Research Projects

India’s transition toward sustainable technologies and energy security has spurred a rising demand for critical minerals such as rare earth elements, lithium, and graphite. These minerals are pivotal in clean energy applications, but their beneficiation remains challenging due to complex ore textures and fine particle liberation characteristics. Among the key unit operations used in their processing, hydrocyclones play a critical role in particle classification and desliming. However, their performance is highly sensitive to feed variability, wear, and operational fluctuations—factors that often go undetected in real time with conventional monitoring tools. This project proposes a novel vibrometry-based monitoring framework to enable near real-time prediction of hydrocyclone separation efficiency in critical mineral beneficiation circuits. Vibrometry captures vibration signals induced by internal flow conditions. These signals inherently carry dynamic information about flow regime shifts, particle behavior, and equipment wear. Despite its cost-effectiveness and non-invasiveness, vibrometry remains largely underutilized in mineral processing for predictive diagnostics. Hypothesis: Structural vibration signals captured externally from a hydrocyclone contain extractable features that correlate strongly with its internal separation performance (d₅₀, Ep), and these correlations can be effectively modeled using machine learning algorithms to predict the separation efficiency in near real time. To test this, the project will: 1. Develop a vibration-based sensing setup using MEMS accelerometers positioned at critical locations (spigot and vortex finder) 2. Operate a laboratory-scale hydrocyclone with critical mineral slurries under varying feed pressure, solids concentration, and particle size distribution; 3. Continuously acquire and analyze vibration signals using time and frequency domain techniques (RMS, PSD peaks, frequency peaks) 4. Sample underflow and overflow streams to calculate d50 and Ep as performance benchmarks 5. Train supervised machine learning models to predict performance indicators using extracted vibration features. 6. Validation will be conducted across different ore types, including monazite, graphite, and ilmenite-bearing feedstocks. Comparative analysis of predicted vs. measured values will be used to assess model accuracy. Expected outcomes include: • A lab-scale vibrometry-integrated hydrocyclone monitoring system; • Machine learning models capable of near real-time prediction of separation efficiency; • A benchmark dataset for future studies on vibration-based diagnostics in mineral processing • A user-friendly monitoring interface for alert generation and performance tracking. This research will advance the integration of intelligent sensing and data-driven diagnostics in mineral processing, opening pathways for predictive control and smarter resource utilization—particularly critical for strategic mineral beneficiation.