Anomaly detection tool for gas turbine condition monitoring under variable operating and environmental conditions

Different industries have widely employed condition monitoring to evaluate their system’s health. For gas turbines, current methods have limitations related to certain operating and environmental conditions and are mainly based on supervised approaches requiring annotated data that is hard to acquire. This project will focus on developing an unsupervised anomaly detection mechanism that alarms users about a potential fault in the machine. The false alarm in typical models is big concern and, therefore will be addressed by a hybrid approach. The developed mechanism will be tested on datasets before implementation in real-life. Developing such tool will help reduce maintenance costs and increase the airliner’s profit. This product can be improved in the next phase by adding modules for fault diagnosis, and prognosis and sold as a maintenance support tool after commercialization. Lab2Market and Mitacs Accelerate will also foster the growth of innovative technologies within the Canadian economy.

Faculty Supervisor:

Xihui Liang

Student:

Partner:

North Forge

Discipline:

Engineering

Sector:

Education; Management of companies and enterprises; Professional, scientific and technical services

University:

University of Manitoba

Program: