Organization and Analysis of Measurement While Drilling (MWD) Data
Obtaining sufficient and reliable in-situ geologic substrate data and characterizing the subsurface conditions for engineering design purposes has always been a challenge to the natural resources and civil engineering industries. Availability and accuracy of such information is key, however, for successful planning, design, construction, and operation of many engineering projects including transportation infrastructure.
Measurement While Drilling (MWD) technology has shown a lot of potential for improving the subsurface characterization process in some industries. Since the 1980s, for example, MWD has been critical to the development of directional drilling within the petroleum industry. In the geotechnical engineering industry, however, MWD technology is in its early research stages.
Utilizing a $50,000 contract funded in early 2020 through FHWA's Every Day Counts (EDC) 5 Initiative, the Montana Department of Transportation (MDT) is currently evaluating the MWD technology on their Central Mine Equipment (CME) 1050 ATM drill rig. For the past several months, MDT has been collecting continuous and consistent measurements of MWD data at several of their projects. The collected data include drilling depth, drilling rate, rotation speed, own pressure, hold-back pressure, mast vibration, flow rate, and fluid pressure. Beginning this spring, MDT will continue to collect more MWD data with an attempt to also collect accurate mechanical torque data. It is worth mentioning that other data including the standard penetration test (SPT), vane shear test (VST), cone penetration test (CPT), as well as geophysical survey data will also be collected. This data will be collected at MDT project sites that have proposed ruts, embankment fills, culverts, and bridge foundations. The projects from which MDT chooses to collect MWD data will be located throughout Montana. The challenges with MWD technology include a combination of organizing large amounts of collected data and correlating this data to the desired subsurface characteristics such as the subsurface soil and rock strength parameters. Finding meaningful and reliable correlations is especially challenging as the multivariable nature of such correlations will not allow the simple regression analyses to be used. These challenges could be addressed by creating a database and using Machine Learning (ML) methods such as Artificial Neural Networks (ANNs), and Deep Learning (DL) algorithms.