Machine learning-based rock characterisation models for rotary-percussive drilling

Abstract Vibro-impact drilling has shown huge potential of delivering better rate penetration, improved tools lifespan and borehole stability. However, being resonantly instigated, the technique requires a continuous quantitative characterisation drill-bit encountered rock materials in order to maintain optimal performance. The present paper introduces non-conventional method for downhole using measurable impact dynamics machine learning algorithms. An impacting system that mimics bit-rock actions is employed this study, various multistable responses have been simulated investigated. Features from acceleration signals were integrated with operated parameters methods develop intelligent models capable quantitatively characterising strength. Multilayer perceptron, support vector regression Gaussian process networks explored. Based on performance analysis, multilayer perceptron showed highest real-time considered features.