Filtriranje signala primjenom dinamičke neuronske mreže u regulaciji istrošenosti reznih alata

master's thesis

• Title (english): Filtering of estimated flank wear widths using recurrent neural network in the tool wear regulation process
• Author: Sara Bešić
• Mentor: Danko Brezak (mentor)
• Granter: University of Zagreb; Faculty of Mechanical Engineering and Naval Architecture; Zagreb
• Defense date and country: 2020-07-14, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Mechanical Engineering

Abstract

Adaptivno vođenje alatnih strojeva prema istrošenosti reznog alata se, zbog nemogućnosti direktnog mjerenja stanja oštrice tijekom procesa obrade, oslanja na indirektno mjerenje signala povezanih s trošenjem i njihovu estimaciju kako bi se dobile informacije o stanju rezne oštrice u realnom vremenu. Unutar snimljenih signala obuhvaćeni su i oni signali koji nisu izravno povezani s procesom trošenja, a javljaju se prilikom obrade odvajanjem čestica. Pogreška estimacije neizbježna je u procesu regulacije. Kako bi se smanjile neželjene smetnje, poremećaji i šumovi, signale je potrebno obraditi, odnosno filtrirati. Filtriranjem se nepoželjni utjecaji potiskuju na prihvatljivu razinu, odnosno smanjuje se pogreška estimacije i njezin utjecaj u regulacijskom krugu nadzora obradnog stroja prema istrošenosti reznog alata. U ovom radu korištene su dvije vrste filtra, FIR i IIR filtar. Oba filtra korištena su u tri modifikacije kao filtri 2., 4. i 6. reda. Realizacija filtara ostvarena je putem dinamičke neuronske mreže. Cilj je usporediti rezultate obrade signala svih šest filtara te izvesti zaključak utječe li vrsta filtra i njegov red na uspješnost potiskivanja poremećaja u signalu obrade. Signal obrade predstavljaju eksperimentalno prikupljeni podaci iz procesa obrade tokarenjem.

Abstract (english)

Due to the impossibility of directly measuring the condition of a tool during the machining process, adaptive machine control for maintaining the tool wear rate in the predefined cutting time relies on measuring signals that are related to tool wear. Those signals are then estimated in real-time. Based on that estimation, tool wear condition is predicted. Measured signals also include some signals that are not related to tool wear, but they occur during the machining process. In regulation the estimation error is inevitable. To reduce the effect of different kinds of interferences, disturbances and noises, signals are processed. Filtering suppresses interferences to an acceptable level, i.e. it reduces estimation error and its influence in the tool wear regulation process. This paper reviews two types of filters: FIR and IIR and their modifications as filters of 2nd, 4th and 6th order. Filters are structured using a recurrent neural network algorithm. The objective of this paper is to analyse and compare the results of mentioned filters and to draw a conclusion if the filter type and the filter order affect filter’s performance. Experimentally collected data from the turning process are used as input signals for filters.

• Keywords: nadzor istrošenosti reznog alata; šum; obrada signala; filtriranje; FIR filtar; IIR filtar; dinamička neuronska mreža
• Keywords (english): tool wear regulation process; noise; signal processing; filtering; FIR filter; IIR filter; recurrent neural network
• Language: croatian
• URN:NBN: urn:nbn:hr:235:685351
• Study programme: Title: Mechanical Engineering; Study programme type: university; Study level: graduate; Academic / professional title: magistar/magistra inženjer/inženjerka strojarstva (magistar/magistra inženjer/inženjerka strojarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Open access; Embargo expiration date: 2020-07-14
• Created on: 2020-07-13 11:00:29