Drill Cuttings Characterization During In-Line Mud Logging Using Acoustic Sensors

Abstract This study presents a low-cost, non-invasive method for real-time diagnostics of multiphase flow in drilling systems using clamp-on MEMS microphones. The technique is based on the analysis of acoustic emission (AE) signals produced by the interaction of solid and water particles in a flow. The definition of characteristic grain size through solid type classification and the estimation of the solid mass flow rate are the two primary tasks that are addressed. A hybrid signal processing pipeline is proposed for the classification of sand types. It combines Wavelet Decomposition (DWT) to extract statistical features of AE signals, which are then fed into a linear classification model. Welch's method is used to calculate the AE signal's Power Spectral Density (PSD) in order to estimate the mass flow rate. Band-wise root mean square (RMS) values of the PSD are used as regression features. To validate the suggested method in a controlled laboratory setting, an experimental testbed was created. The classification model achieved a maximum F1-score of 0.917 after testing five different flow conditions, including a stream of pure water and four types of sand. For mass flow rate estimation, the method was tested across a range from 0 to 250 g/liter of solid, with predictions reaching a mean absolute error below 5 g/liter.