The prediction of failures and the estimation of the health status of components are two critical functions of predictive maintenance. This thesis is the subject of an investigation that aims to predict several defects using vibration signals captured from rotors bearing by MEMS accelerometers sensors. These machine components are widely used in the industrial world, particularly in electric motors. Moreover, artificial intelligence and signal collection have greatly improved, especially with the advent of the Industrial Internet of Things. The use of recorded signals in the prediction of failures and the estimation of the remaining useful life span are currently trending subjects of research in the field of industrial maintenance.
This project is divided into two parts. The first part of the research is conducted on an opensource data base to thoroughly evaluate the models for anomaly detection and classification, as well as to predict the health of the bearings. A second section is devoted to designing a test bench to record vibration data from two bearings using two MEMS accelerometers Adxl335 and Adxl1002Z in order to evaluate their performance in the context of predictive maintenance.
We were able to predict defects before their appearance using the signals recorded by the two accelerometers with a maximum F-score of 98.26% for fault classification using the SVM model. Additionally, a methodological approach has been adapted to this problem for estimating the state of health of these components.
| Date | 16 Mar 2023 |
|---|
| Original language | French |
|---|
| Awarding Institution | - École de technologie supérieure
|
|---|
| Supervisor | Ricardo Izquierdo (Supervisor) |
|---|
Jemli, M. (Author),
Izquierdo (Supervisor),
16 Mar 2023Student thesis: Master's thesis › Master in Engineering: Electrical Engineering