Signal processing and feature extraction for automated machine diagnostics: methodology and experiments

Thesis event information

Date and time of the thesis defence

Topic of the dissertation

Signal processing and feature extraction for automated machine diagnostics: methodology and experiments

Doctoral candidate

Master of Science (Technology) Konsta Karioja

Faculty and unit

University of Oulu Graduate School, Faculty of Technology, Intelligent Machines and Systems

Subject of study

Machine diagnostics

Opponent

Professor Matti Vilkko, Tampere University

Custos

Professor Enso Ikonen, University of Oulu

Add event to calendar

Methods for automation of machine diagnostics

The objective of the study in this doctoral thesis was to determine methods suitable for automated machine diagnostics. Successful automated diagnostics requires methods that are sufficiently sensitive to faults, and often the capability of identifying the fault and keeping track of its progression is desirable.

The methods introduced in the thesis allow e.g. automatic determination of the frequency range in which the observed changes have taken place. Separating overall level changes from phenomena that cause shock-like vibration is also possible. Moreover, the proposed stress evaluation methodology can be utilised to monitor cumulative stress levels and generate forecasts of the risk level for the faults that occur. The thesis includes an overview of methods and study of their applicability through practical experiments. The primary focus of this study is on vibration measurements, but the methods can be applied to any type of signal.

Succesful diagnostics provides opportunities for optimising maintenance actions. Proper maintenance significantly improves the reliability and safety of machinery. In addition, well-performed maintenance can save energy.
Last updated: 20.3.2025