I N NOVATI N G WH I LE DR I LLI N G
UT Austin’s automated forensics process
relies on software to accurately evaluate
bit damage and failure so that changes
can be made to the bit and/or BHA before
the next run.

every blade, distinguishing if the damage is in the cone, nose or
shoulder, how much is the loss area, and what kind of damage
occurred.” The next step, which could be enabled via Trax’s bit
forensics, is an even more detailed analysis on a per-cutter basis,
down to the wear or damage seen at a micro level.

The final objective is to improve overall quality assurance,
which is critical in iterative bit design improvements. “When we
analyzed one bit with a depth-of-cut (DOC) limiter, for example,
we noted a difference in the visualization between what it was
supposed to be and what it actually was,” Mr Schmitz noted.

“When we’re talking about the very precise designs that they’re
trying to develop now, in terms of limiting the DOC to exactly
what they want in order to avoid damage while maximizing DOC
to drill as fast as possible, differences of that magnitude start to
become important.”
Drill bit failure forensics using field
photography Drill bit forensics is also a topic of study in the world of aca-
demia. An ongoing research initiative at the University of Texas
(UT) at Austin involves the development of a software algorithm
that can automatically analyze photos taken of the bit at the rig
site and identify, from those photos, the root cause of bit damage
and failure. The goal of the project is for the software to be able to
accurately evaluate a used bit so that changes can potentially be
made to the BHA and/or bit before the next run .

The methodology for this automated forensics process involves
four steps. First, the algorithm is given a set of drill bit photo-
graphs that clearly show individual blades, allowing the software
to identify all the cutters on that bit. The software then quanti-
fies the damage to each cutter, drawing on a database of surface
sensor and downhole vibration data, as well as offset well rock
strength information, to characterize drilling dysfunction relative
to the damage seen on the bit. After calculating cutter location, the
software then uses a classifier to determine the average damage
in various parts of the blades, thereby enabling it to infer the root
cause of damage.

Jian Chu, a PhD student working on bit forensics research,
explained that the initiative continues to advance. Previous
versions of the software were not always able to detect all the
cutters based on the photographs available and lacked precision.

Additionally, not all damage could be calculated, and when it was,
it was only quantified as a whole number. Further, the algorithm
was affected by lighting, and the forensics process did not include
EDR data. However, the team is now focusing on identifying the
damage type (such as worn, chipped) and improving the precision
of damage grading via semantic segmentation.

Mr Chu said it was critical that noise from the drilling environ-
ment is removed in order to develop something truly useful. “One
big issue we had is that we didn’t have a lot of data,” he explained.

“For traditional machine learning algorithms, they would need
millions of data points to train the neural network. We didn’t have
that much data, so we had to remove all the noise and really focus
on what we had.”
Through precise area isolation, damage categories identifica-
tion and integration of expert systems, the process will eventually
be entirely automated, with the algorithm becoming more and
more accurate. Since the photography used to drive the algorithm
is still prone to human error, Mr Chu said that maybe one day,
computer vision on the rig floor could take photographs of the bit
as the BHA is pulled, capturing all angles and feeding that data to
the algorithm.

There is additional potential should a company choose to
develop an application for the algorithm that could be deployed
on a smartphone or tablet, which would make the process even
easier and reduce the concerns associated with limited connec-
tivity on remote sites. For now, Mr Chu said he isn’t worried about
the commercial potential of the platform. He is more interested in
advancing the field of drill bit forensics: “I want to make some-
thing useful for the industry .” DC
D R I L L I N G C O N T R AC T O R • M A R C H/A P R I L 202 2
19