| Autor: |
LUPEA, Iulian, LUPEA, Mihaiela |
| Předmět: |
|
| Zdroj: |
Proceedings of the Romanian Academy, Series A: Mathematics, Physics, Technical Sciences, Information Science; Apr-Jun2022, Vol. 23 Issue 2, p151-160, 10p |
| Abstrakt: |
A fault detection system based on vibration analysis and machine learning techniques is proposed in this paper. A test rig comprising a flexible rotor supported by oscillating ball bearings with a central disc driven by a DC motor and a timing belt has been build. Six artificial faults are imposed on the central disc and on the timing belt transmission. The input dataset is a balanced one, containing 80 observations for each of the seven test rig health states (classes), corresponding to one healthy state and six faults. Twenty one features (in time and frequency domains) are extracted from only one uniaxial accelerometer and the tachometer sensor. An in-depth data analysis, by applying supervised and unsupervised processing techniques, aims at selecting the most relevant features used further as predictors in the multi-class classification task. A large set of classifiers from Matlab were trained and tested in order to find the best classification model that predicts the seven health states. The best results were provided by Quadratic Discriminant and Neural Network (Wide) with the accuracy 95% and 94% respectively. [ABSTRACT FROM AUTHOR] |
| Databáze: |
Complementary Index |
| Externí odkaz: |
|
