Last update:
12/27/2005
 |
Neural Networks for Chemical Process Analysis |
Deterministic modeling of complex chemical and separation processes is often infeasible when such processes involve a series of complex steps and are affected by numerous variables. Recent, advances in neural network system design have made it possible to map complex process behavior, given sufficient operational data, and to be able to make generalizations regarding the process and even assess potential process upsets. Although, there have been a number of scoping studies on the application of neural networks to process modeling, UCLA is in a unique position having developed, in collaboration with the University of Tarragona (Spain), a novel superior neural network approach for modeling complex systems.
The study will involve a collaboration with a number of membrane manufacturers as well as a commercial membrane facility and a pilot plant in California. The outcome of the study will be a unique analysis system to explore system performance and pilot plant optimization. Given, the growing utilization of membranes in both the chemical and pharmaceutical industries and on a much larger scale by water utilities, the results of this research should have an immediate impact on this rapidly expanding area of membrane technology.
Fundamental areas: process analysis, transport phenomena, separations, process thermodynamics, neural networks, cognitive systems, computer modeling.
References:
- Recent papers on neural networks by Professor Cohen's Research Group
- Bowen, W. R., Jones, M. G., Yousef, H. N. S., Prediction of the rate of crossflow membrane ultrafiltration of colloids: A neural network approach. Chem.Eng.Sci., 1998, V53(N22):3793-3802.
- Bowen, W. R., Jones, M. G., Welfoot, J. S., and Yousef, H, N. S., Predicting salt rejections at nanofiltration membranes using artificial neural networks. Desalination, 10, 2000, V129(N2):147-162.
- Elhawary, M. E.. Artifical neural networks and
possible applications to Desalination, 1993, V92(N1-3):125-147.
