Detection of roller bearing faults at different rotational speeds using envelope analysis features and support vector machine

Intelligent fault diagnosis is important for modern maintenance planning to enhance rotating machinery efficiency and reliability. Recently, is a critical issue in developing a general diagnosis technique independent of the operational constraints of machines especially the rotational speed variatio...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Najim, Haider Suhail, Ali, Jaafar Khalaf
Format: Tagungsbericht
Sprache:eng
Schlagworte:
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Intelligent fault diagnosis is important for modern maintenance planning to enhance rotating machinery efficiency and reliability. Recently, is a critical issue in developing a general diagnosis technique independent of the operational constraints of machines especially the rotational speed variations between training and testing domains. In this study, a different-rotational-speeds diagnosis approach based on employing features obtained from time and frequency domains codded as inputs to Support Vector Machine (SVM) is proposed. Envelope technique is used to capture more distinguishing features at basic fault frequencies from the original signal, while several time domain parameters are calculated from the original and filtered signals. The diagnosis outcomes were achieved experimentally through three data set collected at three different rotational speeds and each one has 400 samples with different levels of fault categories. In addition, the overall learning process is divided into two types based on the features kind are used, that is, all the time frequency features and only the captured frequency features. Finally, the practical experimental results indicate by using all features the diagnosis accuracy was between (92.5%-99.75%) while by just the frequency features a better and more stable accuracy rate was between (99%-99.5%) for different rotational speeds. Moreover, the proposed model can effectively recognize faults even when the SVM model is learned at a certain rotational speed and then employed to detect faults at another different speed.
ISSN:0094-243X
1551-7616
DOI:10.1063/5.0191535