A review of recent advances in vibrational signal processing in the minerals industries

Vibrational signal processing is becoming an increasingly important source of information for monitoring, diagnosis, and control in mineral processing circuits. While early applications focused mainly on equipment condition monitoring, recent advances in sensing, signal processing, and artificial intelligence have enabled vibration data to be used more directly for process monitoring, soft sensing, and control supervision. This paper reviews the state-of-the-art in vibrational signal analysis with specific emphasis on mineral processing systems, including comminution circuits, hydrocyclones, industrial screens, and fluidised beds. Beyond a purely data-driven perspective, the review emphasises the interpretation of vibration signals as projections of underlying process dynamics in complex multiphase systems. This perspective is used to relate signal characteristics to physical phenomena such as particle interactions, flow regimes, and equipment–material coupling. Developments in time–frequency analysis, empirical and variational mode decomposition, and deep learning have reshaped feature extraction and modelling strategies for non-stationary vibration signals. Supervised, unsupervised, and self-supervised learning frameworks, including convolutional neural networks, autoencoders, and transformers, are critically examined for their roles in fault diagnosis, anomaly detection, and operational state classification. Importantly, the paper highlights key practical and scientific challenges that limit wider industrial adoption, including domain shift across operating conditions, data limitations, model interpretability, and the integration of vibration-derived information into process control systems. Differences in the maturity of applications across equipment types are also discussed, with comminution systems and hydrocyclones being more extensively studied than industrial screens and fluidised beds, which are identified as important areas for future research. Industrial patents and commercial developments are reviewed to assess progress toward real-time deployment. The paper concludes by outlining a set of open research questions and directions aimed at enabling more robust, interpretable, and control-oriented use of vibration signals in mineral processing.