Online Fault Detection And Diagnosis For Batch Polymerization Reactor Using Artificial Neural Networks
Loading...
Date
2012-08
Authors
Md Nor, Norazwan
Journal Title
Journal ISSN
Volume Title
Publisher
Universiti Sains Malaysia
Abstract
Batch reactor is an essential unit operation in almost all batch-processing industries, especially in polymers, pharmaceuticals and fine chemicals manufacturing, but with the probabilities of failure in any situation. This has encourages the researchers and industries to find appropriate techniques to detect the process failure as early as possible. This study proposed the development of fault detection and diagnosis system for batch polymerization reactor using an Artificial Neural Network (ANN). In this study, batch polymerization reaction models were solved using MATLAB. After the validation process, the normal and faulty operating condition were simulated to generate the input data for the ANN. Two type of neural network model was proposed as part of the fault detection system, each for process variables estimation and fault classifying purpose. All the proposed approaches was built using MATLAB/Simulink® platform. The process estimator networks were adapted based on Elman networks whereas the fault classifying networks were built based on pattern recognition approach. The optimal structure of both network models and their effects to the network performance were studied throughout this research, and the results were presented. The developed detection system was successfully in detecting both single and multiple faults that were prescribed during the offline training and validation session. Finally the developed detection system was implemented as an online detection system with the real-time simulation process.
The results showed that the proposed model successfully implemented and produced remarkable results as online fault detection and diagnosis system for batch polymerization reactor.
Description
Keywords
Modeling of fault detection , in chemical batch reactor