Drilling Automation

1
Drilling Automation

• Dr. Brian Glass
• Howard Cannon
• Samantha Domville

2
Drilling is unpredictable

• Discovery -- hard to predict whats down there
• Seismic surveys used on Earth
• Rugged environments
• Matching drill behavior to lithology (Sand?
  Basalt? Tuff? Worn-out bit?)
• Diagnostics and prognostics
• Sticking (collapses, creep, ice layers)
• Retrieval of cores and cuttings up a long, thin
  hole

Crater Ejecta Volcanic Flows Weathering
Products Sedimentary Deposits
Basement Fractured in situ
Self-Compaction Depth
Steve Clifford 1993
3
Drilling Diagnosis Difficult

• Highly uncertain environment
• limited number of measurements
• performance dependant on local strata
• environment changes over time
• Need to react quickly
• Situations require different techniques with
  associated costs and risks important to
  identify the correct state.
• Example Auger choking versus hard material.

4
Current Activities

• SMD Projects
• MIDP (PI Brian Glass) 3-year development
  funding for Drilling Automation for Mars
  Exploration (DAME), for software development and
  field testing of a Honeybee drill at the Haughton
  Crater Lunar/Mars-analogue site in the High
  Arctic
• ASTEP (PI Carol Stoker) 3-year MARTE field
  campaign to drill in the unique, highly acidic
  Rio Tinto bioanalogue site in Spain, with a
  Honeybee drill
• -- subtask provides human-centered computing,
  remote science operations simulation, drill
  automation, field wireless communications
• MTP (PI Brian Glass) Adapt DAME-like automation
  and diagnostic software for a new planetary deep
  drill being designed for JPL by Swales, beginning
  in FY05
• ASTID (PI Geoff Briggs) 3-year development
  funding for a Baker-Hughes/JSC Mars Arctic Deep
  Drill (MADD), targeted towards bit development
  for cold-region Mars-analog field site tests in
  the Arctic
• -- subtask provides limited heuristic diagnosis
  and data acquisition, controls and displays
• MIDP (PI Carol Stoker) 3-year development
  funding for a shallow/near-subsurface,
  percussive-drilling Mars Underground Mole (MUM),
  based on the ESA Beagle-2 flight hardware design
  (Note Code TI (H. Vu) is also the lead for the
  MUM mechanical redesign work)
• -- subtask provides diagnostics, mole controls
  and firmware design
• ESMD Projects
• ECP (PI Jerome Johnson/Army CRREL) Regolith
  Characterization Project to build a suite of both
  surface and subsurface (using a modified Honeybee
  drill) instruments for prospecting lunar ISRU and
  surveying the subsurface. Targeted to provide a
  low-mass suite for use in 2009 or later lunar
  surface missions.
• -- subtask provides software architecture and
  integration of instruments, executive controller
• ARC/TI is the drilling automation technology
  provider of choice for all three of the competing
  US planetary subsurface vendors (Honeybee,
  Baker-Hughes/JSC, and Swales) and also
  collaborates with the Canadians

5
Drill Automation for Mars Exploration (DAME)

• Description
• Purpose develop enabling information
  technologies for planetary subsurface exploration
• Approach iterative spiral implementation and
  field testing of drilling automation
• Deliverables
• Low-power, dry Mars-prototype drill, bits
  modified for Arctic conditions (permafrost)
• Control and automation software
• Dynamic shaft flexural model and dynamic failure
  models
• Hybrid diagnostic module

FY04-FY06 Critical Milestones FY04 Design and
fabricate drill field test to characterize
performance, faults FY05 Build diagnostic models
of drill initial diagnostic software field test
in parallel FY06 In-the-loop drill automation
tests at analog site
Co-Investigators / Participants ARC/Code TI,
ARC/Code SS, Honeybee Robotics, Georgia Institute
of Technology, SETI Institute, QSS, Mars
Institute, RIACS . PI Dr. Brian Glass, ARC,
[phone], brian.glass_at_nasa.gov
6
2004 DAME Field Season
Full-scale Mars-prototype deep drill has been
tested for 1st time under field conditions at a
high-fidelity Mars-analog site (Haughton Crater).
Drilled 2.2m in permafrost and regolith-like
breccia, July 2004 Two sites at Haughton Crater
on Devon Island in the the Canadian Arctic,
operating at Mars-relevant power levels
(max150-200W). Level 1 milestone completed
successfully
7
Cognitive Models Comparison to FDIR Architecture

• Some cognitive models see humans comprised as a
  team with multiple intelligences --
  analytical/classificational, emotional, and
  reactive selfs
• Balances of these -- theories of autism, etc
• Relevance to planetary surface robotics in
  replicating/supplementing human functions in
  exploration
• In our current FDIR automation approach, there
  are several internal agents with defined roles...
  
• Quick reflex ("ow, my hand's on fire!")
  implemented as very fast rules or sensor
  limit-checks
• Fuzzy neural net that takes past training
  examples of known problems or faults and then
  perceives incoming data, coming up quickly with
  the closest match it knows about from its neural
  weightings
• Model-based reasoning module which carries around
  an internal running simulation of how things are
  supposed to behave (from the physics, or
  underlying principles). The latter is slowest and
  gets invoked when there's a novel fault or
  situation that the other agents can't handle (or
  by running in parallel, subtleties that they
  didn't detect)
• Overall executive takes the hypotheses from the
  neural or model-based agents, and weighs between
  them and decides what course of action to invoke
  and how it can be made to be compatible with
  higher-level goals.

8
Vibration Classification Module

• Stick-slip-whirl-bounce model describes primary
  vibration characteristics of system
• Failure modes change model parameters
• Fuzzy-neural learning techniques used to learn
  model parameters, and help classify system state

9
Initial Drill Model
Control System
WOB, RPM Inputs from Exec
isensed
dz/dt desired
wq desired
Fz sensed
Power Supply
power
Motorq
Motorz
dz/dt sensed
wz
tz
wq sensed
Theta axis encoder
Z drive
Z axis encoder
tq
dz/dt
wq
Fz
Auger Force Sensor
dVcuttings/dt ejected
Auger
wbit
tbit
dzbit/dt
Fbit
WOB Sensor
dVcuttings/dt created
TOB Sensor
Bit
dzfloor/dt
10
Auger Model

• Nominal
• Drag due to carrying material is minimal
• Torque is a function of depth only
• TTlkdepth
• Choked
• Volume of material in auger can be calculated by
  integrating material displaced and limiting to
  auger capacity.
• Torque is a function of depth and volume
• Equation derived from Payzone model
• TTlkdepthK1/K2exp(k2vol-1)

11
Model Development

• Auger choking selected for initial tests due to
  high degree of uncertainty, and necessity to
  predict (can lead to severe reduction in
  performance potential freeze-up).
• Initial models show reasonable correlation to
  events in 2004 field season.
• Further lab tests conducted recently to refine
  improve models.

12
Leveraging a Common Software Architecture
MARTE 2005 Configuration
MInI Dispatcher
Commands
Contingent Executive
Drill Server
Drill
Ops Data
CSHS Server
CSHS
BHIS
BHIS Server
CRL Plan File
Execution Repository
Remote Sensor Servers
RSI
CRL Plan File
Science Data
Telemetry Interface
Borehole Science Repository
Mission Ops
13
Leveraging a Common Software Architecture
DAME Configuration
MInI Dispatcher
Commands
Contingent Executive
Drill Server
Drill
Ops Data
Hybrid Diagnosis Module
CRL Plan File
Vibration Classification Module
Execution Repository
CRL Plan File
14
MARTE Instrument Interface (MINI)

• Flexible Middleware that relays commands and data
  between multiple clients and servers.
• Built on top of CORBA so is relatively platform
  and OS independent.
• Allows communication without need to share source
  code. Clients can query servers for commands and
  parameters at run time.
• Utility provided which automatically constructs
  servers from a configuration file. Working
  server can be created in minutes.
• GUI provided which allows communication to any
  MINI server.
• MInI is complete and stable. Continuous uptimes
  of over 96 hours demonstrated, with as many as 3
  clients and 10 servers.

15
Contingent Executive

• Developed at NASA Ames for rover applications.
• Same Executive as that used in 2004 K9 rover
  field demonstrations
• Takes in a plan via Contingent Rover Language
  (CRL) issues commands and retrieves data from
  other subsystems via MInI
• Plans support contingent concurrent operation
• Specifies sequential or concurrent tasks.
• Allows specification of temporal relationships
  between tasks.
• Can specify start and maintenance conditions
  based on resources and state.
• Can specify plan branches based on resources and
  state.

16
MADD Automation

• JSC/Baker-Hughes advanced inchworm drill
• Eureka Arctic test site
• Original task was only data archival, later
• upscoped to include controls, displays and
• automation
• Data Acquisition Display
• Load cell, ROP, extension, commanded ROP, auger
  motor speed, auger motor current, auger motor
  Temp, AFOB Motor Temp, AFOB motor current, anchor
  motor current
• Alarm / Warning panel with text
• Motor Controls (includes Drill, AFOB and Anchor)
• Speed, Direction, and Enable/Disable
• Field-test window stability (user control
  safeguards)
• Archival of data and field notes
• Control loops and diagnostic algorithms

17
MADD DAQ Control and Automation
18
MADD Auger Choking Algorithm

• Drill model compared to HB drill model, similar
  fault modes, Labview controls

19
Summary

• Drilling is a MEPAG objective -- follow the
  water
• Drilling is difficult on Earth, an art form
• Automation of drilling is necessary for planetary
  subsurface exploration
• Progress made in drill fault detection and
  recovery, remote operations, controls and in
• Umbrella software architectures that integrate
  diagnosis, execution, sensors and instruments
  on-board
• Successful field tests at lunar and Mars-analog
  sites in 2004

20
Backups
21
Universal Executive (PLEXIL)
CASPER Planner (on-board)
PICO Contingent Planner (off-board)
Universal Executive
CLARAty Functional Layer

• Develop a common execution language that supports
  multiple planning systems (PLEXIL).
• Develop a generalized, robust Universal Executive
  within the CLARAty Decision Layer framework for
  PLEXIL based plans.
• Demonstrate the Universal Executive on the K9
  rover using plans from both CASPER and the PICO
  Contingent Planner.

22
Plans / FY04 FY06 DAME milestones

• The first year, the project designed and
  fabricated a Mars-prototype drill (6/04) and
  conducted manual drilling operations in the field
  to characterize fault modes (7/04)
• Second year, a set of initial automation will be
  tested alongside the drill in an observe-only
  fashion (8/05)
• Final season, in-the-loop automation tests in the
  field (7/06)

23
Automated Drilling Analogue Field Tests

• Overview
• Drill hardware and automation software field
  tests
• Automation and drill tests on Devon Island
  (crater, permafrost)
• Drill and operations, system of systems tests in
  Spain (subsurface life in rocks)
• Component tests (Eureka)
• Reference Mission Class
• Human and robotic subsurface exploration ISRU
  acquisition
• Analogue Features
• Subsurface ice and permafrost (Devon, Eureka,
  Antarctica)
• Regolith-like breccia layers (Devon)
• Absence of vegetation, roots, etc (Devon Eureka,
  Antarctica)
• Cold, dusty (gloved human operations) (Devon,
  Eureka, Antarctica)
• Unknown subsurface layers/strata (Devon, Spain)
• Planetary power and mass limits (max 150-200W
  power)
• Dry drilling without lubricants
• Major Accomplishments
• Full-scale planetary-prototype deep drill has
  been tested for 1st time under field conditions
  at a high-fidelity analogue site (DAME 7/04)
• Drilled frozen breccia, tested permafrost bit
  designs
• Tested automated drill controls

Drill hardware and automation test on Devon Island
Drilling and remote science ops at Rio Tinto,
Spain
24
Measurement Set-up
Accelerometer 2
Laser Vibrometer
Accelerometer 1
Load Cell
Shaker
Bushing
July 28, 2004
25
Channel 1 Load cell Channel 2 Accelerometer
near shaker Channel 3 Accelerometer near clamped
end Channel 4 Laser vibrometer
Clamped free configuration
File name Shaker03.mat
Frequency peaks Hz
Frequency peaks Hz
28.7500 162.5000 378.7500 567.5000
903.7500
28.7500 161.2500 378.7500 567.5000
805.0000
Frequency peaks Hz
28.7500 162.5000 378.7500 805.0000
892.5000
July 28, 2004