Sensing and computational frameworks for improving drill-string dynamics estimation

Jean Auriol (1, 2) , Nasser Kazemi (3) , Silviu-Iulian Niculescu (1, 2, 4)

1 L2S - Laboratoire des signaux et systèmes
2 CNRS - Centre National de la Recherche Scientifique
3 University of Calgary
4 Inria Saclay - Ile de France

In this paper, we consider the axial motion of a directional multi-sectional drill-string. The drillstring dynamics are represented by a distributed dynamical model (wave equations) coupled with an ordinary differential equation at the downhole boundary (bit-rock interaction). The interaction between the drill-bit and rock can introduce severe vibrations in the drill-string and result in safety and performance issues. Consequently, the performance of drilling is interwoven with our knowledge of the subsurface. To address these problems, we propose a sensing and computational framework for estimating the drill-string dynamics and the specific intrinsic energy of the rocks while drilling. By exploiting the derived models' particular structure, we combine the drill-string dynamics modeling with top-drive hook-load (force) and hook-speed (velocity) measurements to estimate the force-on-bit without requiring the knowledge of the sub-surface. Then, we record and model the seismic radiation patterns of drill-bit rock interactions near the surface (i.e., seismic while drilling). Such an idea allows deriving an appropriate estimation of the rocks' intrinsic energy while drilling. We introduce two alternative rock property estimation algorithms based on direct parameter estimation and machine learning concepts to complete the analysis. The different approaches are tested and validated in simulations. We discuss their respective advantages and drawbacks. Finally, we show how to extend our methodologies in the presence of non-linear Coulomb friction terms and of coupled axial-torsional oscillations.

Mots clés: Dynamical system; Distributed system; backstepping; Seismic-While-Drilling; Parameter estimation; Neural Networks

Domaines: Equations aux dérivées partielles [math.AP]; Géophysique [physics.geo-ph]; Machine Learning [stat.ML]

Jean Auriol, Nasser Kazemi, Silviu-Iulian Niculescu. Sensing and computational frameworks for improving drill-string dynamics estimation. Mechanical Systems and Signal Processing, 2021, ⟨10.1016/j.ymssp.2021.107836⟩. ⟨hal-03167464⟩