 Font Size: a A A

# Fuzzy Models For The Identification Of Nonlinear System

Posted on:2006-06-21Degree:MasterType:Thesis
Country:ChinaCandidate:J L YangFull Text:PDF
GTID:2178360182455123Subject:Mechanical Manufacturing and Automation
Abstract/Summary: PDF Full Text Request
Fuzzy model has been recognized as a powerful tool which can facilitate the effective development of models by combining information from different sources, such as empirical models, the expert knowledge or the system input-output data. And the result model is more interpretable than other black-box methods. By now there are many way to build the fuzzy model, but there also are some unsolved problems.In this thesis, we will discuss how to extract the model from system input-output data, how to insure the model precision and interpretability. We will major discuss two types questions:A. The estimate of Fuzzy model's structures and parameters. Firstly, construction methods based on fuzzy C-means clustering originate from data analysis and pattern recognition, where the concept of fuzzy membership is used to represent the degree to which a given data object is similar to some prototypical object. Then, extract the antecedent membership functions by projecting the cluster onto the individual variables. The number of clusters is equal to the number of rules in the rule base.B. The between model precision and interpretability. Contrast with others method, the most advantage of fuzzy model is the interpretability of the result. However, many methods don't handle this issue. Moreover, in general no model reduction is applied, while this may simplify the model by removing redundant information. The model is very complex and there are heavily redundancies from which we can't know the behavior of the system. In this thesis, we merge similar fuzzy sets based on similarity measure between fuzzy sets in order to reduce the rule base. The simulation results show it's effective.C. At last, use optimization algorithm to estimate the rule base's parameters. In case, it can be formed as a fuzzy network model, the cluster can be defined as linear subspaces of the system. And use the error back propagation or a hybrid learning algorithm to training the parameters. Compare whit other methods, it is more advantage which can get a more compact model with less cost. These three items place different emphasis on the fuzzy model identification issue, they are interrelated. When deal with a problem, repeatedly iterates between them, however, they can be rearranged corresponding to specific needs.The results of simulations and experiments show that proposed methods can build a fuzzy model from the data and expert knowledge. The model has good performancein precision and interpretability.
Keywords/Search Tags:nonlinear system identification, fuzzy system, hybrid learning algorithm, fuzzc means clustering, interpretability PDF Full Text Request
Related items
 1 Research On Genetic Fuzzy System For Knowledge Extraction 2 Study On Fuzzy Modeling Based On Interpretability And Precision 3 Fuzzy C-means And K-means Clustering Algorithm And Its Parallel 4 Evolutionary Strategy Learning, Convergence And Escape Capacity Research And Application 5 Nonlinear System Identification Based On T-S Fuzzy Model 6 Research On Nonlinear System Identification Using Fuzzy Clustering 7 Probabilistic K-means Models Via Nonlinear Programming 8 The Application Of Fuzzy C-means Clustering In The Stock Investment 9 Research On The Extension Of Fuzzy C-means Clustering Algorithm 10 Fusion Ant Heap Of Clustering And Fuzzy C-Means Clustering Algorithm Research And Analysis