Title
Upravljanje dinamičkim sistemima primenom adaptivnih ortogonalnih neuronskih mreža
Creator
Milovanović, Miroslav B. 1987-
Copyright date
2017
Object Links
Select license
Autorstvo 3.0 Srbija (CC BY 3.0)
License description
Dozvoljavate umnožavanje, distribuciju i javno saopštavanje dela, i prerade, ako se navede ime autora na način odredjen od strane autora ili davaoca licence, čak i u komercijalne svrhe. Ovo je najslobodnija od svih licenci. Osnovni opis Licence: http://creativecommons.org/licenses/by/3.0/rs/deed.sr_LATN Sadržaj ugovora u celini: http://creativecommons.org/licenses/by/3.0/rs/legalcode.sr-Latn
Language
Serbian
Cobiss-ID
Theses Type
Doktorska disertacija
description
Datum odbrane: 16.03.2018.
Other responsibilities
mentor
Antić, Dragan 1963-
član komisije
Nikolić, Vlastimir
član komisije
Mitić, Darko
član komisije
Milojković, Marko 1980-
član komisije
Stojanović, Sreten
Academic Expertise
Tehničko-tehnološke nauke
University
Univerzitet u Nišu
Faculty
Elektronski fakultet
Group
Katedra za elektroniku
Alternative title
Control of dynamical systems by using adaptive orthogonal neural networks
Publisher
[M. B. Milovanović]
Format
177 str.
description
Bibliografija: listovi158-171
Abstract (en)
The goal of the research in the PhD dissertation is control of dynamical systems by using new types of orthogonal endocrine neural networks, in order to improve their performances. Standard artificial neural networks are described, as well as their historical development and basic types of learning algorithms. Further, possibilities for neural networks applicability within control logic of dynamical systems are presented, as well as the current state of the art of orthogonal and endocrine neural networks. Performance improvement of the laboratory model of a servo system by using a standard neural network with the backpropagation type of learning is analyzed. In addition, a method for selection and optimization of training data, as an efficient way of information preprocessing for the purpose of improving performances of a neural network, is presented.
A detailed description of orthogonal functions and implementation methods of endocrine factors inside standard neural networks are provided. By implementation of orthogonal activation functions of neurons, verification of their applicability in control of dynamical systems was performed. The laboratory model of the magnetic levitation system was used to test the designed orthogonal neural network. Furthermore, the endocrine orthogonal neural network based on the biological processes of excitation and inhibition is designed. Network performance checkup is performed by testing its predictive abilities when working with time series data.
Final dissertation researches refer to development of hybrid systems. The implemented adaptive endocrine neuro-fuzzy hybrid system is tested through modeling of a laboratory servo system. Other hybrid structure, based on a combination of an orthogonal endocrine neural network and an orthogonal endocrine neuro-fuzzy hybrid system, is designed with the aim to form symbiosis of the positive characteristics of the individual networks. Verification of this structure was performed by using it for PID controller parameters adjustments.
Authors Key words
neuronska mreža, ANFIS, ortogonalne funkcije, endokrini faktor
Authors Key words
neural network, ANFIS, orthogonal functions, endocrine factor
Classification
004.7:681.5(043.3)
Type
Elektronska teza
Abstract (en)
The goal of the research in the PhD dissertation is control of dynamical systems by using new types of orthogonal endocrine neural networks, in order to improve their performances. Standard artificial neural networks are described, as well as their historical development and basic types of learning algorithms. Further, possibilities for neural networks applicability within control logic of dynamical systems are presented, as well as the current state of the art of orthogonal and endocrine neural networks. Performance improvement of the laboratory model of a servo system by using a standard neural network with the backpropagation type of learning is analyzed. In addition, a method for selection and optimization of training data, as an efficient way of information preprocessing for the purpose of improving performances of a neural network, is presented.
A detailed description of orthogonal functions and implementation methods of endocrine factors inside standard neural networks are provided. By implementation of orthogonal activation functions of neurons, verification of their applicability in control of dynamical systems was performed. The laboratory model of the magnetic levitation system was used to test the designed orthogonal neural network. Furthermore, the endocrine orthogonal neural network based on the biological processes of excitation and inhibition is designed. Network performance checkup is performed by testing its predictive abilities when working with time series data.
Final dissertation researches refer to development of hybrid systems. The implemented adaptive endocrine neuro-fuzzy hybrid system is tested through modeling of a laboratory servo system. Other hybrid structure, based on a combination of an orthogonal endocrine neural network and an orthogonal endocrine neuro-fuzzy hybrid system, is designed with the aim to form symbiosis of the positive characteristics of the individual networks. Verification of this structure was performed by using it for PID controller parameters adjustments.
“Data exchange” service offers individual users metadata transfer in several different formats. Citation formats are offered for transfers in texts as for the transfer into internet pages. Citation formats include permanent links that guarantee access to cited sources. For use are commonly structured metadata schemes : Dublin Core xml and ETUB-MS xml, local adaptation of international ETD-MS scheme intended for use in academic documents.