CONSTRUCTION OF AN
AUTONOMOUS LABORATORY
SCALE-DRILLING RIG FOR
TESTING AND CONTROL OF
DRILLING SYSTEMS
Authors: Erik Andreas Løken (UiS) & Suranga
C.H. Geekiyanage (UiS)
Co-Authors: Dan Sui (UiS) & Robert Ewald (IRIS)
OUTLINE
• Introduction
• Rig structure (system design, components, sensors)
• Data acquisition & data quality improvement
• Control system design (WOB regulation, fault detection)
• Rig performance optimization
• Conclusion
• Future work
1.
INTRODUCTION
• Team
• Drillbotics™ 2017
competition
• Fully autonomous
laboratory-scale rig
2.
RIG STRUCTURESYSTEM DESIGN
3.
RIG STRUCTURE
• Power transmission
• Hoisting
• Circulation
• Drill string, BHA & bit
SYSTEM DESIGN
Rock
Actuator
hoisting
system
Multi-
filtration
system
Dual-Pump
setup
Leak
valve
Pressure
valve
Return line
Swivel
Top Drive
Drill pipe
BHA
PDC bit
4.
RIG STRUCTURE
• Hollow-shaft brushless top
drive (0 – 3500 RPM)
• 2.86 Nm rated torque (8.59
Nm instantaneous torque)
• Direct torque transfer to drill
string assembly
• Low vibration-signature
• Programmable auto-braking
POWER TRANSMISSION
5.
RIG STRUCTURE
• Three actuators w/ stepper
motors and brakes
• 500 N WOB capacity (hook
load = 168 N)
• Variable speed adjustment
for different drilling stages
HOISTING SYSTEM
6.
Actuator #1
Actuator #3Actuator #2
Z-axis
X-axis
Fy
Fz
Fx
Top Drive
RIG STRUCTURE
• Pmax = 4.1 bara
• Q = 11.5 l/min (min. velocity
range of 0.5 – 0.7 m/s for
cuttings transportation)
• Attached w/ swivel (1000 RPM
capacity)
• Filtration deposit system
CIRCULATION SYSTEM
7.
RIG STRUCTURE
• Interchangable string
(DP), BHA & bit
• Aluminum pipe
(WT = 0.889mm)
• Stainless steel BHA
w/ three stabilizers
• PDC bit with raked
cutters
DRILL STRING, BHA & BIT
8.
DATA ACQUISITION
• Three triaxial load cells (LC),
0-100N range (Fx, Fy, Fz)
• RPM and torque (TTD) in top
drive sensors
• Ferromagnetic torque sensor
(T)
• Pressure transmitter (P)
• Reference height sensor (H)
SENSOR INTEGRATION
Rock
LC LC
LC
P
T
RPM
&TTD
∆ h
H
9.
DATA ACQUISITION
• Torque Sensor sample rate
<< 17 ms at 1000 RPM
• Arduino DUE loop-time for
other sensors
• Data logging:
uploaded every
100ms
DATA LOGGING & SAMPLING
10.
DATA QUALITY IMPROVEMENT
• T, RPM, P, Fx/Fy, ∆h –
low-pass (LP) filtered
• ∑Fz (WOB) – median,
moving avg. & LP
filters
• Frequency-analysis
for cut-off frequency
FILTERING OF DATA
11.
CONTROL SYSTEMHIERARCHICAL TWO-LAYER STRUCTURE
STRATEGIC DECISION CONTROLLER + VISUALIZATION
HOISTING SYSTEM
CONTROLLER
3 TRIAXIAL LOADCELLS & HEIGHT
SENSOR
ROTATIONAL SYSTEM
CONTROLLER
2 TORQUE SENSORS + RPM FEEDBACK
CIRCULATION SYSTEM
CONTROLLER
PRESSURE SENSOR
Main computer
Arduino DUE (3x)
Sensor distribution
Arduino DUE
programmed as finite-
state machines (FSMs)
Non-
deterministic
control loopNormal Drilling
Fault Detection
12.
CONTROL SYSTEM
• Complete closed-loop PID-
controller for WOB
• Phases (1. calibration, 2.
pilot hole drilling, 3. normal
drilling, 4. optimization, 5.
completion)
• Incremental increase of
RPM and WOB setpoint
(sweep algorithm)
NORMAL DRILLING STATE
RPM
WOBmax
max
Input
13.
CONTROL SYSTEM
• Coordinator implemented
• Detects axial vibrations,
stick slip, lateral vibrations,
pack-off (OP), leak, key-
seating & overpull
• Problems simulated in
laboratory
• Example: stick-slip
management
FAULT DETECITON & INCIDENT MANAGEMENT
Stick slip detected (T increase, RPM
decrease)
Pick-up off-bottom (10mm) Increase
RPM by 10%, reduce WOB 10-
15%. Retag bottom
YesDrilling
commences
If still stick slip at RPM/WOB-thresholds, pick-up off-bottom. Set very low WOB & very high RPM.
Continue slow drilling before sweep can commence
Stick slip?
No
14.
CONTROL SYSTEMFAULT DETECTION & INCIDENT MANAGEMENT
• Stick slip & lateral
vibrations get detected
• System attempts to correct
the faults and reduce
vibrations
15.
CONTROL SYSTEMHUMAN MACHINE INTERFACE
16.
RIG PERFORMANCEROP CALCULATION
• ROP calculated (∆ℎ
∆𝑡)
• Instantaneous- and
average-ROP
calculated using
arrays of 15 s & 180 s
• Effect of DP deflection
will be implemented
(depth-based
measurement)
1st second < 1 2 3 … 14 15>
2nd second < 2 3 4 … 15 16 >
3rd second < 3 4 5 … 16 17>…
Etc. 𝑠𝑢𝑚 ∆ℎ𝑒𝑖𝑔ℎ𝑡𝑎𝑟𝑟𝑎𝑦
# 𝑜𝑓 𝑑𝑎𝑡𝑎𝑝𝑜𝑖𝑛𝑡𝑎𝑟𝑟𝑎𝑦∗ 60 = 𝑅𝑂𝑃 [
𝑚𝑚
𝑚𝑖𝑛]
Replace oldest value with newest every second
17.
RIG PERFORMANCERESULTS, NORMAL DRILLING
0 2 4 6 8 10 12 14400
600
800
RP
M
time(min)
0 2 4 6 8 10 12 140
5
10
RO
P(m
m/m
in)
time(min)
0 2 4 6 8 10 12 140
0.5
Torq
ue o
f bit(N
m)
Time(min)
0 2 4 6 8 10 12 140123
WO
B(k
g)
Time(min)
0 2 4 6 8 10 12 14 16 18 20200400600800
RP
M
time(min)
0 2 4 6 8 10 12 14 16 18 200
1
2
RO
P(m
m/m
in)
time(min)
0 2 4 6 8 10 12 14 16 18 200
0.5T
orq
ue o
f bit(N
m)
Time(min)
0 2 4 6 8 10 12 14 16 18 200
2
4
WO
B(k
g)
Time(min)
Weak formation (cement) Hard-drilling formation (floor tiles)
18.
RIG PERFORMANCENEW FORMATION DETECTION
0 2 4 6 8 10 12 140
100
200
300
MS
E(M
Pa)
Time(min)
0 2 4 6 8 10 12 140
20
40
60
UC
S(M
Pa)
Time(min)
0 2 4 6 8 10 12 14 16 18 200
500
1000
MS
E(M
Pa)
Time(min)
0 2 4 6 8 10 12 14 16 18 200
100
200
300
UC
S(M
Pa)
Time(min)
Weak formation (cement) Hard-drilling formation (floor tiles)
𝑀𝑆𝐸 =4 𝜏𝑏𝑖𝑡 𝑅𝑃𝑀
𝜋 𝑑𝑏𝑖𝑡2 ROP
+4 𝑊𝑂𝐵
𝜋 𝑑𝑏𝑖𝑡2
𝑈𝐶𝑆 = 0.35 ∗ 𝑀𝑆𝐸19.
CONCLUSION
• Understanding, evaluating
and investigating current
solutions available
• Experiments for developing
models, control system
algorithms etc.
• Prototyping of BHA /
sensors / components / bits
ADVANTAGES CHALLENGES
• Not necessarily scalable to
industry
• Some components (i.e.
actuators for hoisting) only
work well on lab-rig
• Limited sensors &
processing power
20.
FUTURE WORK
• Drill string dynamics model
development
• Machine-learning fault detection
• Kalman filtering for WOB and ROP
estimation
• Gauss-Seidel Constraint
Control for ROP-optimization
• New BHA design w/ real-time
downhole measurements
• Mud design & cuttings transportation21.
QUESTIONS?
22.