Development of GC Analyzer Model Using Neural Network

Gas chromatography (GC) is the most widely used teclmique in analytical
chemistry. It is an analytical scientific teclmique to separate a mixture of vaporizable
substances and resolve the mixture into single components. Analyzer as hardware
has high initial cost, requires frequent maintenance and sometimes fails to provide
the accurate outputs. Moreover, the increasing complexity of industrial processes and
the struggle for cost reduction, availability, safety and higher profitability requires
efficient and reliable instruments. Thus, this final year project is an attempt to
develop prototype software that is capable of predicting efficiently plant output and
optimize the performance of the model. MATLAB Neural Network and system
identification toolboxes were utilized to recommend the best structure to develop this
predictive model. The purpose of this report is to show the success and applicability
of using neural network in predicting plant output and obtain an alternative
measuring system. It presents the followed methodology in achieving project's
objectives by giving an overview on neural network and system identification
toolboxes and shows a comparison of the performance of Back Propagation Feed
Forward Neural Network (BFN) and other System I identification toolbox models.
Results demonstrated that neural network model trained using LM provides an
adequate result and is suitable for this purposes.