الفهرس 
SummaryOnly 14 pages are availabe for public view
 |  | from | 140 |  |  |
| | | from | 140 | | |
Abstract
Estimating an unknown function from a set of input-output (I/O) data pairs has been and is still a key issue in a variety of scientific and engineering fields. The problem of determining a mathematical model for an unknown system (also referred to as the target system) by observing its input-output data pairs is generally referred to as System Identification. One of the main reasons for making system Identification is that accurate mathematical models do not always exist nor can they be derived for all complex environments. System identification consists of three basic sub problems: 1:) Structure specification; 2) Parameter estimation; 3:) Model validation. Structure specification involves finding the important input variables from all possible input variables, specifying membership functions, partitioning the input space, and determines the number of fuzzy rules comprising the underlying model. Parameter estimation involves the determination of unknown parameters in the model using some optimization method based on both linguistic information obtained from human experts and numerical data obtained from the actual system to be modeled. Model Validation involves testing the model based on some performance criteria (e.g. accuracy). PI (Performance Index) and APE (Average percentage Error) have been used as Performance criteria in our algorithm. There are many methods to make system identification, the best of that is intelligent controllers. Intelligent controllers (sometimes termed soft controllers) are characterized by their ability to establish the functional relationship between their inputs and output from empirical data, without recourse to explicit models of the controlled process. This is a radical departure from conventional controllers, which are based on explicit functional relations. Unlike their conventional counterparts, intelligent controllers can learn, remember and make decisions. Intelligent Control includes Fuzzy, Neural, Neuro-Fuzzy and Evolutionary Control. Being able to deal with linguistic information and numerical information in a systematic and efficient manner, ability to handle non-linearity, and interpretability are salient features of fuzzy models. However, there are no standard methods exist for transforming human knowledge or experience into the rule base and there is a need for effective methods for tuning the membership function (MFs) so as to minimize the output error measure or maximize performance index. It is possible to obtain input-output relationships of a nonlinear system using a three-layered BP-type NN. The rules acquired by the NN are, however, distributed in the network and hard to understand. In short, the ANN is essentially a “black box” model and difficult to interpret. Due to the advantage and disadvantage of NN and Fuzzy we will marry them in ANFIS (Adaptive-Network-based Fuzzy Inference System). In our algorithm we present a quick and straightforward way of input selection (Fuzzy Curves) for Neuro-Fuzzy modeling. The proposed method of input selection was tested on a simple Example. We have used the Subtractive clustering to get the number of Membership functions in each input. The proposed Neuro-Fuzzy Identifier gives a very accurate fuzzy modeling of unknown systems and has optimal fuzzy variables such as the center points and shapes of input and output fuzzy membership function and the whole proposed identifier was tested on 5 cases.