Calculating a Tortuosity Index Metric Using Machine Learning Techniques

Extremely tortuous wells pose many wellbore quality repercussions and poorly affects several well drilling and production-based operations. To date, many indices have been developed for accurate tortuosity identification, but few have had the capability to efficiently mirror and quantify micro-tortuosity in real-time. This study applies a previously-proposed novel algorithm studied by some researchers to quantify well trajectory tortuosity using simple and readily available survey data. The process is followed and validated using twenty wells located in the Permian Basin. Python code was written to identify proper inflection points at the mid-point of the curve turns and using inclination and azimuth indices, a 3D overall TI index was generated for each well. The technique is inspired from the discipline of ophthalmology, specifically a method to determine tortuosity from retinal blood vessels. The approach successfully produced a tortuosity metric with three different risk categories characterizing three ranges of the index. The indices generated were matched against operator reports of drilling incidents and NPT. The methodology matched highly tortuous wells with greater downhole tool failures rates ranking it in the high-risk zone.