Identifikacija i upravljanje nelinearnim vremenski promjenljivim procesima primjenom neuronskih mreža

doctoral thesis

• Title (english): Neural network based identification and control of non-linear time variable processes
• Author: Mato Baotić
• Mentor: Nedjeljko Perić (mentor)
• Committee member: Ivan Petrović (predsjednik povjerenstva)
• Committee member: Nedjeljko Perić (član povjerenstva)
• Committee member: Želimir Kurtanjek (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; (Department of Control and Computer Engineering); Zagreb
• Defense date and country: 2000-10-24, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing; Process Computing
• Universal decimal classification (UDC): 007 - Cybernetics. Self-acting systems

Abstract

Opisana je identifikacija nelinearnih vremenski promjenljivih procesa, pri čemu je posebna pažnja posvećena primjeni neuronskih mreža u identifikaciji nelinearnih procesa. Objašnjena je veza između općeg ulazno-izlaznog nelinearnog modela i statičkih neuronskih mreža. Detaljno su opisane korištene strukture neuronskih mreža: višeslojna perceptronska mreža (MLP), neizrazita neuronska mreža (FNN) i CMAC mreža. Opisani su i ispitani nerekurzivni i rekurzivni algoritmi učenja neuronskih mreža. Prezentiran je algoritam trenutačne linearizacije. Predložen je sustav prediktivnog upravljanja koji se sastoji od nelinearnog modela zadanog procesa realiziranog neuronskom mrežom, postupka trenutačne linearizacije i poopćenog prediktivnog regulatora. Svojstva sustava upravljanja ispitana su simulacijom na računalu na matematičkom modelu nelinearnog sustava. Temeljem jednadžbi očuvanja tvari i energije izveden je fizikalni model šaržne destilacijske kolone. Model je uspoređen s eksperimentalnim rezultatima dobivenim na laboratorijskoj destilacijksoj koloni. Simulacijom na računalu provjerena je primjena predloženog sustava prediktivnog upravljanja pri upravljanju šaržnom destilacijksom kolonom.

Abstract (english)

System identification of non-linear time variant processes is presented, with special attention given to the identification of non-linear processes using neural networks. Connection between black box input-output non-linear models and static neural networks is described. Structures of MultiLayer Perceptron (MLP), Fuzzy-Neural Network (FNN) and Cerebellar Model Articulation Controller (CMAC) are discussed in great detail. On-line and off-line learning algorithms for these neural networks are described and tested. Instantaneous linearization method is derived. System of predictive control is given, which is comprised from nonlinear neural model of given process, instantaneous linearization method and generalized predictive controller. Characteristics of given system were proved through computer simulations. Mathematical model of batch distillation column based on heat and material balances is derived. Model is compared with the experimental results obtained on a laboratory distillation column. Predictive control of batch distillation column was tested through computer simulations.

• Keywords: Identifikacija procesa; nelinearni vremenski promjenljivi procesi; neizrazita neuronska mreža; CMAC neuronska mreža; trenutačna linearizacija; prediktivno upravljanje; šaržna destilacijska kolona
• Keywords (english): Process identification; non-linear time variant processes; fuzzy neural network; CMAC neural network; instantaneous linearization; predictive control; batch distillation column
• Language: croatian
• URN:NBN: urn:nbn:hr:168:596805
• Study programme: Title: Postgraduate doctoral university study; Study programme type: university; Study level: postgraduate; Academic / professional title: doktor/doktorica znanosti (doktor/doktorica znanosti)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2019-04-10 13:37:46