You are on page 1of 18

SPE/IADC-184718-MS

Real-Time Borehole Condition Monitoring using Novel 3D Cuttings Sensing


Technology

Runqi Han, Pradeepkumar Ashok, Mitchell Pryor, and Eric van Oort, The University of Texas at Austin; Paul Scott,
Isaac Reese, and Kyle Hampton, ConocoPhillips

Copyright 2017, SPE/IADC Drilling Conference and Exhibition

This paper was prepared for presentation at the SPE/IADC Drilling Conference and Exhibition held in The Hague, The Netherlands, 14–16 March 2017.

This paper was selected for presentation by an SPE/IADC program committee following review of information contained in an abstract submitted by the author(s).
Contents of the paper have not been reviewed by the Society of Petroleum Engineers or the International Association of Drilling Contractors and are subject to correction
by the author(s). The material does not necessarily reflect any position of the Society of Petroleum Engineers or the International Association of Drilling Contractors,
its officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written consent of the Society of Petroleum Engineers or
the International Association of Drilling Contractors is prohibited. Permission to reproduce in print is restricted to an abstract of not more than 300 words; illustrations
may not be copied. The abstract must contain conspicuous acknowledgment of SPE/IADC copyright.

Abstract
Wellbore instability and stuck pipe incidents are large contributors to drilling-related non-productive time
(NPT). Drilling cuttings/cavings monitoring is crucial for early detection and mitigation of such events.
Currently, monitoring is done manually and lacks a streamlined approach. Automating this process would
be very beneficial, and is possible due to recent advances in sensing technology. Real-time cuttings/cavings
monitoring can be used to quantify cuttings volume, measure size distribution, and analyze shape. By
correlating these measurements with ongoing drilling operations, the hole condition (in particular hole
cleaning/cuttings transport efficiency, wellbore stability situation, etc.) can be automatically assessed in
real-time. This makes pro-active prevention and mitigation of NPT related to hole cleaning and wellbore
instability possible.
In this paper, we detail a system designed and prototyped to allow us to measure cuttings/cavings in real-
time. A highly portable device employs a 2D high-resolution camera and a 3D laser sensor to determine
the physical properties of cuttings. The 3D point cloud/depth data obtained by this device provides cuttings
size distribution, volume and shape characteristics. Comparisons and discrepancies between expected and
sensed quantities can then be used for alarming purposes and taking appropriate corrective action.
A prototype experimental setup was constructed to evaluate the ability to quantify relevant cuttings
properties and profiles in the presence of drilling fluids. In a controlled environment, the cuttings
slide down a shaker table's clearing chute while simulating various realistic external variable scenarios.
The environmental impact on the accuracy, repeatability and robustness of the various sensors under
investigation was determined to identify the sensors best suited for the task at hand. The optimum device
configuration was then implemented and evaluated to verify that the system is viable for use in the field.
The automated cuttings monitoring system can warn drillers to potential hazards associated with poor hole
cleaning conditions, ongoing wellbore breakout, and the likelihood of stuck pipe events.

Introduction
In oil and gas well drilling, inefficient cuttings transport may result in poor hole cleaning problems. Such
problems can be exacerbated by wellbore instability which result in hole enlargement, lower annular
2 SPE/IADC-184718-MS

velocities and large cavings to be cleaned out of the wells. Poor hole cleaning can lead to a series of costly
drilling inefficiencies such as mechanical pipe sticking, excessive torque and drag on drill string, slower
rate of penetration, bottom hole assembly damage, poor cement job, etc. (OCTG Procter 2000). Incidents
like wellbore instabilities and stuck pipe can halt drilling operations for days and cost millions of dollars
(C. Aldea 2005). All these problems may also jeopardize the safety of drilling personnel on a rig.
Monitoring solids returned from the well provides a direct assessment of hole cleaning efficiency. One
of the traditional methods to assess the efficiency of cuttings recovery is by assigning a mud logger
to continuously monitor cuttings on a shale shaker system (Morton-Thompson and Woods 1993). This
individual generates a detailed record of wellbore formation and condition, with information on drilling
cuttings properties and drilling fluid conditions. The physical lithological data of downhole formations can
be recovered by analyzing collected cuttings and cavings samples with regard to their physical appearances.
However, this is a tedious and labor-intensive process, and it lacks a thorough and structured approach.
Cuttings return volume also cannot be measured quantitatively based on a simple visual check from the mud
logger. In addition, a mud logger is continuously exposed to high hydrocarbon mist and vapor at the shale
shaker house. Because of the routine duties such as washing with high-pressure guns using hydrocarbon-
based fluid, collecting cuttings, and measuring drilling mud, workers’ health and safety are affected both
by inhaling vapors and direct skin contact (IPIECA 2009). Therefore, automating this process will be very
beneficial to both the drilling operation as well as personnel health and safety.
There has been a growing trend in drilling automation during the last decade. Focus has been on real-time
data monitoring for safety improvement and drilling performance optimization. With the fast development
of surface and downhole sensors, huge amounts of data are available for analyzing the performance of
drilling operations. This work will present a unique computer vision system to monitor cuttings and cavings
on the shale shaker system in real-time. This system can thus assist mud loggers and drillers to assess the
downhole condition in real-time. In this research study, the objective of the proposed design is to quantify
volume, analyze size distribution, and regenerate surface profiles of moving cuttings in real-time. The results
presented here provide a controlled baseline upon which to evaluate the effectiveness of the approach in
the field.

Historical Review
Cuttings and cavings monitoring while drilling is of great importance for examining hole cleaning efficiency
and diagnosing wellbore instability. As indicated, the most common and traditional practice to analyze
drilling cuttings is that a mud logger manually collects and examines cuttings samples at different time
intervals. After a cuttings sample is acquired from the flowline, shaker screen, or at the end of a shale shaker,
the mud logger usually uses a series of sieves to sort out different size of cuttings and exposes the sieves to
sunlight for drying the cuttings samples. The benefit of this simple sieve system is that it is easy to operate
and costs little except for the labor cost. The mud logger can also differentiate cuttings and cavings on the
sieve. As shown in Figure 1, a large chunk caving can easily be distinguished on the sieve. The accuracy
highly depends on how frequently the mud logger performs the measurement and how representative the
collected sample is. If a high percentage of clay is present in the drilling, clay/mud and cuttings tend to
stick together in lumps as shown in the middle of Figure 1. Then it is impossible to collect and filter out
a representative cuttings sample and perform accurate measurement. The main shortcomings of this old-
fashioned procedure are that it requires continuous human intervention and is not able to provide a real-
time, quantitative measurement on cuttings/cavings volume, sizes, and shape.
SPE/IADC-184718-MS 3

Figure 1—Representation of sieve system (M. Karimi 2013)

In order to quantify return cuttings, some Cuttings Flux Meters (CFM) were designed and tested
previously. One of the common methods is to estimate the weight of cuttings flux off a shale shaker via a
collection device and strain gauges. M. Naegal, et al. (1998) presented the automated cuttings flow meter
shown in Figure 2. Such a device measures return cuttings weight at the outlet of the shale shaker. After
the cuttings pass by the shale shaker screen, they accumulate in the collecting gutter. During this period,
the gutter is locked and strain gauges measure the increasing weight of cuttings. After the container fills up
with cuttings (preset weight limit is reached), the gutter flips to remove the cuttings. A pneumatic system is
used to control and activate all mechanisms. After each flip, software recalibrates a new zero weight of the
gutter. The software extrapolates cuttings flow based on previous measurements. Such a design can provide
an accurate real-time cuttings return volume measurement based on known rock density (Naegel, et al.
1998). In 2014, Schlumberger commercialized a real-time cuttings monitoring system with a similar concept
design (Figure 2). A mud-effect correction factor called the Equivalent Dry Cuttings Ratio (EDCR) was
applied to correct the errors caused by the coating of mud over the cuttings. This allows for the volume of
equivalent dry cuttings to be calculated. This real-time system also collects drilling parameters to calculate
theoretical cuttings volume. By comparing the measured cuttings volume and theoretical return volume, the
percentage of cuttings recovery can be used as a factor to reflect hole cleaning efficiency. However, drilling
fluid not only coats the surface of the cuttings, but is also absorbed by porous cuttings. The amount of fluid
invasion in the cuttings depends on mud rheology, lithology of the cuttings, temperature, and interaction
time. Various factors may affect the EDCR accuracy. In addition, this system is not capable of detecting
cavings, analyzing size distribution, or identifying cavings shapes (Ferranod and Hbaieb 2015).

Figure 2—a) Schematic view of the Cutting Flow Meter (left) (Naegel, et al. 1998) and b)
Schlumberger hole cleaning and wellbore risk reduction service (right) (Ferranod and Hbaieb 2015).

A few other non-intrusive techniques approach cuttings measurement through image processing. Marana
et al. (2010) published a paper on real-time monitoring and analysis of cuttings concentration over the shale
shaker screen. A high definition camera was installed above the shale shaker, and captured images were
sent to a computer for further analysis (Figure 3). The system consisted of two modules: image acquisition
and data analysis. During the drilling process, images of cuttings on shale shaker are classified based on
4 SPE/IADC-184718-MS

predefined classes. The analysis focused on quantifying the concentration of cuttings on the surface based
on the data analysis module. This non-invasive measurement approach requires no field workers to operate.
The algorithm analyzes each frame automatically and alerts a field operator if an abnormal pattern is spotted.
A field trial was conducted on an offshore drilling rig, and the system was demonstrated to be able to monitor
cuttings loading on shale shaker in real-time. However, the proposed system and algorithm cannot quantify
cuttings volume, size, or shape. Cavings also cannot be differentiated from cuttings. The concentration of
cuttings can only be used as a Key Performance Indicator (KPI) (Marana, et al. 2010).

Figure 3—Captured images are processed by the monitoring system proposed by Manara et al. (2010)

Graves and Rowe (2013) described a down hole cuttings analysis system, which applies surface computer
vision system to monitor the characteristic features of return cuttings. Their patent briefly describes a system
that employs an imaging device to acquire image/video of cuttings for analysis. 3D facial recognition
algorithm can be used for cuttings size distribution, volume, and shape analysis in real-time. A possible
programmable data acquisition and memory system can be adapted. The video can be streamed to remote
workstation and analyzed data can be synchronized with other drilling operations parameters. This patent
detailed a fully automated computer vision system concept for real-time cuttings monitoring. However, no
detailed 2D/3D vision techniques are described in the patent and no cuttings transport model is mentioned.

Figure 4—Patented concept for surface cuttings monitoring and analysis (Graves and Rowe 2013)
SPE/IADC-184718-MS 5

Although several of the procedures and designs described above were tested, few systems have been
prototyped to perform comprehensive measurements. None of the tested systems can monitor cuttings
volume, size distribution and cavings shape in real-time. Therefore, there still exists a need for a solution
to automate cuttings monitoring to improve drilling efficiency and safety.

Introducing Computer Vision Techniques


Three-dimensional (3D) vision techniques measure the physical distance between a target surface and the
sensor's reference position. Various technologies can be applied for range measurement, 3D modeling,
object detection, and many other sensing applications. In the past decade, with the technological advances
in photonics, electronics, and computer vision, 3D sensing technologies have been widely adapted in range
measurement, 3D scanning, and automation systems. The three major depth sensing techniques are: stereo
vision, time-of-flight, and structured light. These will be discussed briefly in the below sections.

Stereo Vision
This technique employs two Charged Coupled Device (CCD) cameras placed horizontally at a small fixed
distance. By analyzing the differences in the two images, the relative depth can be calculated. This is
similar to human binocular vision. To calculate the depth value Z of target P, UL and UR represents the x-
coordinates of P projection on two image planes, where and where f is the focal length
of the camera. Then a disparity value d is calculated as . Finally, the actual distance
between target P and the stereo sensor can be calculated as (National Instruments 2012). The
depth resolution of the sensor relies on camera's focal length, image resolution, and baseline. Stereo vision
unfortunately requires long computational time in order to apply multiple feature detection to compare two
images. Therefore, it is more suitable to 3D scanning of static objects or scanning at low frequency.

Figure 5—Principle of operation for stereo vision (National Instruments 2012)

Time of Flight
A 3D Time-of-Flight (ToF) camera measures the distance to a target by emitting modulated light to the target
and observing the reflected light. The basic concept is illustrated in Figure 6. Two methods can be used
to calculate the object's distance. One method is measuring the reflection time and calculating the distance
based on the speed of light. The other method is to translate the phase shift between the emitted signal and
received signal to a distance value. Most illumination sources use a solid-state laser or a Light-Emitting
Diode (LED). In order to detect the phase shift between the emitted signal and the received signal, either a
pulsed or continuous-wave modulated signal can be used. The advantages of ToF 3D vision system are fast
6 SPE/IADC-184718-MS

computational speed, high resolution, and low cost. One concern is that if the emitted light wavelength is
in the range of background light, the detection sensitivity may be suppressed (Li 2014).

Figure 6—3D Time-of-Flight camera operation (Li 2014)

Structured Light
The operation principle of structured light is to project a band of light onto a target's surface. A camera
or multiple cameras are used to observe the distorted light pattern on the illuminated surface (David Fofi
2004). The band of light has a predefined pattern as gray codes, light stripes, sine waves, or speckle patterns
(Geng 2011). A single camera structured light system is used as shown in Figure 7a. A projector emits a
beam of patterned light on to an object. Each pixel of the projector has a specific local configuration of the
projected pattern. By comparing with the distorted projection pattern on the target surface, the 3D geometric
shape or profile of the target surface can be computed by various algorithms (Pietro Zanuttigh, et al. 2016).
There are two major methods to generate a strip pattern: the laser interference and the projection techniques.
The laser interference technique emits a single strip of laser light (Figure 7b). The interference has regular
and equidistant dots in the line pattern. Then the height of the target surface is calculated by triangulation
method at precise accuracy. The laser interference can scan at high frequency and is less affected by the
ambient lighting condition. The projection technique emits incoherent light with a coded pattern (Figure
7c). Instead of scanning one line each time, the method takes a "snap-shot" over a target's surface. Compared
to the laser interference technique, the coded projection detection sensitivity is strongly suppressed by the
ambient lighting environment.

Figure 7—a) Illustration of structured light (Geng 2011); b) laser scan based on
triangulation (Gerig 2013) and c) an example of color-coded grids (Geng 2011)
SPE/IADC-184718-MS 7

It is important to understand the working mechanism of each technology in order to select a suitable
vision system to meet the desired accuracy. Table 1 summarizes the advantages and disadvantages of each
system with respect to: depth accuracy, scanning speed, distance range, low light performance, outdoor
performance, software complexity, and cost.

Table 1—3D vision system comparisons (Michael Brading 2016)

Structured Light
Stereoscopic Vision Time of Flight
Fixed Pattern Programmable Pattern

μm to mm
mm to cm mm to cm
Variable patterns &
Depth Accuracy Difficult with mm to cm Depends on
different light sources
smooth surface resolution of sensor
improves accuracy

Medium
Fast Fast/Medium Fast
Scanning Speed Limited by software
Limited by camera speed Limited by camera speed Limited by sensor speed
complexity

Very short to mid range Very short to mid range Short to long range
Distance Range Mid range Depends on Depends on Depends on laser
illumination power illumination power power & modulation

Low Light Performance Weak Good Good Good

Weak/Fair Weak/Fair Fair


Outdoor Performance Good Depends on Depends on Depends on
illumination power illumination power illumination power

Software Complexity High Low/Middle Middle/High Low

Material Cost Low Middle Middle/High Middle

Proposed UT 3D Cuttings Sensing System


Before designing an automated real-time cuttings monitoring system, the first step was to determine the
objectives of such a measurement system: cuttings volume, size distribution, and shape profile. The location
to place the equipment is also crucial, as it should not interfere with drilling operations. In addition, the shale
shaker system presents a dirty environment with events that will challenge the integrity and reliability of
high-end vision systems: high-pressure cleaning wash, drilling mud spills, high vibration levels, presence
of high-temperature oil or water vapor, limited room for installation, and exposure to all types of weather
conditions. According to Occupational Safety and Health Administration (OSHA) in Texas, the area around
a shale shaker system is a hazardous working zone (OSHA 2016). Class I Division I explosion protection
enclosure should be applied for all electronic devices during hydrocarbon drilling activities. Therefore, the
design should meet the following criteria:

• automated real-time measurement,

• minimum number of components and easy to install on current rigs,

• robust to external factors, including the presence of drilling byproducts and inclement weather,

• no interference with ongoing drilling processes, and

• meeting all safety requirements.

After careful examination on a land rig, the best cuttings-monitoring location was chosen to be above the
cuttings ramp, where cuttings initiate their slide down the ramp into a collecting pit at a relatively steady
speed. This area is below the shale shaker floor which is largely isolated from dirty environment and is less
affected by shale shaker vibrations.
8 SPE/IADC-184718-MS

In order to apply adequate 3D depth sensing technologies, a few of commercially available 3D structured
light cameras and 2D/3D laser profile scanners were tested and evaluated for real-time cuttings measurement
feasibility. Sensors were evaluated based on their advantages and disadvantages with regard to field of
view (FOV), measuring speed, accuracy, cost, ability to operate in ambient lighting, and environmental
performance. A brief evaluation of each sensor is presented in the table given in the Appendix. Even though
it is difficult to find a perfect solution to satisfy all design requirements, sensor selection can be narrowed
down on performance. The benefit of 3D structured light cameras is to take a snapshot of the measuring
target and generate 3D depth profile. However, with respect to the field of view and accuracy, current
commercial 3D structured light cameras did not prove to have enough resolution and are strongly affected
by variable ambient lighting conditions. In addition, the scanning frequency of sensors might not be suitable
to monitor moving cuttings. On the other hand, laser profile scanners are able to cover a large scanning
range and have a higher resolution (up to 1mm in both horizontal and vertical directions). Although laser
profile scanners only provide a 2D profile of the target surface, with accurate moving speed input, its high-
frequency depth data can be reconstructed by computer vision techniques to produce a 3D depth profile
and thereby measure volume. The Gocator 2380 (LMI 2016) smart profile scanner was considered to be
most suitable even though its hardware is not specifically designed for oil and gas industry applications. An
explosion proof enclosure, however, can be used to protect sensors in the hazardous outdoor environment.

Physical Design
The Gocator 2380 profile laser scanner allows measurement of the depth profile of the cuttings. Based on the
measuring range, the z-resolution ranges from 0.092 mm to 0.488mm, and x-resolution ranges from 0.375
mm to 1.100 mm. This laser sensor has a compact gasket aluminum enclosure with dimensions of 272 mm ×
49 mm × 75 mm (L × W × H) (LMI 2016). In order to produce a detailed 3D profile of the moving cuttings, it
is essential to determine the moving speed of the cuttings. An integrated 2D and 3D computer vision system
is proposed to extend existing technology to accurately measure cuttings/cavings volume, size, and shape
in real-time, as shown in Figure 8. By using image processing techniques, a 2D HD Camera was included
to measure cuttings moving speed and analyze size distribution of the cuttings. With the synchronization
of measured cuttings speed and 2D laser scanning frequency, a 3D profile of cuttings can be generated and
volume can be determined. The Point Grey Blackfly 2.3MP Color GigE PoE camera was selected to capture
high-resolution images at high speeds. Figures 8a and b illustrate the design concept. Figure 8c present the
UT testbed setup and Figure 8d shows the field setup with an explosion proof enclosure.
SPE/IADC-184718-MS 9

Figure 8—UT Austin Cuttings Monitoring Testbed Design

Software Design and System Logic Map


In order to design a real-time cuttings monitoring system, Microsoft Visual Studio and C++ were selected as
the programming environment and language respectively. Visual Studio allows controlling the Gocator 2380
laser scanner and Point Grey Blackfly HD camera at the same time. Concurrently, open source SDK libraries
of each sensor can be added into Visual Studio together with Open Source Computer Vision (OpenCV 2016)
and Microsoft Direct3D (Microsoft 2016) imaging processing libraries. A software architecture, shown in
Figure 9, illustrates the levels of control in the software system. It indicates the roles of each electrical
component and the flow of data. A detailed procedure is as follows:
1. Point Grey 2D HD camera and Gocator 2380 laser scanner are triggered via Visual Studio.
2. Both sensors are calibrated at certain fixed resolution and scanning frequency.
3. High-resolution 2D images are passed back to the communication interface via Gigabit Ethernet.
4. Cuttings moving speed is calculated by computer vision algorithms (active binarized image difference
analysis) such as: createBackgroundSubtractorMOG(), findContours(), and computeCentroid() are
applied to perform speed measurement (OpenCV 2016).
5. Cuttings size distribution is generated by computer vision algorithms (e.g. Canny() edge detection,
counterArea(), etc.) and data is stored for display (OpenCV 2016).
6. 2D profile depth data is passed back to the communication interface in the form of raw data points
via Gigabit Ethernet
7. Measured speed (ν) is then used with the 2D profile data for volume calculation and 3D profile
reconstruction.
8. 3D profile reconstruction is performed either in real-time or off-line mode.
9. 3D profile data is then finally stored.
10 SPE/IADC-184718-MS

Figure 9—Software architecture for cuttings analysis

Experimental Design
The objective of the validation experiment described here is to evaluate the proposed computer vision
technique and generate an experimental threshold for void space correction when calculating cuttings
volume. The experiment includes two parts: 1) uniform-shape objects testing and 2) real cuttings testing.
The purpose of testing uniform-shape objects is to validate the volume measurement against the detected
moving speed, since uniform shaped objects have either fixed void space ratio or no void space. The three
uniform-shapes targets are ping-pong balls, aluminum bars, and aluminum rods as shown in Figure 10. For
drill cuttings testing, the sample used in this experiment were all from the same formation interval collected
from the field. The shapes are random, as shown in Figure 10, and size distribution of cuttings is relatively
consistent. The cuttings have a composition of 70% shale and 30% sandstone. Experiments were conducted
by sliding various samples down the yellow ramp (Figure 8c) at different tilting angles.
SPE/IADC-184718-MS 11

Figure 10—a) Ping-pong ball (Sellercube 2016) radius = 40mm b) Aluminum bar (15mm × 30
mm × 70mm) c) Aluminum rod (OD= 25 mm, length = 100 mm) d) Drilling cuttings sample

Experimental Results and Discussion


The 2D laser profile scanner only generates depth data of the target's surface. Therefore, the vertical space
below the top surface is also included in volume calculation. With known dimensions of the three uniform-
shape objects (spheres, cubes and rods), the over-calculated volume ratio can be calculated as followings:

(1)

(2)

Then these ratios were applied to the uniform-shaped objects’ measured volume:

(3)

Table 2 shows the calculated volumetric percentage error of the experimental measurement on the three
uniform-shaped objects.

Table 2—Volumetric measurement error percentage for uniform-shaped objects

Ramp Angle in degrees

45° 50° 60° 70° 75° Average

Sphere 0.97 % 2.65% 4.30% 0.08% −1.09% 1.38%

Rectangular Bar 2.05% 0.62% 1.99% 3.21% 2.78% 2.13%

Rod 2.91% 3.74% 3.88% 0.72% 2.16% 2.68%


12 SPE/IADC-184718-MS

While testing uniform-shaped objects, it was found that the average measurement error for the volume
was less than 3%. The experiment was performed by varying the moving speed of the objects. This
demonstrated that the speed measurement algorithm and volume calculation are relatively reliable and
accurate. It was furthermore demonstrated that the proposed computer vision system is capable of tracking
an object's moving speed and measuring an object's volume in the laboratory environment.
Compared to uniform-shaped objects, drilling cuttings have irregular shape and size variance, which
make it impossible to calculate a theoretical void space threshold for a pile of moving cuttings. The void
space threshold is critical for accurate volume estimation. Also, different amounts of cuttings moved down
the yellow ramp at various speeds. Figures 11 and 12 plot over-calculated volume in percentage against
known sample volume and cuttings moving speed respectively.

Figure 11—Over-calculated volume against cuttings actual volume

Figure 12—Over-calculated volume against cuttings moving speed.


SPE/IADC-184718-MS 13

No obvious correlations have been spotted between cuttings moving speed, sample size, and the
calculated volume. The over-calculated volume, which can be identified as the void space below the cuttings
surface, is in the range of 30% to 45%. An average volume fraction is calculated from Figure 12 as
Ø=37.18%. This is an experimental threshold which can be used to correct the volume measurement of dry
cuttings in the laboratory environment.
3D depth data of cuttings was stored simultaneously. Figure 13 demonstrates a reconstructed 3D profile of
moving cuttings down the ramp. This 3D profile data can then be applied to detect larger cavings and analyze
the shape of cavings. The shapes of cavings can be categorized as tabular, angular, and splintered (Karimi,
Moji 2013). The shape of cavings indicates the mechanisms of wellbore collapse. The size of cavings can
be measured based on length, width, and thickness. Cavings provide an indication of possible wellbore
instability, formation overpressure, and overall well behavior. Cavings can be induced by underbalanced
drilling, stress relief, pre-existing planes of weakness, or result from mechanical action by drilling tools
(D. Kumar 2012). Real-time monitoring of cavings during drilling operations can help optimize drilling
performance by using appropriate actions to prevent NPT.

Figure 13—A generated 3D profile of moving cuttings

By adjusting noise level and threshold value, the majority of cuttings were differentiated from the
background environment as shown in Figure 15. With the help of Canny edge detection algorithms,
cuttings were presented as contours in the image. The Canny edge detection algorithm is a computer image
operator that detects a wide range of edges in images by applying multi-stage algorithm like Gaussian filter,
localization, intensity gradients, etc. (Parker 2010). Contours that have the same color or intensity can be
represented by a curve joining all the continuous points along the boundary (Maire 2009). These detected
contours can then be used to calculate the size of each cutting. This is a preliminary test for cuttings size
distribution measurement, in which high levels of processor loading were observed. The processing time
takes between 10 and 50 seconds depending on the level of details in the image. The success of the algorithm
also depends on the lighting environment and adjustment of Canny threshold value. This algorithm proposes
a feasible machine vision technique to quantify cuttings size distribution.
14 SPE/IADC-184718-MS

Figure 14—A summary of cavings types (Karimi, Moji 2013)

Figure 15—a) Original image (upper) and processed image (lower) by


Canny edge detector (left) and b) Cuttings size distribution analysis (right)

Benefiting from various advanced computer vision techniques, the proposed system has the ability to real-
time quantify cuttings volume accurately and generate a 3D profile. Comparing most of the current industry
methods for cuttings monitoring, a cuttings flux meter has a more promising accuracy in real-time cuttings
return quantity measurement. However, such measurement accuracy relies on known formation density and
wet cuttings correlation threshold adjustment. In the meanwhile, as shown in Table 3, the proposed 2D
and 3D integrated computer vision system not only provides a direct assessment on cuttings volume but
also stores raw 3D depth data. With those geometric data, larger cavings can possibly be identified and the
shape can be analyzed accordingly. This automated design can also remove rig personnel from this hazard
working zone.

Table 3—Comparison between current industry method and the proposed system

Traditional Cuttings 2D Video Monitoring UT Proposed 2D and


Cuttings’ Properties
Sieve System Flux Meter (Marana, et al. 2010) 3D Integrated System

Size Distribution ◐ ●

Return Volume ● ◐ ●

Weight ●

Caving Detection ●

Shape Profile Analysis ◐ ●

Real-time Monitoring ● ● ●

Safety ● ● ●

System Integration for


● ● ●
Drilling Automation
SPE/IADC-184718-MS 15

To extend the work one step further, it is essential to incorporate an advanced cuttings transport model for
estimating the theoretical cuttings return volume based on drilling parameters. Cayeux (2013) proposed a
transient cuttings transport model that integrates cuttings transport with fluid transport model and drill-string
mechanics. Such a system can then continuously compare the difference between the measured volume
and the theoretical return volume of cuttings in real-time. In addition, cuttings size distribution and cavings
also represent borehole quality and formation strength. With the assistance of a comprehensive cuttings
transport model, the traveling time of cuttings and cavings can be estimated. A field operator can then relate
the measured cuttings characteristics and cavings to the depth where this debris was produced and the time
the debris has resided in the well. As an important element of the real-time cuttings monitoring system,
this proposed computer vision prototype sets a foundation for integrating this system into a control scheme
which actively evaluates cuttings transport efficiency and monitors wellbore instability. Furthermore, it
provides real-time raw data and/or metrics that could be evaluated remotely by an operator, which provides
the opportunity to remove workers from hazardous working areas.

Conclusions
The current procedures for measuring cuttings either require human handling or lacks certain functionalities
such as 3D model reconstruction. With the development of 3D sensing technology during the last decade,
various commercial 3D scanners are now capable of volume measurement. After evaluating and testing
some advanced 3D scanners, the Gocator 2380 laser profile scanner was chosen in this study for its high
resolution (±1mm), fast scanning speed, compact size, and large FOV. The trade-off for high resolution
and fast scanning speed is that it only provides high frequency 2D profile data. The next challenge was
to measure how fast the cuttings were sliding down the ramp. By applying a series of digital image
processing techniques, the cuttings moving speed and size distribution were estimated. Subsequently the
synchronization of input speed and 2D depth data enabled the volume of the scanned target to be calculated.
A 3D surface profile of the cuttings was also generated. To verify the accuracy of proposed system, uniform-
shaped objects were tested and an accurate volumetric measurement was achieved. As cuttings slide down
the ramp, the top surface of the cuttings covers some underlying void space. By running the experiment at
different speeds, an empirical volume correction could be established. The error rate was identified and its
low variance in the experimental data is encouraging.
This paper proposes a state-of-art non-intrusive computer visualization system to quantify cuttings/
cavings volume, size distribution, and surface profile. An explosion proof enclosure has been designed and
the laboratory results presented here justify the pursuit of field trials, which will be conducted using the
design presented in this paper. Future work is required to test and improve the vision processing techniques
in the field, where cuttings are coated with drilling fluid and illumination conditions vary. For the next stage,
the following topics will be addressed as a part of the continuing research:

• Shape characterizations and analysis –With the 3D model of the reconstructed cuttings/cavings
surface, machine vision algorithms can be developed to characterize the shapes of cavings based on
length, width, thickness, and corner angle. Such algorithms will allow the system to automatically
detect cavings and identify wellbore instability. Both feature-based criteria and modern machine
learning algorithms (based in part on expert training) will be considered.
• Field Trials – The speed tracking vision algorithm (background subtraction method) depends on
adjusting certain parameters for varying lighting environments. Therefore, it is crucial to test this
algorithm on an actual drilling rig under different lighting conditions. Field trials are also necessary
to evaluate robustness to the dirty environments and weather, and to garner user feedback.
• Cuttings size and shape model – Different drill bits will generate different shapes of cuttings. It is
essential to develop a model based on various drill bit types to theoretically estimated cuttings void
space ratio. Additionally, 3D depth data will need to be collected in field trials in order to obtain
16 SPE/IADC-184718-MS

appropriate empirical void space correction ratios. Initial efforts will focus on identified generally
homogeneous field and drilling conditions in order to accelerate deployment opportunities.
• Cuttings transport model – In order to correlate measured cuttings volume with theoretical return
volume, a comprehensive cuttings transport model should be integrated with the vision system's
algorithm. Drilling information, such as fluid rheology, flow rate, well path, etc., should be used
in real-time for cuttings traveling velocity estimation. Only with the assistance of an accurate
cuttings transport model can the system relate detected cavings and cuttings size distribution to
the formation depth.

Acknowledgment
The authors would like to thank ConocoPhillips for financial support and technical guidance on this project,
as well as for providing field assistance during the test stage.

Glossary
CCD = Charged Coupled Device
CFM = Cuttings Flux Meter
EDCR = Equivalent Dry Cuttings Ratio
FOV = Field of View
KPI = Key Performance Indicator
LED = Light-Emitting Diode
NPT = Non Productive Time
OSHA = Occupational Safety and Health Administration
OpenCV = Open Source Computer Vision
PoE = Power over Ethernet
ToF = Time of Flight

Symbols
L = length
r = radius diameter
Ø = over-calculated volume ratio
ν = measured object moving speed

References
O.C.T.G. Procter Consultancy Ltd. 2000. OCTG NS-17 ABC of Hole Cleaning. Aberdeen, Scotland: O.C.T.G. Procter
Consultancy Limited.
C. Aldea, A. W. Iyoho, M. Zamora. 2005. "Hole Cleaning: the Achilles' Heel of Drilling Performance." Houston, Texas:
American Association of Drilling Engineers.
Cayeux, E., T. Mesagan, S. Tanripada, M. Zidan, and K. K. Fjelde. 2013. "Real-Time Evaluation of Hole Cleaning
Conditions Using a Transient Cuttings Transport Model." SPE/IADC Drilling Conference. Amsterdam, The
Netherlands: Society of Petroleum Engineers. doi:http://dx.doi.org/10.2118/163492-MS.
D. Kumar, S. Ansar2, S. Wang, J. YiMing, S. Ahmed, M. Povstyanova, and B. Tichelaar. 2012. "Real-time Wellbore
Stability Analysis: An Observation from Cavings at Shale Shakers." Singapore.
David Fofi, Tadeusz Sliwa, Yvon Voisin. 2004. "A Comparative Survey on Invisible Structured Light." http://
www.le2i.cnrs.fr/IMG/publications/fofi04a.pdf.
Ferranod, Pierrick, and Slim Hbaieb. 2015. "Schlumberger CLEAR Hole Cleaning and Wellbore Risk Reduction Service."
Accessed 09 03, 2016. http://www.slb.com/~/media/Files/drilling/industry_articles/2015-clear-ep.pdf.
Geng, Jason. 2011. "Structured-light 3D surface imaging: a tutorial." Advances in Optics and Photonics pp. 128–160.
Gerig, Guido. 2013. "Structured Lighting." http://www.sci.utah.edu/~gerig/CS6320-S2013/Materials/CS6320-CV-
S2013-StructuredLight.pdf.
SPE/IADC-184718-MS 17

Graves, W.V. Andrew, and Matthew Dennis Rowe. 2013. Down Hole Cuttings Analysis. US Patent WO 2013089683 A1.
June 20.
IPIECA. 2009. "Drilling fluids and health risk management." International Petroleum Industry Environmental
Conservation Association. http://www.ogp.org.uk/pubs/396.pdf.
Karimi, Moji. 2013. "Drill-Cuttings Analysis for Real-Time Problem Diagnosis and Drilling Performance Optimization."
SPE Asia Pacific Oil and Gas Conference and Exhibition, 22-24 October. Jakarta, Indonesia: Society of Petroleum
Engineers. doi:http://dx.doi.org/10.2118/165919-MS.
Li, Larry. 2014. "Time-of-Flight Camera – An Introduction." Accessed 07 03, 2016. http://www.ti.com/lit/wp/sloa190b/
sloa190b.pdf.
LMI, Technologies. 2016. "Gocator 2300 Series." http://downloads.lmi3d.com/system/files/Gocator/documents/Gocator
%202300%20Series/DATASHEET_Gocator_2300_US_WEB.pdf.
Maire, Michael Randolph. 2009. Contour Detection and Image Segmentation. Thesis, California Institute of Technology.
Marana, A.N., J.P. Papa, Marcus V.D. Ferreira, Kazuo Miura, and Francisco Assis Cavalcante Torres. 2010. "An
Intelligent System To Detect Drilling Problems Through Drilled-Cuttings-Return Analysis." IADC/SPE Drilling
Conference and Exhibition, 2-4 February. New Orleans, Louisiana, USA: Society of Petroleum Engineers. doi:http://
dx.doi.org/10.2118/128916-MS.
Michael Brading, Kenneth Salsman, Manjunath Somayaji, Brian Dipert, Tim Droz, Daniël Van Nieuwenhove,
Pedro Gelabert. 2016. "3-D Sensors Bring Depth Discernment to Embedded Vision Designs." Embedded Vision
Alliance. http://www.embedded-vision.com/platinum-members/embedded-vision-alliance/embedded-vision-training/
documents/pages/3d-sensors-depth-discernment.
Microsoft. 2016. Direct3D. Accessed 08 04, 2016. https://msdn.microsoft.com/en-us/library/windows/desktop/
hh309466%28v=vs.85%29.aspx?f=255&MSPPError=-2147217396.
Morton-Thompson, Diana, and Arnold M. Woods. 1993. "Mtds 10 - Development Geology Reference Manual." AAPG.
Naegel, M., E. Pradie, T. Delahaye, C. Mabile, and G. Roussiaux. 1998. "Cuttings Flow Meters Monitor Hole Cleaning
in Extended Reach Wells." European Petroleum Conference, 20-22 October. The Hague, Netherlands: Society of
Petroleum Engineers. doi:http://dx.doi.org/10.2118/50677-MS.
National Instruments. 2012. "Parts of a Stereo Vision System." Accessed 08 03, 2016. http://zone.ni.com/reference/en-
XX/help/372916M-01/nivisionconceptsdita/guid-cb42607e-f256-40f5-ab6e-28ec3a33bcda/.
OpenCV. 2016. "OPENCV-OPEN SOURCE COMPUTER VISION." http://opencv.org/.
OpenCV Tutorials. 2016. Background Subtraction Methods. Accessed 07 27, 2016. http://docs.opencv.org/3.0-beta/doc/
tutorials/video/background_subtraction/background_subtraction.html.
OSHA. 2016. https://www.osha.gov.
Parker, J. R. 2010. Algorithms for Image Processing and Computer Vision. 2. John Wiley & Sons.
Pietro Zanuttigh, Giulio Marin, Carlo Dal Mutto, Fabio Dominio, Ludovico Minto, and Guido Maria Cortelazzo. 2016.
Time-of-Flight and Structured Light Structured Light Depth Cameras: Technology and Applications. Switzerland:
Springer International Publishing.
Sellercube. 2016. Sellercube. http://img.sellercube.com/UploadFile/6/SKU032820/20120515150716945.JPG.
18 SPE/IADC-184718-MS

Appendix
Table A—Characteristics of five 3D depth sensors

Kinect 3D Acuity 820 –


SICK LMS400-2000 Gocator-2380
Sensor 1000

Measurement Type Structured Light Laser Laser Laser

Scanning Frequency Slow High High High

X an Z resolutions Low Medium High High

Meet FOV Requirement YES YES YES YES

Size Compact Compact Large Compact

Cost Low High High High

You might also like