الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 165 |  |  |
| | | from | 165 | | |
Abstract
The aim of neuro-fuzzy systems is to combine collectively the benefits of both neural networks and fuzzy logic approaches. It was noted that neural networks have two main benefits. First, they are capable of learning non-linear mappings of numerical data. Second, they perform parallel computation. However, the operation of neural networks have also many weaknesses. For example, in the popular multilayer perceptron network, the knowledge of the system is distributed into the whole network as synaptic weights. Therefore, it is very hard to understand the meaning of weights, and the incorporation of prior knowledge into the system is usually impossible. The knowledge in neural network cannot be easily extracted into linguistic rules. Fuzzy logic uses human understandable linguistic terms to express the knowledge of the system. This makes possible a close interaction between the system and human operator which is very desirable property. Neuro-fuzzy systems allow incorporation of both numerical and linguistic data into the system. The key advantage of neural fuzzy approach over traditional ones lies on that : (1)The former doesn’t require an accurate mathematical description of the system while modeling. (2) Moreover, in contrast to pure neural or fuzzy methods, the neural fuzzy method possesses both of their advantages; It brings the low-level learning and computational power of neural networks into fuzzy systems and provides the high-level human-like thinking and reasoning of fuzzy systems into neural networks. The objective of this study is to design a control system that does not require any explicit information about the object dynamics. This is done with the help of the identification process of the controller which includes both structure and parameter learning. To identify a plant, we do not need its dynamic equation. We can collect the training data by applying randomly generated input data to the plant, and collect the input-output data pairs. Then, train the proposed system by these collected data, but since in supervised learning method we must know the I/O training data, hence, we must get the I/O information from the outside by either providing from the expert knowledge or deriving from the known I/O equation . The proposed controller is a self-constructing neural fuzzy inference network possessing neural network’s learning ability. There are no rules initially in the proposed controller. They are created and adapted as on-line learning proceeds via simultaneous structure and parameter identification. The structure as well as the parameter learning phases are done simultaneously in the proposed self-constructing neural fuzzy inference network by performing them both for each incoming data, this makes the controller faster. In this thesis, controller is applied to model a control system for an autonomous underwater vehicle, cart-pole balancing problem, water bath temperature control & a prediction system for soil liquefaction Computer simulations were conducted to demonstrate: (1)The self-organizing learning and self-adapting abilities of the proposed system. In order to verify the performance of the controller system .(2)The better performance of the proposed controller as compared to an adaptive controller.