Dinamička neuronska mreža u regulaciji stupnja istrošenosti reznog alata

Rajaković, Marko (2014) Dinamička neuronska mreža u regulaciji stupnja istrošenosti reznog alata. = Application of recurrent neural network in the tool wear control. Undergraduate thesis , Sveučilište u Zagrebu, Fakultet strojarstva i brodogradnje, UNSPECIFIED. Mentor: Brezak, Danko.

[img]
Preview
Text
17_09_2014_Marko_Rajakovic.pdf Jezik dokumenta:Croatian

Download (3MB) | Preview

Abstract (Croatian)

U ovom završnom radu opisana je struktura i način funkcioniranja umjetnih neuronskih mreža, te njihova povezanost s biološkim neuronskim mrežama. Zatim je opisan dinamički neuron i struktura dinamičke neuronske mreže. Neuronska mreža u strukturi skrivenog sloja ima dinamičke neurone drugog, četvrtog ili šestog reda, a učena je metodom povratnog rasprostiranja pogreške (eng. error back-propagation). Radi ubrzanja učenja u mrežu je dodan momentum prvog reda. Kasnije se funkcionalnost mreže testira na proporcionalnom članu prvog reda i mreža se testira u svrhu filtriranja odaziva estimatora trošenja reznih alata.

Abstract

This paper describes the structure and the operating principle of artificial neural networks, as well as their link to biological neural networks. The structures of a basic dynamic neuron and a dynamic network are defined. The network's hidden layer contains either second, fourth or sixth order dynamic neurons. The network is trained via the error backpropagation method and the training speed is increased with the use of a first order momentum. The network’s functionality is tested using a first order proportional controller and later for the purpose of filtering the response of a cutting tool wear estimator.

Item Type: Thesis (Undergraduate thesis)
Uncontrolled Keywords: strojarstvo; dinamička; neuronska mreža; trošenje alata; mehatronika
Keywords (Croatian): engineering; dynamic; neural network; tool wear; mechatronics
Divisions: 900 Department of Robotics and Production System Automation > 910 Chair of Engineering Automation
Date Deposited: 22 Sep 2014 18:00
Last Modified: 02 Dec 2020 13:12
URI: http://repozitorij.fsb.hr/id/eprint/2897

Actions (login required)

View Item View Item

Downloads

Downloads per month over past year