Comments On “Neural Network Modeling of Structured Packing Height

phase pressure gradient (ΔP/H) in counter-current gas- liquid structured packing towers. The proposed tools are shown to outperform, in terms of predi...
0 downloads 0 Views 15KB Size
Ind. Eng. Chem. Res. 2000, 39, 4437

4437

Comments On “Neural Network Modeling of Structured Packing Height Equivalent to a Theoretical Plate” and“HETP and Pressure Drop Prediction for Structured Packing Distillation Columns Using a Neural Network” Faı1c¸ al Larachi* and Bernard P. A. Grandjean Department of Chemical Engineering, Laval University, Sainte-Foy, Que´ bec, Canada G1K 7P4

Sir: The two research notes1,2 recently published in Ind. Eng. Chem. Res. by Eldridge and co-workers describe a currently fashionable approach for correlating a number of macroscopic hydraulic and transport parameters in multiphase reactors. Hence, on the basis of a wide hydrodynamic data set, these authors propose a set of general perceptron-like artificial neural network (ANN) correlations for the prediction of the height equivalent to a theoretical plate (HETP) and the twophase pressure gradient (∆P/H) in counter-current gasliquid structured packing towers. The proposed tools are shown to outperform, in terms of prediction capability, the well-trodden empirical correlations or phenomenological models existing in the field. Although these authors are successful in demonstrating their concept, the impact of their contribution, can, to our opinion, be further reinforced. The procurement in a publication of the full expression and parameters of a correlation is the sole guarantee that such a tool can be useful to readers from industry and academia. In this work, the authors neglect to provide for both the HETP and ∆P/H in their derived correlation equations together with the numerical values of the weights. This unfortunately makes it impossible for users to tangibly take advantage of these two papers. Basically, the most valuable aspect of this work is more the correlation equations rather than the methodology implemented to extract the correlations. For the benefit of the readers of Ind. Eng. Chem. Res., it is highly suggested that the authors provide the complete set of equations allowing for the computation by their tools of the hydrodynamic parameters in the chosen configuration. Neural network computing is becoming increasingly fashionable among the chemical engineering circle. It is a powerful “black-box” approach used to map complex * Corresponding author. Tel.: 418-656-3566. Fax: 418-6565993. E-mail: [email protected].

nonlinear system behaviors such as those encountered in multiphase-flow hydrodynamics problems. Unfortunately, despite the numerous investigations of ANN correlations that abound in the literature, the relevance of publishing many of them is questionable merely because the authors often neglect to publish the full sets of equations. We are aware that neural network correlation equations are cumbersome and could significantly burden a paper. This perhaps has contributed to the reluctance of contributors to fully quote them. To change this tendency and to help the perpetuate such useful information, a number of solutions can be envisioned, such as providing “Supporting Information” to be archived by the journal, or adjoining the full equations in appendices accompanying the papers, or releasing Internet electronic source codes, spreadsheets, etc. that can be downloaded, or when appropriate, providing compact graphical plots. Finally, by quoting the introductory sentence in the recent article by the respected professor S. W. Churchill3 “Correlation is of direct importance to all engineerssboth those who generate new information and those who wish to make use of it” (sic), we hope that we have made the contributors aware of the importance of what is meant by a useful correlation. Literature Cited (1) Whaley, A. K.; Bode, C. A.; Ghosh, J. G.; Eldridge, R. B. HETP and Pressure Drop Prediction for Structured Packing Distillation Columns Using a Neural Network. Ind. Eng. Chem. Res. 1999, 38, 1736-1739. (2) Pollock, G. S.; Eldridge, R. B. Neural Network Modeling of Structured Packing Height Equivalent to a Theoretical Plate. Ind. Eng. Chem. Res. 2000, 39, 1520-1525. (3) Churchill, S. W. The Art of Correlation. Ind. Eng. Chem. Res. 2000, 39, 1850-1877.

IE000679F

10.1021/ie000679f CCC: $19.00 © 2000 American Chemical Society Published on Web 09/22/2000