Original title: Ball Mill Fault Prediction Based on Deep Convolutional Auto-Encoding Network
Authors: Xinkun Ai, Kun Liu, Wei Zheng, Yonggang Fan, Xinwu Wu, Peilong Zhang, LiYe Wang, JanFeng Zhu, Yuan Pan
This article tackles the challenge of detecting faults in ball mill bearings, crucial components in mining operations. By employing Deep Convolutional Auto-encoding Neural Networks (DCAN), it explores an anomaly detection method utilizing vibration data gathered during normal mill operations. Unlike traditional supervised methods, this approach doesn’t require labeled data, overcoming common issues like data imbalance. DCAN comprises specialized modules for extracting and reconstructing features from the vibration data. The study showcases the practical application of DCAN for detecting faults in ball mill bearings, using real-world data from Wuhan Iron & Steel Resources Group and NASA’s bearing vibration dataset. Results confirm the DCAN model’s effectiveness in identifying abnormal vibration patterns linked to bearing faults. This innovative approach shows promise in bolstering fault detection, reducing operational disruptions, and cutting maintenance expenses in mining operations.
Original article: https://arxiv.org/abs/2311.13571