Loading...
| Friend's email | |
| Your name | |
| Your email | |
| enter code | |
This page was sent successfuly
- Type of Document: Ph.D. Dissertation
- Language: Farsi
- Document No: 41482 (06)
- University: Sharif University of Technology
- Department: Chemical and Petroleum Engineering
- Advisor(s): Bozorgmehry, Ramin; Shahrokhi, Mohammad
- Abstract:
- The accurate fault diagnosing system design helps the process safety and also helps increasing the products quality of the process. In this project, the fuzzy system and neural network have been used for fault detection and diagnosis of a yeast fermentation bioreactor. In one case, parameters of membership functions are selected in a conventional manner. In second case, the optimal values of these parameters have been obtained using the genetic algorithm. In another case, the neural network system is used for fault detection. These three cases are compared based on their performances in fault diagnosis of a yeast fermentation bioreactor for three different conditions. The results indicate that the fuzzy-genetic system and the neural network are superior in multiple fault detection for the conditions where the minimum and maximum deviations from normal conditions occur in the process variables. In many cases, multiple fault diagnosis of plant wide systems based on steady state data is impossible. To solve this problem, a neuromorphic framework, whose neural network used as the diagnoser is trained, based on a hybrid set of steady and dynamic characteristic data have been proposed. The dynamic characteristic data include overshoot and undershoot values of measured variables and their corresponding occurrence times. The Tennessee Eastman process has been used as the performance assessment benchmark. Various combinations of concurrent faults have been considered in this assessment. The results indicate the generality, flexibility and accuracy of the proposed framework. Because of the large size and complexity of the many industrial processes, the implementation of a comprehensive detection system for an entire process can be difficult and/or impossible. In these conditions, the multi-agent approach can be an appropriate solution to fault diagnosis system. In the presented approach, the TE process divided into four small units. An integrated neural network system has been designed for three units and two diagnoser based on the fuzzy system have been designed for another unit and the undiagnosed faults. The results indicate the design of the proposed algorithm is faster than the design of one diagnoser based on the neural network
- Keywords:
- Fuzzy Logic ; Genetic Algorithm ; Neural Network ; Object Oriented Approach ; Fault Detection ; Fault Diagnosis ; Tennessee-Eastman Process ; Fermentation Bioreactor
Digital Object List-
محتواي پايان نامه - view
Bookmark