Evaluation of Top Compositions in Debutanizer through Artificial Neural

CATEGORY: DEBUTANIZER
International Conference on Mechanical, Automobile and Robotics Engineering (ICMAR'2011)
Evaluation of Top Compositions in Debutanizer through Artificial Neural
Hamed Sahraie*, Ali Ghaffari, Majid Amidpour
H.Sahraie is with the National Iranian Oil Co.(NIOC)
Hamed.Sahraie@yahoo.com
A.Ghaffari and M.Amidpour are with K.N.Toosi University of Technology,Tehran, Iran. They are now with the Department of Mechanical Eng
Ghaffari@kntu.ac.ir
Amidpour@kntu.ac.ir
Abstract
An artificial neural network estimator is designed to evaluate the top products from the secondary measurements. Based on the real data in a refinery, the soft computing approach is designed to minimize the root mean square error between the actual system output and the estimator. It is shown that the results are quite satisfactory and acceptable.
DISTILLATION of multicomponent mixtures is one of the most common separation operations in the chemical industry and refineries. Distillation is a constrained, coupled, nonlinear, no stationary process and has different dynamic behavior. The refinery community has recognized the importance of the optimization of process automation because of the benefits in terms of both profitability and tight control on product quality. In certain control applications, there are situations when some of the parameters cannot be measured economically online because of either the instrumentation is very expensive or measurement introduces lags, which designing an effective control system will be impossible. In such situations from the secondary measurements, inference is made for the desired parameters, artificial neural network (ANN) best suits for this application. The control of many industrial processes is difficult because online measurement of product quality is complicated. This is due to the lack of measurement technology. In 1972, Weber and Brosilow [1] proposed using secondary measurement to control the variables that their measurements are difficult or impossible and in 1978 Tong and Brosilow [2] introduced the structure and dynamic system of inferential control. Patke [3] (1979) applied the inferential control technique to single composition control of laboratory-scale distillation column separating a mixture of n-propanol and methanol.
A new method to select the appropriate temperature measurement locations in a distillation column for feedback control has been developed by Bequette and Edgar [4] in 1989. The proposed method was based on the trade-off of measurement sensitivity vs. inferential accuracy.
Free Full Text Source: http://psrcentre.org/images/extraimages/1011083.pdf