Vibration analysis using deep learning for predictive maintenance

Goh, Qi Chen (2024) Vibration analysis using deep learning for predictive maintenance. Project Report. Universiti Teknikal Malaysia Melaka, Melaka, Malaysia. (Submitted)

Text (Full Text) Vibration analysis using deep learning for predictive maintenance.pdf Download (4MB)

Abstract

In the era of Industry 4.0, predictive maintenance through vibration analysis has gradually improved across various industrial sectors, especially those incorporating rotating components like bearings and shafts. Conventional maintenance methods, whether reactive, preventive, or proactive, often entail high risks and costs. Reactive maintenance, in particular, results in significant drawbacks, causing downtime, resource wastage, and substantial monthly repair costs. The project proposes employing a Long Short-Term Memory (LSTM) autoencoder deep learning model for predictive maintenance, utilizing acceleration data collected from accelerometers. This data undergoes preprocessing before being fed into the LSTM autoencoder model, implemented using Python with TensorFlow and Keras frameworks in Jupyter Notebook. The project concludes with the LSTM autoencoder model demonstrating low losses (0.0017), highlighting its effectiveness in flagging anomaly conditions and its potential to enhance predictive maintenance in industries with rotating machinery components.

Actions (login required)

View Item