Rock Characterization Using Time - Series Classification Algorithms

Development of measurement while drilling (MWD) systems for ground control applications in mining applications has recently been an active research topic. The main goal is to utilize the drilling information for characterizing the ground condition. Such characterization is the key to the accurate and efficient mapping of hazards and proper planning of ground support. However, most of the existing drilling units and measurement while drilling systems are mainly developed for joints and voids detection, and much less research is done to estimate rock strength. This paper focuses on rock classification based on estimated strength from the drilling data. For this purpose, data from J.H. Fletcher roof bolters instrumented with the vibration and acoustic sensors as well as torque, thrust, penetration rate, RPM and position signals is used. All the available data was utilized for rock classification using pattern recognition algorithms. The algorithms were developed based on data from drilling, and into three different rock types. The classification is performed using a simple feature extraction algorithm with a well-known pattern recognition algorithm. The results demonstrate the suitability of the proposed algorithms in identifying the rock types based on their strength properties, which can be adopted in measurement while drilling systems.