Anomaly Detection in Industrial Machinery Using Machine Learning and Deep Learning Techniques with Vibration Data for Predictive Maintenance

Authors:

Abstract
Industrial machinery often operates under harsh conditions, where early de-tection of anomalies is crucial for avoiding unplanned downtime and costly repairs. This study explores the application of machine learning and deep learning techniques for anomaly detection in industrial machinery using three-axis vibration data. By utilizing data collected before and after sched-uled maintenance, three models One-Class Support Vector Machine (SVM), Isolation Forest, and Long Short-Term Memory (LSTM) Autoencoder are trained to identify anomalies in the vibration signals. The models are trained on normal operating data and tested against a dataset containing both normal and anomalous conditions. The One-Class SVM achieved the highest anomaly detection accuracy at 99.5%, outperforming the LSTM Autoencod-er with 94.1% accuracy and the Isolation Forest with 92.9% accuracy. The results highlight the efficacy of machine learning in predictive maintenance applications, providing a robust approach to early fault detection in industrial equipment.

This publication has 14 references indexed in Scilit: