An Intelligent Method for Bearing Anomaly Detection

Anomaly detection of machinery is a crucial aspect of predictive maintenance. It involves monitoring specific parameters related to the machinery’s condition to identify significant changes that indicate a potential fault. Bearings, which are important components in rotary machinery, are responsible for over 40 percent of rotary machinery failures. Consequently, researchers have been focusing on detecting anomalies in bearings. While there are various methods available for bearing anomaly detection, most of them require high engineering expertise to detect faults. Alternatively, techniques solely based on machine learning models are not always reliable and accurate. Additionally, detecting weak faults accurately has been challenging due to the complexity of bearing vibration signals. As a result, there is a lack of an intelligent method that can automatically and precisely detect faults. In this project, our goal is to develop an intelligent anomaly detection model, and conduct multiple validation tests using various datasets, including three public datasets, one lab dataset generated by the applicants, and industrial data from our partner organization to achieve an accurate, reliable, and robust model.

Faculty Supervisor:

Xihui Liang;Parimala Thulasiraman;Carson Leung

Student:

Partner:

StandardAero

Discipline:

Engineering

Sector:

Aerospace; Artificial Intelligence; Energy and Utilities

University:

University of Manitoba

Program: