Roman Shor

University of Calgary

Estimating formation properties and bit dynamics in realtime using physics and data driven methods

Estimation of the behavior of long dynamic systems with limited sensing remains an open question. In this presentation, I will consider the rotational behavior of a deep drilling system and compare several methods to estimate, in real time, the drillstring behavior, along-string friction, and bit-rock dynamics using only available surface data.  Along string friction is estimated as a Coulomb source term arising from interaction of the drillstring with the borehole wall, and bit-rock dynamics are assumed to be either energy based or frictional cutting based.  An observer framework will be shown that utilizes the physics based model to estimate these parameters and convergence will be discussed.  A second methodology, using data driven methods will be explored, both as a method to estimate these parameters as well as a technique to provide an initial guess for the physics based approach.  Data, from FORGE and other wells, will be used as validation.