FUZZY DRILLING DIRECTION (FDD) CONTROLLER - 1
FUZZY LOGIC & DIRECTIONAL DRILLING
Hole Deviation (THD) was a building block that
resulted from searching for a solution to control logic for truly
automated rotary steerable directional drilling systems. Just like the
cruise-control in cars and the auto-pilot mode in aircraft, so too will true "auto-guidance"
really enter directional drilling operations.
page and Next present an overview of our
patented Fuzzy Logic controller for directional
steering. For reference we'll call it the Fuzzy
Drilling Direction (FDD) Controller. The discussions will
convey how Technical Hole Deviation plays a vital role in assessing the
current state of the directional drilling system and how Fuzzy-processing
such information leads to an informed change in "current" directional tool settings.
This document contains the following sections.
If Technical Hole Deviation
(THD) in general is new for you, we
kindly suggest you first read this.
STATEMENT AND BACKGROUND
1996, Stoner Engineering attacked the following problem: design a controller for truly automated directional drilling. At that
time, 3D rotary steerable directional drilling systems--the ones that provide direct
lateral force magnitude AND direction close to the bit--were very new. However,
given the many published benefits gained by imposing inclinational and azimuthal directional control while rotating, we
guessed these systems would eventually become standard for long and medium radii
a 3D rotary steerable directional drilling system, controller OUTPUT was the change in the
effective Tool Force Magnitude AND Orientation (TFMO)
acting at the adjustable stabilizer (or otherwise near the bit, depending on
system specifics); with respect to the adjacent sketch, this
is vector components DTFy
To keep the controller
practical, DTFMO must be determined from only the
current actual well path and the current preferred well path, that is, no models of
directional drilling were to be employed. The far
from simple challenge
was to create a controller--the logic implemented as software in a computer
chip--that could "do" what a "good" directional driller does.
direction in which a bit drills cannot accurately be predicted. Thus, a drilling
direction controller based on a rigorous (or otherwise) model of
drilling is in the practical sense, futile. That doesn't mean simulators don't
have value; it means that using classical control theory to design a controller
for directional drilling is severely limited in the best it can ever do. To
think that the mental data-processing capabilities of a directional driller can be
"temporarily" replaced with one or a few linear equations
far-fetched. Thus, in our opinion a "good" controller actually mimics--what
we think are--the
thought processes of a directional driller. One way to attempt this is with Fuzzy control theory. Fuzzy control theory provides a
means to design a controller for highly complex dynamic systems, with a
structure that can be entirely explained with words and intuitive phrases we can understand. Fuzzy control theory is also based on mathematical concepts that better model how
humans think and convey information.
DIRECTIONAL STEERING AUTOMATION?
There are two primary reasons
for directional steering automation:
Less 24/7 dependency on "human touch"
Better well bore for the operator
Pioneering technical analyses of bottomhole
assemblies began with the published works of
Arthur Lubinski in the 1950's. He showed that
imposing minor changes to controllable operating
parameters produces a well bore with minimum
dogleg severity variance. This equates to a
"smooth", usable well bore. If system controller
software (the "brains") is sufficient to
"temporarily" compete with the directional
control performance of the best directional
drillers, then the software
can reside in the downhole tool or otherwise be
made to automatically process frequently.
Why automation? A better well
bore, drilled on the preferred path...it's that simple.
COMMERCIAL STATE-OF-THE-ART STEERING AUTOMATION
portions of this page may be outdated.)
Before we continue with a
very brief introduction to Fuzzy Logic, we would like to summarize commercial
state-of-the-art automated rotary steerable directional drilling systems. Let us
begin by defining (truly) automated rotary steerable directional drilling:
A directional drilling system possessing a drilling mode
whereby tool settings (e.g., TFMO) are automatically determined by
system software, and frequently and automatically changed/controlled by
system hardware, while drilling.
Furthermore, the system controller software addresses
AND lineal deviation, and only requires
"knowledge" of the actual and planned well path trajectories. Thus,
the system controller software temporarily assumes the role of a
traditional directional driller.
At least three commercial 3D rotary
steerable directional drilling systems exist. They are:
Baker Hughes Inteq's AutoTrak
Baker Hughes Inteq's AutoTrak can only attempt to automatically control
angular deviation in
"hold mode", and otherwise requires manual "steer mode" to
attempt to control lineal deviation.
In other words, their system automatically
tries to be "on inclination" and "on azimuth"--be they
changing or constant--but not necessarily also "on
depth". AutoTrak for example is certainly a big step towards true automated
directional drilling. However, a comparable analogy is this: "the car is
headed straight, just like you asked; however, it's in the river instead of on
the road." We are not aware of the "automation"
features of PowerDrive or Geo-Pilot, but surmise they are similar to AutoTrak.
If drilling the payzone at
the best trajectory is important, then controlling lineal deviation is of utmost
concern. Controlling angular deviation alone is far easier than
controlling lineal deviation, because controlling lineal deviation inherently
requires the ability to also control angular deviation.
PRIMER ON FUZZY LOGIC
"Fuzzy" is a
catch-all term that refers to a system or methodology
that to some degree employs Fuzzy sets. A Fuzzy set is a general mathematical concept that
describes how an element belongs to a particular notion (set) of some domain of
definition. Classical, or Boolean set theory declares that membership of an
element to a set is either completely true or completely false (black/white,
on/off, 1/0, binary). Alternatively, Fuzzy set theory declares that the degree of
membership (DOM) of an element within a set lies within a continuum from true
to false [1.0, 0.0]. Additionally, and unlike Boolean set theory, an element can
"belong" to more than one set...even its
What does this mean? Consider
drilling a horizontal well bore. A hypothetical Fuzzy rule that relates vertical
deviation and the change in the vertical tool force component is the following:
Let us examine the
"IF" part of the rule. msVD represents an Input variable defined over
a domain (e.g., -40 feet to +40 feet). msVD could be described with 5 sets,
named VERY LOW, LOW, RIGHT-ON, HIGH, and VERY HIGH. The adjacent figure presents
an example "fuzzification" of msVD, in which the preceding 5 fuzzy
sets are defined. If they were Boolean sets, discrete true/false boundaries
would exist that separate the various notions about vertical deviation.
Let us consider msVD = 16 ft, for example.
The graph shows that +16 ft belongs to the set of HIGH vertical deviation to a
degree of 0.5, while it also belongs to the set of VERY HIGH vertical deviation
to a degree of 0.5 (and to a degree of 0 for VERY
LOW, LOW, AND RIGHT-ON). With Boolean sets, DOM is binary; move an infinitesimal
amount from the boundary (e.g. to 16.00001 ft) and set membership flips. Clearly,
Boolean sets do not model how humans--including directional drillers
(ha)--categorize elements and process information. A black and white world is
rare; typically it's many shades of gray.
How do Fuzzy rules work? In
short, the consequent ("THEN" part) is true to the same degree as the
antecedent ("IF" part) is true, and all rules are "fired" (computed). The ending result is a modified/scaled fuzzification of the
respective OUTPUT variable, which is then defuzzified to arrive at a discrete
result (e.g., DTFy
= -100 LB).
APPLICATIONS THAT USE FUZZY
multiple 10,000's of
documents have been published about Fuzzy Logic theory and applications. It is not our intention to even
begin to teach Fuzzy Control Theory to you. A query with search string
"Fuzzy Sets" at Amazon.com returns over 400 books! What should be
retained from this primer on Fuzzy Logic is that it's not just theory. Several
industries have successfully applied Fuzzy technology to solve real problems for monetary
benefit. See the
(dated) table below that lists a few commercial applications that employ Fuzzy
Logic. The sources are:
- 1) Kosko, Bart. 1993. Fuzzy Thinking: The New Science of Fuzzy
Logic. New York, New York: Hyperion.
- 2) McNeill, Daniel, and Paul Freiberger. 1994.
Fuzzy Logic. The Revolutionary Computer Technology That Is Changing Our
World. New York, New York: Simon & Schuster Inc.
||Hitachi, Matsushita, Mitsubishi, Sharp
||Honda, Mitsubishi, Nissan, Saturn, Subaru
|cement kiln control
||Isuzu, Nissan, Mitsubishi
||Fujitec, Mitsubishi Electric, Toshiba
|golf diagnostic system
|health management system
|iron mill control
||Hitachi, Matsushita, Sanyo, Sharp, Toshiba
|space shuttle docking
|subway control system
||Goldstar, Hitachi, Samsung, Sony
|traffic control system
||Hitachi, Matsushita, Toshiba
||Canon, Matsushita, Sanyo
||Goldstar, Hitachi, Matsushita, Samsung, Sanyo, Sharp
Fuzzy set theory was invented
Dr. Lotfi Zadeh in 1965. The first commercial applications of Fuzzy Logic
addressed control problems (e.g., controller for a cement kiln; controller for a
high-speed train). A few Fuzzy systems exist within the Petroleum Industry, and
most of those are fuzzy expert systems (e.g., fluid selection for stimulation).
As of 1996, we
were not aware of any commercial Fuzzy applications in the
which is why we applied Fuzzy technology to one of the best control problems we
have: directional drilling.
WELL PATH TRAJECTORIES WITH THE FDD CONTROLLER
Before we continue with an
explanation of the FDD controller, let us "jump to the bottom line"
to recognize why we believe our work has significant value. We created a
directional drilling simulator with which to test and design the FDD controller. The
simulator was a 3D finite element model incorporated with a drill-ahead model.
The finite element model was a static analysis of a typical rotary-steerable
bottom hole assembly. Our drill-ahead model was based on laboratory data and a
simulation model, proposed by Millheim and Warren in 1978, and Brett et al. in
1986, respectively. We setup many formation types by altering the respective
parameters and stochastically varied those parameters to implement extra
complexities with which the controller would have to deal. For example, directional
control is different when drilling at 200 feet per hour versus 10 feet per hour.
Progress was slow. In the
beginning, we experienced all the common problems associated with complex
controllers: (well path) instability and controller-parameter sensitivity. Of
course, Technical Hole Deviation and the FDD controller were being invented simultaneously. However, "in the end",
we produced a controller that possessed very important characteristics.
Consider the following two
vertical section views, which were created with the simulator and the FDD controller. The first one presents six TVD corrections for a horizontal well,
where initial vertical deviations varied from 3 feet to 8 feet in one foot
increments, and initial well bore inclinations were 90 degrees. The second
vertical section view below presents 3 entire horizontal wells, modeled from the
KOP through the horizontal section. The
significance of these plots is given below!
Section View for six TVD
Section View for three
smooth well bores were "drilled" with the identical Fuzzy controller! By identical,
we mean the controller parameters were kept constant in all cases, while initial
conditions, well plans, and formation parameters were significantly varied.
this mean? It means GENERALITY--a systems engineer's dream.
This technology has
performed for so many other
Every controller has control
parameters that directly affect the computed OUTPUT, and those parameters must
be tuned. Classical controllers (e.g., P, PI, PID) typically have a very small
number of parameters, therefore, "choosing" or tuning to find the
"right ones" usually does not result in a general controller. In other
words, take a tuned classical controller and simply change the initial
conditions; directional control performance becomes heavily degraded.
FDD controller has more than one hundred control parameters, but common-sense
gets most of them "close-enough" and the remaining few are tuned.
That's the power of Fuzzy. Academia cares about theoretical correctness, while
industry cares about solving problems efficiently and effectively, period; for
this reason, Fuzzy Logic still (after 40+ years) is
sometimes met with criticism in Academia.
To view more details of the FDD controller,