Razvoj modela za prediktivno održavanje rotacijske opreme temeljenog na ultrazvučnoj analizi

master's thesis

• Title (english): Development of a predictive maintenance model for rotating equipment based on ultrasound analysis
• Author: Lovro Ugarković
• Mentor: Davor Kolar (mentor)
• Granter: University of Zagreb; Faculty of Mechanical Engineering and Naval Architecture; Zagreb
• Defense date and country: 2024-12-05, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Mechanical Engineering

Abstract

Ovaj diplomski rad istražuje primjenu strojnog učenja u prediktivnom održavanju rotacijske opreme kroz prepoznavanje kvarova na ležajevima temeljem ultrazvučne analize. Glavni cilj istraživanja je razviti aplikaciju koja automatski detektira kvarove na ležajevima te korisnicima omogućuje pravovremeno održavanje čime se smanjuje rizik od neočekivanih zastoja i povećava učinkovitost održavanja. Prikupljanje podataka provedeno je pomoću uređaja SDT 340 koji je snimao ultrazvučne signale u različitim stanjima ležajeva, uključujući normalno stanje i kvarove na unutarnjem prstenu, vanjskom prstenu te kuglici ležaja. Signalni podaci obrađeni su u Matlab okruženju, a kroz razvijenu aplikaciju omogućeno je izdvajanje ključnih značajki signala te prikaz samog signala. Za klasifikaciju kvarova korišteni su različiti modeli strojnog učenja, uključujući SVM s kvadratnom jezgrom i neuronske mreže. Modeli su testirani na temelju preciznosti klasifikacije te su postigli visok stupanj točnosti od 96.4%. Rezultati istraživanja potvrđuju prikladnost ovih metoda za prediktivno održavanje te postavljaju temelje za daljnji razvoj sustava za industrijsku primjenu.

Abstract (english)

This thesis explores the application of machine learning in predictive maintenance for rotary
equipment through bearing fault detection based on ultrasonic analysis. The primary goal of the
research is to develop an application that can automatically detect bearing faults, thereby
enabling timely maintenance, reducing the risk of unexpected downtimes, and enhancing
maintenance efficiency. Data collection was conducted using the SDT 340 device, which
recorded ultrasonic signals under different bearing conditions, including normal state and faults
in the inner ring, outer ring, and bearing ball. Signal data were processed in the Matlab
environment, where the developed application enabled the extraction of key features and the
display of the signal. Various machine learning models, including an SVM with a quadratic
kernel and neural networks, were used for fault classification. The models were evaluated based
on classification accuracy and the models achieved a high accuracy of 96.4%. The research
findings confirm the suitability of these methods for predictive maintenance and lay the
groundwork for further development of industrial maintenance systems.

• Keywords: prediktivno održavanje; ležajevi; ultrazvučna analiza; strojno učenje; Matlab
• Keywords (english): predictive maintenance; bearings; ultrasonic analysis; machine learning; Matlab
• Language: croatian
• URN:NBN: urn:nbn:hr:235:023238
• Study programme: Title: Mechanical Engineering; specializations in: Design, Process and Energy Engineering, Production Engineering, Engineering Modelling and Computer Simulation, Marine Engineering, Industrial Engineering and Management, Materials Engineering, Mechanics and Robotics, Autonomous Systems, Mechatronics and Robotics; Course: Industrial Engineering and Management; Study programme type: university; Study level: graduate; Academic / professional title: sveučilišni/a magistar/magistra inženjer/inženjerka strojarstva (sveučilišni/a magistar/magistra inženjer/inženjerka strojarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Open access
• Created on: 2024-11-28 15:50:26