in agitated vessels
,
1977
.
[5]
Jyeshtharaj B. Joshi,et al.
Mass transfer and hydrodynamic characteristics of the Denver type of flotation cells
,
1981
.
[6]
Richard P. Lippmann,et al.
An introduction to computing with neural nets
,
1987
.
[7]
J. Friedman.
Multivariate adaptive regression splines
,
1990
.
[8]
J. A. Leonard,et al.
Radial basis function networks for classifying process faults
,
1991,
IEEE Control Systems.
[9]
Lyle H. Ungar,et al.
Using radial basis functions to approximate a function and its error bounds
,
1992,
IEEE Trans. Neural Networks.
[10]
C. Aldrich,et al.
Observations on the effect of medium density and viscosity on the rate of induced aeration in agitated vessels
,
1994
.