Automated methods for vibration-based condition monitoring of rotating machines
Thesis event information
Date and time of the thesis defence
Place of the thesis defence
OP-Pohjola auditorium (L6), Pentti Kaiterankatu 1, Linnanmaa, Oulu
Topic of the dissertation
Automated methods for vibration-based condition monitoring of rotating machines
Doctoral candidate
Master of Science in Technology Riku-Pekka Nikula
Faculty and unit
University of Oulu Graduate School, Faculty of Technology, Environmental and Chemical Engineering
Subject of study
Process Engineering
Opponent
Professor Matti Vilkko, Tampere University
Custos
Professor Mika Ruusunen, University of Oulu
Automation facilitates data analysis in condition monitoring
The sustainable and safe use of rotating machines can be enhanced by condition monitoring. Condition is often measured indirectly by using accelerometers but the analysis of measurements can be complicated. Data-driven methods can enhance the time management and accuracy of analysis but their implementation is challenging in real measurement environments.
In this thesis, automated methods were developed to facilitate the implementation of condition monitoring algorithms. Additionally, methods were studied for the automated detection of anomalies in acceleration signals. The methods were studied based on measurement data from azimuth thrusters and a roller leveler, and based on data from rolling element bearings in laboratory and simulation tests.
The results indicated that automated selection of training samples systematized the identification of models and their operating areas. Automated feature selection also revealed previously unknown dependencies between the acceleration signals and operating parameters of a machine. In addition, certain patterns of local faults in rolling element bearings could be detected automatically from short time series that contained only a few fault impulses. The results of this dissertation can be utilized in condition monitoring applications in real measurement environments, where adaptive and automated methods are required.
In this thesis, automated methods were developed to facilitate the implementation of condition monitoring algorithms. Additionally, methods were studied for the automated detection of anomalies in acceleration signals. The methods were studied based on measurement data from azimuth thrusters and a roller leveler, and based on data from rolling element bearings in laboratory and simulation tests.
The results indicated that automated selection of training samples systematized the identification of models and their operating areas. Automated feature selection also revealed previously unknown dependencies between the acceleration signals and operating parameters of a machine. In addition, certain patterns of local faults in rolling element bearings could be detected automatically from short time series that contained only a few fault impulses. The results of this dissertation can be utilized in condition monitoring applications in real measurement environments, where adaptive and automated methods are required.
Last updated: 23.1.2024