Graphics enhanced version of this site

Lithological Classification By Drilling
D. LaBelle
tech. report CMU-RI-TR-01-27, Robotics Institute, Carnegie Mellon University, 2001.

Jump to: Download | Abstract | Text Reference | BibTeX Reference

Adobe portable document format (pdf) [509 KB]
Compressed postscript (ps.gz) [588 KB]

Copyright notice: This material is presented to ensure timely dissemination of scholarly and technical work. Copyright and all rights therein are retained by authors or by other copyright holders. All persons copying this information are expected to adhere to the terms and constraints invoked by each author's copyright. These works may not be reposted without the explicit permission of the copyright holder.

There are many drilling tasks in which drill monitoring is used to improve the quality of a product: detecting tool breakage in manufacturing drilling, exploratory drilling for oil and natural gas reservoirs, collecting soil samples on Mars with a robotic drill. However, in many applications, a human is partially or entirely responsible for controlling and analyzing the interaction between the drill bit and the drilling medium. This research is exploring intelligent drilling that can be applied to multiple applications.

I propose to develop a methodology with which to build a working lithological classifier for different drilling applications. The methodology is a template with which a classifier can be built using the proprioceptive sensors of a drill, and it includes: (1) acquiring drill hole sensor data; (2) using expert knowledge from a mine or drill operator to generate a feature set from drill sensors; (3) automating the selection of a subset of drill features which maximize classification rates and minimize the sensitivity of the classifier to drill and environmental variables; (4) using machine learning to classify rock and detect material boundaries; and (5) extending the classifier to operate with a data stream. The methodology and it's adaptability will be tested with coal mine drilling, but another application, such as well drilling or surgical drilling, will be investigated.

The methodology uses a neural network to classify material lithology where the inputs to the neural network are sensed drill parameters such as thrust, torque, rotary speed and penetration rate, as well as information derived from these sensors over time. In preliminary experiments, five different layers of concrete and three layers of coal mine strata were classified using the sensors attached to a coal mine drill. These results suggest that drill parameters can be used to classify rock strata, and that using additional features derived from the drilling parameters significantly improves classification accuracy.

D. LaBelle, Lithological Classification By Drilling, tech. report CMU-RI-TR-01-27, Robotics Institute, Carnegie Mellon University, 2001.

@techreport{LaBelle_2001_3888,
   author = "Diana LaBelle",
   title = "Lithological Classification By Drilling",
   institution = "Robotics Institute, Carnegie Mellon University",
   year = "2001",
   number = "CMU-RI-TR-01-27",
   address = "Pittsburgh, PA"
}