(0007) Inferential sensing techniques in industrial applications. Masters thesis, KingFahd UNiversity of Petroleum and Minerals.

Microsoft Word (A thesis on : Inferential sensing techniques in industrial applications) Theses-inferential_sensing_techniques_in_industrial_applications.doc Download (36kB)

Abstract

Climate change caused by pollution is considered as one of threats facing humankind. Industrial emission is one of the main sources of the air pollution. There has been many efforts to protect climate at the regional and international level by requiring industries to monitor, limit and report their emissions. Inferential sensing techniques have been gaining momentum recently as viable alternatives to hardware sensors (i.e. Continuous Emission Monitoring System) in various situations. Inferential sensing technique is a method to estimate certain quantities based on a set of conventional measurements. The core of inferential sensing is based on modeling and estimation techniques. In this work, dynamical neural networks are investigated to build inferential sensor for the emissions due to combustion operation in industrial boilers. The emission andpollutants formation in industrial boiler is a dynamical and nonlinear process. Neural networks are powerful tool for modeling highly complex nonlinear systems, especially when the physics of the system is not clearly known or di±cult to determine. A modular approach is used to develope the inferential sensor model. Different types of dynamical neural networks are combined according to system operation and emission behavior. Real data from a boiler plant is used to develop the model. Input variablesare grouped and Principal Component Analysis is used to reduce the total number of input variables to the proposed model. System delays are obtained by approximating the model by a linear model. Genetic algorithm, which is a heuristic optimization technique, is used to ¯nd the system delays of the linear model, which are used in dynamical neural network model. The dynamical neural network results are compared with the static neural network models. The dynamical neural networks provided better results in terms of complexity of network structure, generalization and prediction error.

Item Type:	Thesis (Masters)
Subjects:	Systems
Department:	College of Computing and Mathematics > lndustrial and Systems Engineering
Depositing User:	MR. ATIQ WALIULLAH SIDDIQUI
Date Deposited:	14 Jun 2008 10:45
Last Modified:	01 Nov 2019 13:44
URI:	http://eprints.kfupm.edu.sa/id/eprint/2526