Vibration-based Anomaly Detection on Weather Radar Rotating Machinery using One Dimensional Convolutional Autoencoder

Weather radar systems are vital components of meteorological monitoring, especially in regions prone to adverse weather phenomena and near critical infrastructures like airports. The rotating machinery within these radar systems plays a vital role in their operation. Vibration analysis has emerged as a valuable tool for monitoring the health and performance of rotating machinery in weather radar systems. Detecting anomalies in the vibration patterns of rotating machinery within weather radars is essential for early fault detection and preventing costly breakdowns that could disrupt weather monitoring and forecasting. This paper presents a one dimensional convolutional autoencoder (1D-CAE) for vibration anomaly detection within weather radar systems, aiming to reduce the frequency of manual inspections and maintenance costs. The proposed architecture learns hierarchical feature representations by reducing noise in high-dimensional vibration signals. The comparison between 1D-CAE and other conventional anomaly detection algorithms illustrates the effectiveness of 1D-CAE for vibration anomaly detection.