Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Development of a position sensing system for a robotic guided drill Wilkinson, Nicholas Jay 2006

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Notice for Google Chrome users:
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.

Item Metadata


831-ubc_2006-0342.pdf [ 16.64MB ]
JSON: 831-1.0081152.json
JSON-LD: 831-1.0081152-ld.json
RDF/XML (Pretty): 831-1.0081152-rdf.xml
RDF/JSON: 831-1.0081152-rdf.json
Turtle: 831-1.0081152-turtle.txt
N-Triples: 831-1.0081152-rdf-ntriples.txt
Original Record: 831-1.0081152-source.json
Full Text

Full Text

D E V E L O P M E N T O F A P O S I T I O N S E N S I N G S Y S T E M F O R A R O B O T I C G U I D E D D R I L L by N I C H O L A S J A Y W I L K I N S O N B . A . S c , The Un ive r s i ty of Br i t i sh C o l u m b i a , 2002 A T H E S I S S U B M I T T E D I N P A R T I A L F U L F I L M E N T O F T H E R E Q U I R E M E N T S F O R T H E D E G R E E O F M A S T E R O F A P P L I E D S C I E N C E i n T H E F A C U L T Y O F G R A D U A T E S T U D I E S ( M i n i n g Engineering) T H E U N I V E R S I T Y O F B R I T I S H C O L U M B I A A p r i l 2006 © Nicholas Jay Wilkinson, 2006 ABSTRACT This work details an investigation into possible methods for tracking the underground movement of a robotic, guided drill intended for minimizing borehole deviation in long-hole drilling applications. The work begins with a general overview of the operational constraints, followed by an in-depth exploration of several potential position sensing solutions. After a discussion concerning the benefits and disadvantages of the most interesting concepts, a MEMS gyroscope-based method for detecting the drill end effector tilt rate is chosen as the most technically feasible solution. A test apparatus is conceived and constructed to provide an initial, proof-of-concept evaluation of this approach. Experiments are conducted to determine the general performance characteristics of the gyroscope, and its suitability for the intended application. While the test results cannot be extrapolated to form a framework for establishing expected real-world drill performance, the low-cost gyroscope-based method is demonstrated to exhibit significant promise, and is proven worthy of continued attention and development. Table of Contents Abstract i i Table of Contents i i i Lis t of Tables v Lis t of Figures v i Acknowledgements v i i i Dedica t ion ix 1.0 Introduction 1 1.1 L o n g - H o l e D r i l l i n g and Dev ia t i on 2 1.2 L o n g - H o l e D r i l l i n g Techniques 3 1.3 The Robotic , G u i d e d D r i l l 5 2.0 Target Performance and Des ign Constraints 8 2.1 Target Performance 8 2.2 D r i l l Design-Based Constraints 11 2.2.1 V i b r a t i o n and Shock 12 2.2.2 V o l u m e and Shape 13 2.2.3 Mass 14 2.2.4 Cost 14 2.2.5 U n k n o w n Parameters Conce rn ing D r i l l Des ign 15 2.3 Opera t ional Envi ronment -Based Constraints 15 2.3.1 Cut t ings D i sposa l 16 2.3.2 Temperature 16 2.3.3 Magne t ic Fie lds 17 2.3.4 Scalabil i ty : 17 2.4 Constraint S u m m a r y 18 3.0 Conceptua l Solut ions 19 3.1 Trad i t iona l Gu idance Sensors 19 3.1.1 Cl inometers (Tilt Sensors) 20 3.1.2 Magnetometers 21 i i i Table of Contents, cont'd 3.1.3 Accelerometers 23 3.1.4 Gyroscopes 26 3.2 Non-Trad i t i ona l Approaches to Gu idance 31 3.2.1 External Camera 31 3.2.2 External Rollers 32 3.2.3 B iomimet i c Solut ions 33 3.2.4 F l u i d L e v e l 35 3.2.5 Laser Line-of-Sight 36 3.2.6 Non-Contac t Fiber Opt ic Probe 38 3.2.7 Capaci t ive Sensing : 39 3.2.8 Op t i ca l Shape Sensing 40 3.3 Concept S u m m a r y 41 4.0 M E M S Gyro-Based Posi t ion Sensing System 46 4.1 Gyro -Based PSS M e t h o d of Opera t ion 48 4.2 Test Appara tus Concept ion and Development 50 5.0 Test ing the Gyro -Based Pos i t ion Sensing System 60 5.1 Test Appara tus Specifications and Performance 60 5.2 Init ial CRS03 Test Results 62 5.3 Dri f t Test 68 5.4 Interpretation of Results 70 6.0 Conc lus ions 72 6.1 Future W o r k 74 Bib l iography 77 A p p e n d i x A - Test Appara tus Schematics 81 A p p e n d i x B - T r i a l Da t a 83 A p p e n d i x C - R a w Data Trials 90 List of Tables Table 1 - D r i l l Specifications S u m m a r y 15 Table 2 - Concept Summaries 45 Table 3 - Percentage Errors for T r i a l 1 65 Table 4 - Offsets f rom A c t u a l A n g u l a r Displacements for T r i a l 1 66 Table 5 - T r i a l 2 S u m m a r y 67 Table 6 - T r i a l 3 S u m m a r y 67 V List of Figures Figure 1 - Borehole Paths 2 Figure 2 - U n d e r g r o u n d D r i l l i n g & Blast ing Opera t ion 3 Figure 3 - The Robotic , G u i d e d D r i l l 5 Figure 4 - Robot ic D r i l l i n g Sequence 6 Figure 5 - Dev i a t i on Geomet ry 9 Figure 6 - M a x i m u m Ti l t 10 Figure 7 - The Robotic , G u i d e d D r i l l , Redux 12 Figure 8 - A v a i l a b l e V o l u m e 13 Figure 9 - Sensing Static Accelera t ion -. 25 Figure 10 - External Rollers 32 Figure 11 - Potentiometer N e t w o r k 34 Figure 12 - Laser Line-of-Sight 36 Figure 13 - Non-contact Fiber Op t i c Probe 39 Figure 14 - G y r o M o u n t i n g 47 Figure 15 - Robotic D r i l l i n g Sequence, Redux 49 Figure 16 - Test Appa ra tu s Setup 51 Figure 17 - L inea r Ac tua tor A s s e m b l y 52 Figure 18 - U n i v e r s a l Joint 53 Figure 19 - V i b r a t i o n H o u s i n g 54 Figure 20 - S imu la t i ng 3 D O F Behaviour 54 Figure 21 - V i b r a t i o n A s s e m b l y 55 Figure 22 - V i b r a t i o n C o n t r o l and Data A c q u i s i t i o n H a r d w a r e 56 Figure 23 - CRS03 gyro 58 Figure 24 - Sample G y r o Ou tpu t 63 Figure 25 - Da ta A n a l y s i s M e t h o d o l o g y : 65 Figure 26 - CRS03 Drif t Test . . . 69 Figure 27 - Second CRS03 Drif t Test 70 Figure 28 - Test Appa ra tu s Schematics 81 v i List of Figures, cont'd Figure 29 - G y r o M o u n t Deta i l 82 Figure 30 - V i b r a t i o n A s s e m b l y Deta i l 82 Figure 31 - Data, R a = 0.45deg/sec, d = 1.25cm, S = 1 k H z 83 Figure 32 - Data, R a = 0.45deg/sec, d = 1.25cm, S = 2 k H z 84 Figure 33 - Data, R a = 0.45 deg/sec, d = 1.25cm, S = 2 . 8 k H z . 85 Figure 34 - Data, R a = 0.45 deg/sec, d = 1.25cm, S = 5 .6kHz 86 Figure 35 - Data, R a = 0.45 deg/sec , d = 1.25cm, S = 11 .2kHz 87 Figure 36 - Data, Ra = 0.45 deg/sec , d=0.6cm, S = 2 k H z 88 Figure 37 - Data, R a = 0.23 deg/sec , d = 1.25cm, S = 2 k H z 89 Figure 38 - Unaveraged G y r o Ou tpu t 90 Acknowledgements The author would like to thank Dr. W. Scott Dunbar for offering the opportunity to participate in graduate studies in the Department of Mining Engineering. The academic freedoms he allowed and encouraged fostered myriad newfound interests, and enabled existing ones to flourish. In addition, his steadfast support throughout the entire course of this research, including all of its iterations, has proven invaluable. The author would also like to thank the Department of Mining Engineering. Although contact with faculty and staff may have been infrequent at times, their constant helpfulness and enthusiasm helped smooth away many potential wrinkles of frustration. - The thesis advisory committee is also to be offered profound thanks. Their insightful and constructive feedback was always at the ready and helped to keep this work from becoming too unwieldy. Finally, the author would like to offer thanks to Dr. Greg Baiden and Dr. Yassiah Bissiri for making this wonderfully multifaceted and challenging project available, and for providing financial support. vi i i Of course, to Mom. 1.0 Introduction Imagine y o u ' r e d r i v i n g a car a n d a l l of a sudden y o u real ize that y o u w a n t something on the back seat. W i t h o u t m u c h - if any - conscience thought, you ' r e able to reach w i t h y o u r a r m t o w a r d the seat beh ind y o u , maneuver y o u r h a n d un t i l you 've located the des i red object, and f inal ly, retrieve the object. O f course y o u must also do this w i t h o u t l o o k i n g , as you ' re very m u c h preoccupied w i t h the traffic ahead of y o u . H o w is this possible? H o w are y o u able to coordinate complex movement w i thou t h a v i n g to concentrate on those actions? W i t h o u t the abil i ty to look, h o w do y o u k n o w where your a rm is? P r o p r i o c e p t i o n - the i n t e rna l sensory i n p u t f r o m musc le and jo in t s t retch receptors - is i n large part wha t facilitates both the s imple and h i g h l y intricate actions that people take for granted o n a da i ly basis [1]. W i t h o u t this inherent ' pos i t ion sense', even w a l k i n g w o u l d become extremely diff icul t at best, a n d impossible w i thou t the sense of sight to help guide each footstep. In the mid-1990's, a n e w type of robot for long-hole d r i l l i n g was deve loped by a consor t ium of m i n i n g companies l ed by Inco L i m i t e d i n Sudbury , Canada . In a situation somewhat analogous to the d r i v i n g example ci ted above, this robot was tasked w i t h h a v i n g to maneuve r its a r m to a specific u n d e r g r o u n d p o s i t i o n wi thou t be ing able to v l o o k ' b y any t rad i t iona l means. Unfor tuna te ly , robots generally lack the abi l i ty to determine where they or their l imbs are wi thou t that specific functionali ty be ing bui l t into them, and the robotic d r i l l was no different. A posi t ion t rack ing system and m a n y other technical challenges were w o r k e d o n un t i l 2001 w h e n the project was shelved due to f inancial constraints. In that t ime, a feasible so lu t ion for de te rmin ing the robot 's l i m b pos i t ion was not d iscovered. In 2004 the robotic d r i l l project was transferred to Lauren t ian U n i v e r s i t y where w o r k cou ld beg in anew. l The goal of the present work is threefold: to propose a means by which the robotic limb can be tracked, to develop a method for testing this position sensing solution, and to determine the general feasibility of the proposed solution based on initial testing. Ultimately, the overall intent of this project is to lay the groundwork for endowing the robotic dr i l l wi th what could be termed machine proprioception. 1.1 Long-Hole Drilling and Deviation Long-hole dr i l l ing has many applications in a wide variety of industrial and scientific research settings. In mining, long-hole dri l l ing is often used for the creation of blast holes for excavating and fragmenting rock [2], [3]. These holes can be anywhere from several tens of meters to several hundreds of meters deep, and the general intent is to produce holes that follow predictable, straight-line paths [3]. The reality is often quite different however -borehole profi les tend to deviate i n an approximately exponential manner away from their intended path, generally reaching maximum deviations of approximately 10% of the hole depth (figure 1) [2], [3]. These deviations are in large part induced by the unknown underground structure of the rock being drilled through 1. Once a hole has been drilled, its exact profile and end point cannot be predicted and are not known until the dri l l can be removed and a specialized sensor package can be run down the hole to trace the profile of the path. Exacerbating the issue is Figure 1 - Borehole Paths. The dot ted l ine denotes the intended path, and the solid line, the actual deviation path. 1 Borehole deviations are also caused i n part by equipment setup error, however geological structure accounts for the most significant component of borehole deviation [2], [4]. 2 that unlike the two-dimensional representation of the problem in figure 1, real-world deviations occur in three-dimensions. Drill Access In pract ical terms, borehole deviation is a problem for a number of reasons. For example, when attempting to fragment a given amount of rock, it is desirable to dr i l l a hole that w i l l enable the accurate positioning of explosives. If the endpoint of a borehole (i.e. the position of the explosives) is subject to an appreciable amount of uncertainty, ine f f i c i en t a n d some t imes unpredictable blasting can occur. This results in poorly fragmented rock and the overproduction of waste rock, w h i c h leads to increased wear rates for ore processing equipment and dilution of the recovered ore [2], [3]. The importance of being able to produce specific borehole patterns accurately is illustrated in figure 2. Figure 2 - Underground Drilling & Blasting Operation. From the borehole pattern illustrated above it is clear that precise control over the borehole path is of primary significance. 1.2 Long-Hole Drilling Techniques Most long-hole dril l ing equipment makes use of a steel dri l l rod that is subjected to an axial load, pushing it into the ground to generate the down-force required for drilling [3], [6]. When the dr i l l rod has attained its maximum underground depth (the length of the dr i l l rod, typically 1.5 meters (5 feet)), the dr i l l ing 3 operation is suspended while another rod is added to the top of the existing sunken rod to form a continuous dr i l l string. Once the extension is in place and the pneumatic and /o r hydraulic systems are re-pressurized, the d r i l l ing operation can begin again. This procedure is repeated, adding additional dr i l l rods to further extend the string until the desired maximum depth has been achieved. Each particular long-hole drilling operation w i l l vary in its details (drill bit type, hole diameter, top hammering vs. down-the-hole dril l ing, etc.) having some impact on the magnitude of deviation, however they all follow this same general procedure and are all mostly capable of the same approximate level of performance. Consequently, they all suffer from the same general drawbacks [3]. Long-hole drilling is not a quick procedure. The drill ing itself progresses at a rate of approximately 0.3m/ minute (1ft/minute) [3], however the overall operation requires considerably more time due to the mandatory work stoppages. Even a modest 30-meter (98-foot) hole wou ld require 20 stoppages for dr i l l string extensions (at 1.5 meters (5 feet) each) and associated activities. More time is spent preparing the dr i l l than is spent waiting for each segment of the dr i l l string to reach its target depth. One argument against real-time borehole profiling systems is that even if dr i l l operators could be alerted to deviations as they were occurring, there is very little they could do in terms of corrective action. Still, once dri l l ing is complete using conventional bl ind techniques, a considerable amount of time must be spent determining the actual location of each borehole endpoint [6]. A real-time tracking solution would eliminate this extra step. A l l of this time spent determining the borehole endpoint location and preparing dri l l rod extensions coupled wi th the dilution of recovered ore and premature equipment wearing leads to expenditures on the order of several hundred thousand dollars per year for a given mining operation [2]. These expenditures 4 c o u l d effectively be e l imina ted p r o v i d e d t w o key issues were addressed: the m i n i m i z a t i o n of borehole deviat ions and the e l i m i n a t i o n of m a n d a t o r y w o r k stoppages for m a n u a l d r i l l preparat ion and hand l ing [3]. 1.3 The Robotic, Guided Dri l l To address these issues, a n e w a p p r o a c h to long-hole rock d r i l l i n g was conce ived . A n i n -the-hole, robotic gu ided d r i l l connec ted to the surface b y a f l ex ib l e u m b i l i c a l that houses t he p n e u m a t i c / hydraul ic and electrical Figure 3 - The Robotic, G u i d e d D r i l l . Several features distinguish this design from current techniques - chief among l ines c o u l d po ten t i a l ly t h e m / t h e flexible umbilical and steerable end effector. solve each of these prob lems at once (figure 3) 2 . Instead of axia l thrust o n a d r i l l r od , the d r i l l ' s robot ic e n d effector c o u l d generate the r e q u i r e d down- fo rce local ly i n the borehole. In rep lac ing the steel d r i l l rods w i t h a flexible u m b i l i c a l , the robotic d r i l l c o u l d r u n cont inuously wi thout any need for w o r k stoppages. In addi t ion, due to extendable feet on the d r i l l end effector a n d the flexible nature of the umbi l i ca l , it w o u l d be possible to steer the d r i l l accord ing to a p rescr ibed path, or to main ta in one. A sequence of six steps comprises the operat ion of the robotic end effector, as i l lus t ra ted i n f igure 4 (next page). The m o t i o n is ve ry s i m i l a r to that of a n i n c h w o r m . Firs t (figure 4 A ) , the rear feet are he ld f i r m l y against the borehole 2 A l l information pertaining to the robotic dril l was obtained from the following patents [3], [7], [8], except where otherwise noted. 5 Figure 4 - Robotic Dr i l l ing Sequence. The robotic end effector moves in a pattern very similar to that of an inchworm. A sequence of six steps comprises the operat ion of the robotic end effector, as i l lustrated i n figure 4 (next page). The m o t i o n is ve ry s imi la r to that of an i n c h w o r m . First (figure 4 A ) , the rear feet are he ld f i r m l y against the borehole w a l l to p rov ide stabil i ty d u r i n g d r i l l i n g w h i l e the end effector midsec t ion s l o w l y elongates (figure 4B), p r o v i d i n g the necessary down-force for the percussive d r i l l head against the rock face. Once the midsec t ion is fu l ly elongated, the front feet extend (figure 4C) to p r o v i d e stabil i ty as the rear feet disengage f rom the borehole w a l l (figure 4D). The midsec t ion then contracts (figure 4E) to advance the end effector d o w n the borehole. D r i l l i n g is cont inuous throughout this operation. A t this point the rear feet extend to engage the borehole w a l l f rom this new, lower pos i t ion i n the borehole (figure 4F) and the entire procedure is repeated. In this manner, the robotic d r i l l can ' w a l k ' th rough the rock, d r i l l i n g cont inuously , and w i thou t the need for per iodic m a n u a l in tervent ion as w i t h t radi t ional long-hole d r i l l i n g techniques. The presence of retractable feet o n the robotic d r i l l also presents a potent ia l so lu t ion for excessive borehole devia t ion. If data c o u l d be gathered concerning 6 target, straight-l ine pa th i n the event that deviat ions became significant, or even gu ided a long a cu rved pa th if so desired. This concept presupposes the presence of a sensor sys tem capable of detect ing the d r i l l ' s trajectory i n real-time. Cur ren t ly , the robotic d r i l l prototype does not feature such a system. It is the purpose of this research to fu l ly examine the technical and operat ional constraints that must be met by any proposed t rack ing system, to out l ine a feasible means of p r o v i d i n g this functionali ty based o n these constraints, and to offer an in i t i a l evaluat ion of the proposed solut ion. A l t h o u g h the robotic d r i l l has m a n y n o v e l features - the flexible u m b i l i c a l and retractable feet chief a m o n g them - their significance cannot be real ized i n the absence of an effective pos i t ion sensing system. 7 2.0 Target Performance and Design Constraints As with any kind of design, before particular approaches can be conceived and pursued, context must be established to provide each potential solution wi th a framework for evaluation and comparison. In the case of the robotic dr i l l , this context comes in three forms: target performance, indicating an idealized sense of expected dri l l capabilities; constraints introduced by the particularities of the overall dr i l l design as it currently exists; and constraints imposed by the drill 's operational environment. Although only one type of application for the robotic dr i l l has been explicitly mentioned thus far (i.e. the drill ing of blast holes), there are many settings where this type of technology can find use. Addi t iona l applications include core sampling [6], pinpoint ore leaching [9], Horizontal Directed Dri l l ing (HDD) [6], underwater mining [6], and dril l ing on other terrestrial bodies throughout the solar system [6], [10], among many others. A s such, the above-mentioned framework of performance targets and design constraints should be as broad as possible to properly characterize the wide range of potential applications that this dri l l is suited for. While a single overall design won't be able to completely satisfy all potential applications, the intent is to develop an approach to the position sensor system (PSS) that would require very little modification for use in different operational settings. 2.1 Target performance As a general rule of thumb, and as previously cited, most current long-hole drilling systems are capable of accuracies on the order of approximately 10% of the dril l ing depth. In other words, when dri l l ing a 200m (656ft) borehole, one could expect the endpoint to have deviated from the intended, straight-line path by about 20m (65.6ft) in any direction. The robotic dr i l l , wi th its potential for 8 correc t ing dev ia t ions th roughou t the d r i l l i n g process, has an in tended target accuracy of 0.15% 3 of the d r i l l i n g depth, or 30cm (11.8in) of devia t ion at a depth of 200m (656ft). If at tainable, this l eve l of per formance w o u l d represent a q u a n t u m leap over ex i s t ing long-ho le d r i l l i n g techniques, a l l o w i n g for near p i n p o i n t accuracy. E n a b l i n g operators w i t h this leve l of con t ro l and f lex ib i l i ty w o u l d lead to new long-hole d r i l l i n g strategies that w o u l d reduce wastage both i n terms of p roduc t ion and expenditure. In pract ica l terms, wha t does a 0.15%-of-depth dev i a t i on represent? In other w o r d s , w h a t w o u l d any p r o p o s e d sensor sys tem ac tua l ly measure w h e n de te rmin ing the d r i l l ' s pos i t ion underg round? To answer this, w e can begin by examin ing the geometry of the borehole path. Ideally, this pa th is a straight l ine (vert ical or otherwise) , h o w e v e r d r i l l s tr ings tend to deviate i n a c u r v e d pa th from where they were in tended to e n d up . A s a r o u g h approx imat ion , a circle can represent this d e v i a t i o n c u r v e , d e f i n e d by t w o l o c a t i o n s o n i ts per imeter (the borehole start and end points) w i t h the radius located on the surface at the same e leva t ion as the borehole s tar t ing po in t (figure 5). If a vert ical l ine is extended upwards to the surface f rom the borehole end poin t , the sys tem is represented as a r ight-angle tr iangle. The radius of curvature can therefore be calculated as: Figure 5 - Deviat ion Geometry. The dots represent the dr i l l ing start and end points, and the center of the deviation's radius of curvature. 3 The 0.15%-of-depth criteria is based on the statement that the d r i l l features an accuracy of approximately +/- 0.3m (11.8in) at a depth of 200m (656ft), from [3] and [8]. 9 R = T]D2 + (R-d)2 [Eq. 1] where R is the circle radius, D is the borehole depth and d is the deviation. Since the deviation d is a function of the depth D, Equation 1 can be re-written as: R " ~ [Eq. 2] 2p where p is the deviation rate (i.e. 0.0015 for a 0.15%-of-depth deviation). With both the depth and deviation known, this relation can be broken down into a simple expression yielding the circle's radius. A deviation of 30cm (11.8in) at a depth of 200m (656ft) representing a 0.15%-of-depth deviation results in a 66.7km (41.4mi) radius of curvature. Although this figure represents only an approximation, it is apparent that the umbilical bend is incredibly minute and that any system proposed for detecting this bend must be very sensitive. Continuing with this geometric analysis, an approximate minimum allowable angle of tilt can also be calculated. Since it is assumed that the end effector is always tangent to the deviation circle, the approximate angle of tilt can be determined by drawing a tangent line at the drill endpoint location and calculating its angular offset from the vertical (figure 6). This angular offset is calculated as follows: i 1 R - d Figure 6 - M a x i m u m Til t . The white line represents the maximum angular offset from the vertical at the borehole endpoint, which is tangent to the deviation curvature. 1 0 6 = sin - i ^ [E<.3] where D is the borehole depth and R is the radius of curvature. For example, at a depth of 200m (656ft) and wi th a 66.7km (41.4mi) radius of curvature, the approximate angle of tilt at the borehole endpoint is 0.172 degrees (note that the 0.15%-of-depth deviation is implied by the inclusion of the radius of curvature, which is calculated based on that percentage deviation). However , this represents the maximum approximate angular offset likely to be encountered while dr i l l ing since the assumed end effector location is at the end of the borehole. If the location is assumed to be at an intermediate stage during the drilling procedure, the angle is less pronounced and therefore more difficult for sensors to detect. It would be much more desirable to monitor the angle of tilt at specified depths throughout the dri l l ing procedure rather than only once at the end of dril l ing. Therefore, any sensor used to detect the angle of tilt must be sensitive enough to detect minute offsets before they have the opportunity to accumulate into significant deviations that may be more difficult to overcome. 2.2 Dr i l l design-based constraints Generally speaking, the robotic dr i l l consists of three main components: the end effector, the umbilical, and the service unit (figure 7, next page). The end effector in its current embodiment is approximately 1.5m (5ft) tall, wi th a diameter of about 21.6cm (8.5in). The flexible umbilical is roughly the same diameter as the end effector, and houses all of the required the pneumatic and/or hydraulic lines and electrical conduits. The length of the umbilical can vary depending on the application, however the maximum depth is l imited only by practical spool 11 capacities and by the resources required to remove the end effector and u m b i l i c a l f rom the borehole once operations are complete. The service un i t inc ludes the umbi l i ca l spool , end effector pos i t ion ing mast, and a t readed m o b i l i t y system, a n d houses a l l necessary systems for s u p p o r t i n g the p o w e r a n d c o n t r o l requirements of the end effector. Figure 7 - The Robotic, Guided Dr i l l , Redux. It should be noted that the fo l l owing design constraints are largely qual i tat ive i n nature. O v e r a l l , the robotic d r i l l is presently i n a h i g h l y exper imenta l stage and many u n k n o w n s s t i l l exist. Despite this lack of concrete specifications, examin ing the var ious des ign constraints is a useful exercise i n that, at the ve ry least, i t provides rough margins that must be accommodated i n the PSS design. 2.2.1 Vib ra t ion and Shock The current embod imen t of the des ign calls for the use of a percuss ive d r i l l , w h i c h w i l l resul t i n subjecting a l l end effector a n d u m b i l i c a l components to substantial v ib ra t ion and shock. A nove l shock absorber has been incorpora ted 12 into the des ign of the end effector [7], however i t is l i k e l y safe to assume that v i b r a t i o n d a m p i n g w i l l not be 100%. A s such, any potent ia l p o s i t i o n sensing sys tem des ign that m a y be p a r t i c u l a r l y suscept ib le to v i b r a t i o n s h o u l d be considered less favourably than those that are inherent ly more vibration-tolerant. A s p rev ious ly noted, the robotic d r i l l is to be operated con t inuous ly th roughout a l l phases of its operat ion. It m a y be argued that s topp ing the d r i l l w o u l d a l l o w for the col lec t ion of vibration-free measurements. H o w e v e r , this stoppage m a y require t ime- in tens ive re -pressur iza t ion of h y d r a u l i c and pneumat ic l ines, i n add i t i on to potent ia l ly signif icant wa i t times associated w i t h the cessation of a l l r e s idua l d r i l l head osci l la t ions. A t present, i t is not fu l ly k n o w n w h a t impac t occasional d r i l l stoppages migh t have on the overa l l operat ion, howeve r it m a y prove ins ightful to perform a cost /benef i t analysis of this procedure i f v ib ra t ion is found to be a c r i pp l i ng factor concerning PSS implementa t ion . 2.2.2 V o l u m e & Shape The d r i l l has been des igned such that the gap between the end effector / u m b i l i c a l a n d the borehole w a l l w o u l d be 1.3cm (0.5in) o n average [6], a cknowledg ing the fact that the d r i l l i n g process w i l l no t result i n a borehole w a l l w i t h a smoo th su r face ( f igu re 8). T h e m a x i m u m a v a i l a b l e space i n s i d e the u m b i l i c a l w o u l d be a p p r o x i m a t e l y 3.8cm (1.5in), measu red f rom the u m b i l i c a l ' s in te r ior w a l l to the condui t s connect ing the end Figure 8 - Available Volume. This is a top -down v i ew of the borehole / umbi l ica l cross-section. The umbil ical diameter is 21.6cm (8.5in). 13 effector to the service unit on the surface [6]. There will likely be no space available for the addition of a position sensor system in or on the end effector itself. Clearly there is not much room available, either internally or externally, for the inclusion of a sensor system in terms of width or depth, although the height of any additional equipment is not explicitly limited. It should also be noted that any additional equipment might have to take the annulus curvature into account, due to the diminutive gap spacings involved. 2.2.3 Mass It is possible that some potential PSS concepts could involve the addition of large amounts of mass, taking advantage of the relatively unrestricted height available. Such mass could come in the form of proof masses, the use of large quantities of fluids or gels, etc. However, one of the unique overall design characteristics of the robotic drill is its low mass, compared to traditional systems. This feature plays a crucial role in several applications, including drilling on other planetary surfaces. Even on Earth it would be desirable to keep overall mass to a minimum, ensuring that the system remains as mobile and adaptable as possible. 2.2.4 Cost Simply because a given problem may be technically challenging, it doesn't mean that it must be solved with a large outlay of financial resources. While exorbitantly priced solutions may exist, emphasis should be placed on stressing the limits of lower-cost components. 1 4 2.2.5 Unknown Parameters Concerning Drill Design There are many parameters concerning the design of the robotic drill that are yet to be determined. Such parameters include the stiffness of the umbilical, the drilling rate, the rate of travel for the end effector's front and rear feet, the rate of the end effector's midsection expansion / retraction, vibration characteristics and the amount of end effector slippage in the borehole, among others. Without more comprehensive and quantitative specifications, the position sensor system can only be designed and implemented to a first-degree approximation. Still, this initial design should be of significant value to the overall project and is intended to serve as a jumping-off point for further, more detailed development. The reason for this lack of concrete information is that the drill prototype is still in a highly experimental stage. As testing progresses, more quantifiable information concerning the drill will become known and will help to round out what is to be expected from associated systems, such as the PSS. A summary of the most significant drill specifications can be found in table 1. Drill Diameter 21.6cm (8.5iu) Approximate Borehole Diameter 24.1cm (9.5in) Available Width inside Umbilical 1.27cm (0.5in) Approximate Drilling Rate 0.3m/min Target Deviation +/- 30cm @ 200m depth Maximum Tilt at Target Deviation 0.172 degrees Table 1 - Drill Specifications Summary 2.3 Operational Environment-based Constraints To further the goal of establishing a reference framework for conceiving and evaluating position sensor system designs, a significant amount of insight can be gleaned by examining the operational settings of the robotic drill. As with the design-based constraints, the exact quantitative nature of the most likely 15 operational environments is not known. However, since the overall goal of this work is to develop a first-approximation design, the information currently at hand concerning these environments will be generally sufficient. 2.3.1 Cuttings Disposal To dispose of cuttings during operation of the robotic drill, compressed air will be pumped into the borehole, blowing excess material back out the top of the hole [3]. This has a profound effect on any sensor system that might rely on detecting the external environment of the drill end effector. Rock, dust and drilling fluid will obscure the borehole wall and would present a significant challenge for any exposed moving parts and seals. 2.3.2 Temperature Geothermal gradients could contribute to signal error, depending on the particular sensor's susceptibility to changes in temperature. Even over a 200m (656ft) depth, temperatures could be expected to increase by approximately 5°C on average, depending on location and local geology [11]. Sophisticated electronics can be affected by even slight changes in temperature, and finely tuned mechanical systems may be subject to minor, yet potentially deleterious material deformations. If fluids are incorporated into the PSS design, attention must be paid to the effects of temperature especially when operating in particularly cold or warm environments. The cuttings may also introduce thermal effects as they are being disposed, where the drilling fluid, waste rock and compressed air are in contact with the umbilical along its length until being blown out of the borehole. Although these effects can in large part be filtered out with signal processing or mitigated with insulation, the potential effect of thermal drift as a source of signal noise must nevertheless be considered. 1 6 2.3.3 Magnetic Fields It is possible that some of the rock types being drilled through could be ferrous, generating local magnetic fields [6], [9]. At the very least, these fields may interfere with detecting the Earth's global magnetic field for use as a fixed navigational reference. In the worst case, strong local magnetic fields could require the use of shielding to prevent interference with on-board electronics. From the perspective of PSS design, it may not be wise to rely upon magnetic field detection as a means for determining end effector position. 2.3.4 Scalability Depending on the approach, being able to accurately detect the deviation for a 500m (1640ft) borehole may be considerably more difficult than detecting the deviation for a 30m (98ft) borehole, or vice versa. Ideally, the position sensing system should be intolerant to the application's scale and should not require re-calibration depending on the target depth. In addition, the general philosophy of the position sensor design should remain the same regardless of the physical scale of the drill itself. For example, the drill's overall dimensions would most likely be scaled down (with reference to its current embodiment) for inclusion in missions of scientific exploration on the Moon or Mars. For simplicity, and perhaps as a sign of legitimacy, the overall approach to the sensor system design should remain the same, provided that the extreme environment (reduced gravity, radiation exposure, atmospheric differences, etc.) does not dictate otherwise. Similarly, if the drill's dimensions are to be increased for larger scale tasks, the basic PSS design should be robust enough to accommodate these changes without requiring a complete re-think of its underlying premise. 17 2.4 Constraint Summary There are a number of factors restricting potential PSS design configurations, some more pressing than others. At this initial stage, volume, sensor accuracy and tolerance to vibration and shock are seen as the most significant metrics affecting the design. As more becomes known about the exact nature of the drill and its operational environment, a more quantitative evaluation of these parameters can take shape. However, even with the information at hand, enough is known to make general determinations as to what would constitute a technically feasible approach. 18 3.0 Conceptual Solutions With a certain degree of familiarity having been established concerning the parameters that the PSS must adhere to/ existing solutions from similar applications were examined for inspiration concerning the task at hand. Patent searches were conducted to ascertain how the issue of orientation and position determination is handled in traditional long-hole drilling operations and related fields. The information gleaned from this investigation was applied to the specifics of the robotic drill in an attempt determine whether the existing approaches are suitable for incorporation into the PSS or if a new approach is warranted. In a sense, this section is an overview of the tools available for solving the problem of determining drill position and orientation. 3.1 Traditional Guidance Sensors Patents relating to horizontal directed drilling [12]-[16], offshore oil drilling [17], hydrocarbon drilling [18]-[20] measurement-while-drilling systems [18], [21], [22] and borehole surveying technologies [22], [23] were investigated. These applications typically feature four different types of sensors used in different combinations, numbers and configurations for determining orientation and position: • Clinometers (tilt meters) • Magnetometers • Accelerometers • Gyroscopes This section will present an overview of each of these sensor types to determine their general suitability for tracking the robotic drill throughout its operation. 1 9 Subsequent sections will discuss the most promising of these and other sensor types to determine the most feasible approach for PSS implementation. 3.1.1 Clinometers (tilt meters) High-end clinometers have found widespread use in large-scale oil and gas drilling [17]. Although they are available in a wide variety of configurations, clinometers typically employ a proof mass or solid-state equivalent that provides a constant reference, indicating the direction of gravitational acceleration. Tilt is then measured with respect to this constant external reference. Clinometers can be extremely sensitive to changes in orientation, capable of resolving 0.005 degrees or better [24], and many are available in biaxial configurations. In the case of the robotic drill, clinometers would only solve half of the position-sensing puzzle: end effector orientation. The exclusive use of clinometers would offer no help in determining the underground distance traveled by the drill throughout its operation. According to the navigational technique known as 'dead reckoning' [25], the task of position determination can be divided into two halves: displacement and orientation. The path of an object can be broken down into a series of linear steps, and if the length of each step can be measured in addition to the orientation held over the course of each step, the entire path can be recreated to yield the final position and heading of that object. More simply, an object's position is known if the distance traveled by an object and the direction in which it has traversed this distance are both known. Relating this back to the robotic drill, clinometers would be able to relate the direction of travel, but not the distance covered by the end effector. Fortunately, the drill's umbilical offers this information - the underground distance covered by the drill is known by how much of the umbilical has been pulled off of the aboveground spool. Combining this underground distance with tilt data from clinometers would, in theory, yield the drill's underground position. 20 Despite this potential, clinometers feature two significant drawbacks: size and susceptibility to vibration and shock. The kind of highly accurate clinometers required for measuring the minute angular offsets described in section 2.1 tend to be quite large, or at least larger than what could be accommodated inside the drill umbilical. For example, the Columbia Research Laboratories SI-701FNE clinometer, capable of resolving angles as small as 5 x IO 5 degrees, has a footprint measuring 4.1cm (1.6in) wide x 10.5cm (4.1in) long [26]. While its capabilities may be more than adequate for detecting the minute angular offsets that the drill can produce, this footprint wouldn't fit in the 3.8cm (1.5in) wide x 8.3cm (3.3in) long area available within the umbilical. And while these clinometer dimensions may seem close enough to those of the available volume, note that the umbilical curvature has not been taken into account - this significantly reduces the space available to the point where shoehorning a clinometer into this volume is no longer a viable option. Clinometers have found widespread use in industry since large component sizes can be accommodated more easily in larger scale operations such as oil drilling, where more volume is available. Also, all PSS components will be subjected to vibration and shock induced by the end effector's percussive drill head. Many types of clinometers would not react well in this type of environment, particularly the smaller clinometers that are capable of fitting within the volume available inside the umbilical [27]. In this regard, clinometer use represents a catch-22 proposition. 3.1.2 Magnetometers There are three main approaches to taking advantage of magnetic fields for orientation sensing in underground settings. The first is used primarily in horizontal directed drilling (HDD) - a minimally invasive technique for boring holes for underground sewage lines, communications conduits and other types 21 of infrastructure, and is often employed instead of trenching [12]. Although several variations exist, the general approach includes a magnetic field generator located in or on the underground drill head, with aboveground operators detecting the signal with handheld equipment [13], [28]. The strength of the signal measured by the operator indicates the drill head's depth and position, taking advantage of the fact that the boreholes being drilled are horizontal and relatively close to the surface. This technique does not provide information concerning finer movements of the underground drill head, such as pitch, and therefore this approach often incorporates accelerometers (described below) for higher accuracy. Considering that HDD operations generally do not take place more than a few meters below the surface, a magnetic beacon system such as this is not applicable to the design of the robotic drill. When drilling to significant depths, the minute deviations from the vertical would be extremely impractical to detect from the surface when compared to other potential methods. The second approach involving magnetometers is often used in oil and gas drilling. This technique simply measures the Earth's magnetic field, using it as a constant directional reference in much the same way that a handheld compass would [13], [14]. Unlike oil and gas drilling operations however, the robotic drill will likely have to drill through hard rock containing ferrous material that would obfuscate magnetometer readings. In addition, even the drilling equipment itself could emit deleterious magnetic interference rendering a magnetometer considerably less useful [28]. Vector Magnetics Inc. of Ithaca, New York has developed several means for accurately guiding boreholes to depths in excess of 500m (1640ft) using a variation of the above two magnetic field-related schemes. One such technique involves drilling a reference borehole, retracting the drill string, and pulling a guide wire into the borehole that is connected to a source of current [20]. Once in the borehole, this current-carrying wire produces a magnetic field that can be 22 used as a reference for creating adjacent, parallel boreholes. Other techniques developed by Vector Magnetics include determining the direction and distance to a target well bore: current is run through the steel casing within a target well, and magnetometers are used to guide a drill in a separate borehole toward the target [19]. This technique is used in drilling relief wells to control oil well blowouts, and in well avoidance. These artificial beacon techniques do not translate to the present problem simply because the robotic drill will not typically be operating in an environment that includes pre-existing infrastructure or other features that could be targeted. While targetable features may occasionally be present, nominal operation of the drill should not depend on it. 3.1.3 Accelerometers Accelerometers are devices that measure the rate of change of an object's speed, and some are also capable of detecting static accelerations (i.e. gravity). As such, accelerometers are used in myriad applications related to position sensing, impact detection, tilt measurement and vibration monitoring among many others. There are a number of different approaches to accelerometer design, however capacitive accelerometers are currently among the most popular due to their excellent performance and relative low-cost [29]. Capacitive accelerometers employ small, micro-machined elements that produce a capacitance [29]. When these elements are accelerated, deflections occur that alter the capacitance in a manner that is related to the acceleration. Internal electronics then convert this capacitance to an output voltage, indicating the magnitude of acceleration. This signal can be integrated over time once to obtain velocity, and a second time to determine displacement. As such, accelerometers 23 have p r o v e n v e r y effective i n app l i ca t ions re la ted to p o s i t i o n sens ing and navigat ion. In terms of t racking the robotic d r i l l , accelerometers h o l d m u c h promise . D u e to the i r v e r s a t i l i t y , there are at least t w o p r i m a r y app roaches to u s i n g accelerometers that are appl icable to the robotic d r i l l . The first approach calls for a set of accelerometers to be used i n conjunct ion w i t h gyroscopes (described i n greater de ta i l i n the f o l l o w i n g section) i n w h a t is referred to as an Inert ia l M e a s u r e m e n t U n i t , o r I M U [30]. T y p i c a l l y these u n i t s feature three accelerometers and three gyroscopes to detect three degrees of f reedom for pos i t ion (x-, y- , z-motion) and three degrees of freedom for orientat ion (rotations about the x, y a n d z axes). I M U s featur ing reasonable accuracy are t y p i c a l l y b u l k y units (relative to the v o l u m e avai lable i n the d r i l l umbi l i ca l ) [30], however an I M U c o u l d be assembled ' f rom scratch' u s ing smal ler i n d i v i d u a l sensors that are not a l l necessarily contained w i t h i n the same protective enclosure, to reduce vo lume . The acceleration output from the I M U w o u l d have to be integrated twice over t ime to obta in pos i t ion . A l t h o u g h d r i l l accelerations are not current ly k n o w n , the figures are expected to be quite sma l l since the overa l l speed of the end effector is approx imate ly on ly 0 . 3 m / m i n (12 in /min ) ; any changes i n this speed, i n c l u d i n g from rest to ful l -speed operat ion, w i l l be ve ry sl ight. A l t h o u g h there are m a n y sensors capable of detecting such sma l l accelerations, the constant and v io len t end effector v ibra t ions w i l l render acceleration measurement cha l lenging at best. It s h o u l d also be noted that any error i n the detected measurement w o u l d be c o m p o u n d e d further by the integrat ions r equ i red to obta in pos i t ion f rom the accelerat ion data. If the ove ra l l a i m is to determine h o w far u n d e r g r o u n d the d r i l l has t raveled, i t w o u l d be easier by far to m o n i t o r the length of u m b i l i c a l dep loyed f rom the aboveground spool , as described i n section 3.2.1. 24 The second approach for u s ing accelerometers to track the robotic d r i l l ' s pos i t ion is to measure static accelerat ion (i.e. the centr ipetal accelerat ion of the Earth 's rotation). In effect, some types of accelerometers c o u l d be used as t i l t meters i n m u c h the same w a y that c l inometer use was ou t l i ned above [32]. W h e n the sensing element is pe rpend icu la r to the g rav i ty vector, m a x i m u m outpu t w i l l result. A s the uni t is t i l ted , the sensor w i l l detect o n l y the componen t of the grav i ty vector that is pe rpend icu la r to the sensing element (figure 9), un t i l the sensor is para l le l w i t h this vector and no output results. 1 Ful l gravity vector Partial gravity vector F igure 9 - Sensing Static Acce le ra t ion . The accelerometers (grey rectangles) produce output proportional to the angle of the gravity vector w i th respect to the sensing element, going to zero when the two are parallel. A c c e l e r o m e t e r s c a p a b l e of ve ry respectable performance i n m e a s u r i n g s t a t i c acce lera t ion are ava i l ab l e i n v e r y s m a l l fo rm factors. For example , the A D X L 2 0 2 E f rom A n a l o g Devices measures just 5 m m (0.2in) w i d e x 5 m m (0.2in) l ong x 2 m m (0.08in) tall [31] and is sensit ive e n o u g h to detect a change of 15 mi l l i - g ' s per degree of t i l t w i t h i n an approximate range f rom 0 (horizontal) to 30 degrees [32]. H o w e v e r , i f the robotic d r i l l were to t i l t 0.172 degrees f rom an in tended pa th (the m a x i m u m l ike ly to be encountered), g iven a rated sensi t ivi ty of 0 . 0 0 3 g / m V [31], i n theory this sensor w o u l d output on ly 0.86mV. If no vibra t ions were present this m a y be sufficient, however such a smal l s ignal w o u l d not be idea l i n a r ea l -wor ld setting. M o r e sophis t ica ted accelerometer conf igura t ions w i t h 0.1 u,g resolut ions can detect angular offsets of 0.005 degrees [33] - a leve l of sens i t iv i ty m u c h better sui ted to the robot ic d r i l l - h o w e v e r this l eve l of per formance comes at a s igni f icant cost i n terms of bo th c o m p l e x i t y of c o n f i g u r a t i o n and r equ i r ed v o l u m e [33]. 25 In addition to volume and complexity, accelerometers have further drawbacks. For example, most common accelerometers capable of measuring static accelerations are generally better suited to measuring tilt when the initial position has zero offset with respect to gravity [32]. If the intended borehole path is not vertical, the sensor's initial starting orientation wi l l be offset from the vertical and the sensor's usefulness in this situation may be limited. Also, when relying on detection of the Earth's centripetal acceleration, one must be aware that this value changes depending on the latitude where the measurement is taken. While not an insurmountable issue, this fact adds to the complexity of incorporating accelerometers into the PSS design. 3.1.4 Gyroscopes Gyroscopes are sensors that, provided a given input voltage, wi l l output a voltage (or current) proportional to the rate at which they are rotated, in degrees (or radians) per second. There are a number of different implementations of this basic premise/each with their own advantages and drawbacks4. Ring Laser Gyros (RLGs) and Fiber Optic Gyros (FOGs) are the two most accurate types of gyroscopes available [34]. Both involve counter-rotating light pathways produced by a laser and take advantage of the Sagnac Effect: when the sensor is rotated, a frequency difference is generated between the two beams - the beam rotating in the same direction as the sensor shifts toward the red end of the spectrum, and the counter-rotating beam shifts toward the blue. This frequency difference is proportional to the rate at which the sensor is rotated [34]. Some FOGs, such as those used in military applications, are sensitive enough to accurately detect one rotation in approximately 137 years [35]. However, this level of performance comes at significant financial cost (approximately 4 Many types of gyros exist that are not discussed in this work, including mechanical gyros and piezoelectric gyros. White these technologies remain in widespread use in some applications, they are either being phased out by newer technologies (as in the case of mechanical gyros), or are becoming outclassed by the technologies discussed in this work [34]. 26 $10,000USD/unit [33]) and logistical complications (the technology is subject to International Trade in Arms Regulations (ITAR) [33], among other restrictions). Another type of rate sensor employed in drilling applications is the micro-electromechanical systems (MEMS) gyro. Even within this classification there are a wide variety of configurations. Generally speaking, MEMS rate sensors are based on the Coriolis effect - forces that are observed when linear motion occurs within a rotating frame of reference. Coriolis gyros are comprised of an oscillating beam vibrated to produce a standing wave along its length. When subjected to a rotation, this standing wave is displaced on the beam due to the Coriolis force. The motion induces vibration in a new direction, and that change in direction is proportional to the sensor's rate of rotation [36]. Depending on the specific sensor, the oscillating material can take the form of micro-machined beams, tuning forks or rings, all exhibiting varying levels of performance. MEMS gyros are not as accurate as RLGs or FOGs, however they can be found in widespread use in applications requiring a small form factor, reasonable performance and low-cost. For example, the CRS03 gyro from Silicon Sensing measures 29mm (1.14in) x 29mm (1.14in) x 18.4mm (0.72in) and costs approximately $150USD per unit [37]. In current drilling applications, gyroscopes are used in multi-axis configurations and generally in conjunction with either accelerometers (as part of an inertial measurement unit), or with magnetometers (to provide backup when magnetic interference is significant) [13], [14]. As with accelerometers, there are numerous methods for employing gyroscope that may prove useful for tracking the robotic drill. Recalling the above discussion concerning dead reckoning and the breakdown of position determination into orientation and distance traveled, it is not necessary to include sensors for measuring the displacement of the drill end effector since 27 the amount of umbilical deployment already provides this information. Therefore it is unnecessary to pair displacement sensors with gyroscopes/as is generally the case in other applications. As a result, there are two main approaches for using standalone gyroscopes that could provide information concerning the drill's orientation. As discussed in section 2.1, although the drill deviates from the intended path following an approximately exponential profile, this curve can be approximated with a circle where the borehole start and end points lie on the perimeter. Also recall that the robotic drill is intended to operate continuously, moving through rock at approximately 0.3m/min (12in/min). If the drill were to follow a path featuring a radius of curvature of 66.7km (41.4mi) - the result of drilling to a depth of 200m (656ft) with a 0.15%-of-depth deviation - the drill would travel around this circle at an angular rate of approximately 4.3 x IO 6 deg/sec. Many gyros are capable of detecting such small angular rates, however typically all of them are either RLGs or FOGs, and are therefore much too large for the drill umbilical. As an example, the Northrop Grumman FOG 2500 features a drift rate of 2.78 x 10 7 deg/sec, meaning that the inherent sensor output error is an order of magnitude smaller than the drill's angular rate. Despite this, each unit measures 24cm (9.6in) x 27cm (10.7in) x 13cm (5in) and weighs approximately 8kg (181bs) [38]. Although these sensors provide a highly desirable level of performance, the order of magnitude that these dimensions represent is completely out of touch with the robotic drill's constraints. Following the same general approach, if the angular rate were more rapid, a smaller, less expensive gyro could be used instead. According to Equation 4, there are two parameters that affect the angular rate - the drill's speed and the radius of curvature of the borehole: V = 0)XR [Eq.4] 28 where v is the linear velocity, oo is the angular rate in radians/second, and R is the radius of curvature. If the drill could move faster, the rate would increase and less sensitive equipment could be used for detecting motion. This is not a viable option. Nor is the idea of allowing for a tighter radius of curvature - this simply leads to larger borehole deviations. Alternatively, the gyro could be made to move faster, independent of the drill's progression though the rock. If the gyro were able to quickly shuttle up and down the length of the umbilical, the borehole profile could be measured using much less sophisticated sensors due to the increased angular rate5. Equation 4 can be used to determine how fast the gyro shuttle would have to move in order to produce lmV of sensor output, ignoring all sources of noise and other inaccuracy. Assuming the use of a gyro featuring a scale factor of lOOmV/deg/sec and a 66.7km radius of curvature, the gyro shuttle would have to move at approximately 11.6m/s in order to produce lmV of output. Accurately and reliably displacing a cluster of gyros at such a speed would be a significant challenge at the very least, especially considering the volume available within the drill umbilical. In addition, required shuttle speeds depend on drilling depth - if the depth of the hole is reduced, the radius of curvature would be smaller and the gyro could move at a slightly slower speed. Conversely, if the intent were to drill to greater depths, the gyro would require even greater velocity to produce a reasonable output. Adding to the technical drawbacks, this dependence on borehole particulars reduces the system's versatility and is therefore not desirable. Alternatively, gyros could also be used to measure end effector tilt in much the same manner as accelerometers or clinometers. However, employing gyros would involve a slightly different tack than with these other sensor types. 5 Dr. G. Baiden proposed this concept as one possible approach to developing a position sensing system when this project was originally presented to the author. 29 Whereas accelerometers and clinometers measure tilt angles directly, gyros measure angular rates. This means that while it may be difficult to directly measure minute angular offsets (irrespective of the sensor type), if the rate at which an angular displacement takes place is fast enough, it might be possible for small, low-cost gyros to be employed successfully. In terms of analyzing this approach, unfortunately there is no data currently available concerning expected angular rates for the drill's end effector. Despite this, reasonable estimations can be made. On average, and as previously stated, a gap of approximately 1.25cm (0.5in) will exist between the borehole wall and the drill umbilical when vertical. In addition, guesses can be made concerning how much time would be required for the end effector's rear feet to push the drill toward the borehole wall to close this gap. For the sake of argument, it will be assumed that this distance can be covered in 1 second, resulting in an approximate linear displacement of 1.25cm/s (0.5in/s). Keeping in mind that the end effector is approximately 1.5m (5ft) tall and if it is assumed that rotations occur about the drill head (i.e. the radius of rotation equals the end effector height), Equation 4 reveals that, in closing this gap, the drill would rotate at 0.477deg/s. If the end effector rear feet move more quickly, the angular rate would be correspondingly greater and therefore easier to measure. Of course, each of the above assumptions can be tweaked, however the numbers chosen are all reasonable given what's currently known about the drill. In addition, although the numbers used in the above calculation were approximate, they demonstrate the general order of magnitude for required gyro sensitivity. Low-cost gyros, such as the aforementioned CRS03, feature scale factors of 20mV/deg/sec [39] and should in theory be capable of outputting 9.5mV if rotated at 0.477deg/s. If this rate could be accurately detected over an end effector linear displacement of 0.5cm (0.19in), the displacement corresponding to 30 0.172deg of tilt at the end effector / umbilical interface, low-cost MEMS gyros may prove to be an effective solution for position sensing. From the present analysis, it appears that low-cost gyroscopes used for detecting tilt rates may hold significant promise for the robotic drill. An adequate level of performance mixed with a reasonably small form factor makes this particular approach worth further consideration. 3.2 Non-Trad i t iona l Approaches to Guidance In addition to the above-mentioned traditional navigation sensors, non-traditional concepts were also conceived and evaluated in the interests of thoroughness. The following eight approaches all attempt to take advantage of either the drill's motions or its environment to determine position, with varying degrees of feasibility. 3.2.1 External Camera This approach involves the use of cameras aimed radially outward from the robotic end effector. As the drill moves down the hole, the cameras would capture images in rapid succession. Each image would then be compared with the previous image to detect any differences indicative of motion - similar in principle to an optical computer mouse [40]. In theory, if the drill were veering off course, more motion would be detected in one camera than the others. The amount of relative motion detected by each camera could then be translated into heading and orientation. Combined with information concerning the current depth of the borehole, position would be known. In addition to images taken in the visible spectrum, this technique could employ any wavelength, in theory. However, one must keep in mind that significant 31 amounts of mater ia l w i l l be b l o w n out of the borehole by compressed air a n d w o u l d l ike ly b lock any cameras - no matter wha t w a v e l e n g t h is e m p l o y e d -f rom i m a g i n g the borehole w a l l . Since the borehole cut t ings are made of the same mater ial as the borehole w a l l , it m a y prove p roh ib i t ive ly diff icul t to resolve one from the other. In addi t ion , the cutt ings may not a lways t ravel at a constant rate m a k i n g the process of detecting the d r i l l ' s minute l inear m o t i o n cha l lenging at best. For these reasons, it was dec ided that external cameras d i d not represent a h igh ly feasible solut ion. 3.2.2 External Rol le r s 6 In the same v e i n as e x t e r n a l cameras, external rol lers c o u l d be bu i l t into the end effector such that they mainta in constant contact w i t h the borehole w a l l (figure 10). A s the d r i l l progresses d o w n the hole, the ro l l e r s w o u l d rotate. In the presence of d e v i a t i o n , one ro l l e r w o u l d rotate more than the others ind ica t ing the amount of devia t ion . Figure 10 - External Rollers. Deviations from the intended borehole path w o u l d be detected if more displacement were measured w i t h one roller than with the others. A s w i t h the external cameras however , the debr is be ing b l o w n ou t of the borehole w o u l d l i ke ly pose a s ignif icant challenge - dus t a n d d r i l l i n g f lu id c o u l d interfere w i t h exposed m o v i n g parts. A d d i t i o n a l l y , the borehole w a l l may not be a smooth, hard and cont inuous surface. V o i d s , bumps and whee l s l ippage c o u l d a l l combine to render the process of translat ing whee l rotat ion into d r i l l head ing and or ientat ion p roh ib i t ive ly cha l lenging . F ina l ly , i m p l e m e n t i n g a rol ler system 6 Dr. J. Yan originally suggested this concept as one possible approach to developing a position sensing system. as described would require a significant and highly undesirable re-design of the existing end effector prototype since it is unlikely that the required mechanisms could fit within the umbilical's limited volume. 3.2.3 Biomimetic Solutions As alluded to in Section 1.0, it is very easy to ascribe biological traits to the behaviour of the robotic drill. As such, inspiration for new approaches to navigation was sought from many different types of living systems including insects, snakes and humans. Although this line of investigation led to many different, non-traditional approaches to navigation, none stood up particularly well to even cursory analysis. For example, some snakes and worms have fluid-filled membranes, forming partitionable annuli that line the length of their bodies [41]. As the body bends, a change in volume for certain membranes takes place indicating the amount of bend. This approach could be adapted to the robotic drill by lining the interior wall of the umbilical with flexible, partitionable annulus sections. In its unpartitioned state, fluid would be allowed to flow freely between the sections. When a measurement is required, the sections could be closed off and a volume change assessment could be made for each partition to determine the drill's angular offset. It is thought that this approach could potentially add considerable unwanted weight to the system, lead to complications concerning temperature variations down the length of the umbilical, and could not easily compensate for volume changes due to umbilical stretch. Another biomimetic approach was inspired by the concept of joint position sensors found in many living systems. It was thought that a network of small variable resistors could be joined together on the nodes of a mesh lining the interior wall of the umbilical (figure 11, next page). As the umbilical bends, the mesh would flex altering the network's resistance at given cross-sections down 33 the length of the umbilical. The implementation of this solution would likely become quite complex in three-dimensions, and depending on the required number of resistors, quite unwieldy in terms of how many individual channels would be required. In the same vein as the potentiometer network, an artificial skin could be created for the Portion of Umbilical Wall - No Bending Portion of Umbilical Wall Subjected to Compression Figure 11 - Potentiometer Network. Such a network could line the interior wall of the umbilical consisting of strain umbilical. When subjected to bending, the network would flex, altering its resistance. gauges. The basic concept involves a network of these sensors mounted to the interior wall of the umbilical. As the drill deviates from a straight-line path, the umbilical would bend producing tension and compression around the perimeter of the umbilical. The strain gauges could detect these stresses and translate them into bend angles yielding orientation and heading. One major obstacle to this approach is the unknown characteristics of the umbdical material. At present, it is not known what material(s) will be used for the umbilical or how flexible it will be. This complete lack of relevant information concerning the umbilical makes testing this approach somewhat futile. Moreover, it may be very difficult to isolate the tensions and compressions induced by bending from those induced by the intense vibrations of the drill. This rapid vibration-induced tension/compression cycle may also lead to difficulties in reliably mounting a delicate strain gauge network. 34 Many such biomimetic approaches were devised, but all were found to have significant drawbacks, even at the conceptual level. It may be that some of these approaches could be made to work quite effectively for facilitating navigation of the robotic drill, however the amount of work required to render any one of these concepts feasible would likely be substantial relative to the level of effort necessary to implement one of the other more well-established technologies. 3.2.4 Fluid Level Perhaps one of the oldest methods for detecting tilt with respect to gravity is the fluid level. A small vessel containing a fluid (e.g. oil) could be mounted inside the drill end effector, or several along the length of the umbilical. As the drill deviates from a given path, an air bubble in the fluid would move relative to the vessel. Vibration and shock would tend to make this type of measurement very difficult, if not impossible. Alternatively, there is a new class of gels being developed, which partially solidify when exposed to an electric charge [6]. The gel could be solidified while taking measurements then dissolved to allow the air bubble to move within the medium. If several measurements were taken in this manner, a great deal of noise could be eliminated from the signal by averaging the results. However, these materials are still somewhat experimental and are therefore not easily obtainable. In addition, it may be very difficult to manufacture a sensor capable of detecting minute deviations using this principle, rendering this approach perhaps less than ideal. 35 3.2.5 Laser Line-of-Sight Taking advantage of the fact that effector deviates from an intended path, the laser line-of-sight concept involves a laser source and a target spaced several meters apart within the umbilical. The target could come in the form of a Position Sensitive Detector (PSD), a device that outputs the position of a beam of light striking its surface. Referring to figure 12, a small laser source would be mounted on the inside wall of the umbilical. Directly opposite from the source and also mounted on the umbilical's interior wall would be a PSD sensitive to the appropriate frequency of laser light. the umbilical would likely bend when the end Figure 12 - Laser Line-of-Sight. As seen i n a cutaway side view, wi th no bending, the laser light w o u l d strike a reference location on the target. When subjected to bending, the laser light w o u l d strike a different part of the target, indicating the magnitude of bend. With no end effector deviation, the laser would strike the PSD at a reference x-y location - a point that all other locations on the target would be compared to. In the presence of end effector deviation, the umbilical would bend and the laser would strike the PSD at a different location. This displacement (relative to the reference location) could be translated into a bend angle, and therefore drill heading and orientation could be obtained. 36 With the use of a two-dimensional PSD, both the x and y position of the light beam could be detected, allowing for complete characterization of the end effector's orientation with only one sensor setup. In addition, a series of such laser/ target pairs could be mounted along the length of the umbilical to obtain periodic snapshots of the umbilical's overall position, if desired. The farther apart the laser source is positioned from the sensor the better, from the perspective of maximizing horizontal deflections. It is desirable to have larger horizontal deflections since the PSD (or whatever target may be employed) will have limitations concerning its ability to resolve closely spaced locations on its surface. For example, the most accurate Hamamatsu two-dimensional PSDs feature position detection errors ranging from 40-150uJm (0.0016-0.0059in) [42]. If the laser source and target were placed 5m (16ft) apart, the resultant horizontal displacement would be 187um (0.0074in) at a bend angle of 0.172deg. These figures suggest that the laser source and target should be placed farther apart to produce increased displacement. Greater distances between the laser source and target are desirable, however the farther apart these two components become, the grater will be the amount of vibration-induced signal noise. Mounting the laser source and PSD to the interior of the umbilical wall could prove problematic, especially in light of the expected vibrations. Essentially, the laser source and PSD mount would act like cantilever beams resulting in significant unwanted displacements being detected on the PSD. These vibration-induced displacements could easily obscure the signal produced by legitimate deviations, especially when the source and target are placed at large distances from each other. In addition, umbilical sagging could contribute to false readings - depending on the rigidity of the umbilical material, the system may sag under its own weight in the borehole and produce umbilical bends even if no deviations from the 37 intended path have been incurred. Again, more information about the drill design must be known before this approach could be investigated further. 3.2.6 Non-Contact Fiber Optic Probe Non-contact fiber optic probes are used to measure linear distances over very small scales and consist of two sets of optical fiber bundled together (light transmitters and light receivers). When a probe face is in contact with the surface it is measuring the distance to, most of the light emitted by the transmitting fibers is reflected back into those fibers and relatively little light is reflected into the receiving fibers. As the probe moves away from the target surface, the amount of light reflected to the receiving fibers increases until the optical peak is reached - the point at which the amount of light reflected to the receiving fibers begins to diminish with increased distance. The amount of light detected by the receiving fibers is translated into an output voltage, which can be calibrated to represent the distance between the probe and the target. Non-contact probes using this principle are used in motor vibration analysis, turbine blade tip clearance measurement, and surface defect analysis among many other applications [43]. One approach to taking advantage of this type of sensor involves mounting the probe to a rigid pole connected to the interface between the end effector and the umbilical, as in figure 13 (next page). As the drill deviates from the intended path and as the umbilical bends, the pole would remain parallel with the end effector. The probe could measure this change in linear distance between the pole and the umbilical wall. Some fiber optic non-contact probes are capable of resolving distances down to O.OljAm [43], sufficient to detect the umbilical bend produced by a 0.15%-of-depth borehole deviation if the probe were placed on a rigid pole at least 1.15m from the end effector. If the pole were made longer, a sensor featuring less resolution could be employed. 38 T h e r e are t w o s i g n i f i c a n t drawbacks to this approach. The f i r s t i n v o l v e s the fact tha t v ib ra t ions i n d u c e d by the d r i l l w o u l d ve ry l i k e l y lead to a ve ry no i sy s igna l since the po le has s ignif icant length and w o u l d be connec ted to the e n d effector. The second has to do w i t h probe size - probe t ip lengths are often o n the o rder of severa l inches [43], w h i c h is m u c h too large for the space ava i l ab le w i t h i n the u m b i l i c a l . In a d d i t i o n , cab le lengths f rom the p robe to the s igna l process ing e q u i p m e n t are restr icted to re la t ive ly short lengths, o n the o rder of 140cm (54in) [43], to ensure that not too m u c h reflected l i gh t is d i s s ipa ted f rom the r ece iv ing fibers, e n s u r i n g that the s igna l is of adequate s trength for re l iable measurements . Th i s imp l i e s that b u l k y s igna l process ing equ ipment w o u l d have to be bu i l t into the d r i l l i n or ve ry near the end effector, as opposed to h a v i n g this equipment located on the surface where v o l u m e is not as restrictive an issue. 3.2.7 Capac i t ive Sens ing 7 W h e n a set of conduc t i ng plates of a g i v e n area is subjected to a vol tage, the sys tem w i l l exh ib i t a cer ta in capacitance that is i nve r se ly p ropo r t i ona l to the 7 Dr. J. Yan originally suggested this concept as one possible approach to developing a position sensing system. Figure 13 - Non-contact Fiber Opt ic Probe. Attached to the end effector / umbil ical interface, the probe could detect l inear displacements induced by bending of the umbi l ica l by optical means. distance be tween the plates. In an approach ve ry s imi l a r to that d iscussed for non-contact fiber optic probes, capaci t ive sensors c o u l d be p u t to use i n the robotic d r i l l i n g system for de te rmin ing end effector or ientat ion a n d heading. A s the u m b i l i c a l deviates f rom a straight-line pa th a n d the bend angle increases, the distance between the capacitor plates w o u l d increase, i nd ica t ing the amoun t of u m b i l i c a l bend, p r o v i d i n g an orientation. Capac i t ive sensors can be found i n form factors appropr ia te for i nc lu s ion i n the d r i l l system, h o w e v e r p rob lems of d r i l l v i b r a t i o n - i n d u c e d s igna l noise w o u l d s t i l l r emain . In add i t ion , capaci t ive sensors can be ve ry sensit ive to changes i n h u m i d i t y a n d e lec t romagnet ic a n d r ad io interference, further l i m i t i n g their usefulness i n this par t icu lar app l i ca t ion [44]. F ina l l y , readings also depend o n plate a l ignment - were there any angular offset between the plates, this w o u l d constitute a significant source of s ignal error [44]. 3.2.8 O p t i c a l Shape Sensing O n e of the m o r e i n t r i g u i n g n o n - t r a d i t i o n a l approaches to fac i l i t a t ing d r i l l n av iga t ion invo lves the use of a n e w l y emerg ing technology that incorporates f iber opt ic sensors. "Shape Tape" , manufac tu red by M e a s u r a n d [45], is an example of such a sensor. Essent ial ly , the flexible ' tape' employs a series of fiber optic sensors (i.e. opt ica l diffraction gratings) d o w n its l ength to measure shape, p o s i t i o n a n d o r i en t a t i on i n t h r ee -d imens ions . "Shape T a p e " is c u r r e n t l y gene ra t ing s ign i f i can t in teres t i n 3 D m o t i o n - c a p t u r e c o m p u t e r g r aph i c s applicat ions. W i t h regard to de t e rmin ing the robotic d r i l l ' s or ienta t ion, i t is thought that a series of Shape Tapes or s i m i l a r sensors c o u l d l ine the in te r io r w a l l of the u m b i l i c a l to detect the e n d effector's angu la r offsets t h r o u g h o u t the d r i l l ' s opera t ion. Unfor tuna te ly , the capabi l i t ies of this technology are not yet to the 40 point where they would be of tremendous benefit to the robotic drill. The "Shape Tape" features an endpoint accuracy of 1-3% of its length, and an endpoint resolution of 0.5deg. While this level of performance is not currently in tune with the robotic drill's requirements, in time this technology may prove to be a viable option. 3.3 Concept Summary In total, twelve broad approaches to the problem of determining position and orientation for the robotic drill were presented (see table 2 at the end of this section for a complete summary). Many of these approaches comprised several sub-configurations, each featuring varying degrees of feasibility. By far the most easily implementable method for accurately determining the distance traveled by the end effector is achieved by measuring the length of drill umbilical deployed from the aboveground spool. This leaves the issue of orientation to be addressed so that position can be determined. Gyroscopes, accelerometers, clinometers, lasers, and optical shape sensing were the most appropriate options presented based on the geometry of the problem. However, once the real-world constraints of continuous and intense vibration coupled with strict volume allowances were applied, no single technology stood out as an obvious choice for this application. Optical sensing can be eliminated from the list of potential candidates since the technology is not yet mature enough to detect the slight deviations that the drill is likely to experience. If its accuracy could be improved upon, this technology would be highly desirable since it outputs a real-time three-dimensional shape of the object being monitored in an intuitive graphical format. For future versions of the robotic drill, optical shape sensing technology should definitely be investigated further. 41 While the laser line-of-sight approach may be capable of producing desirable results in a controlled lab setting, this method is not inherently tolerant to the intense vibrations found in the real-world application. If the amplitude and frequency of the vibrations were relatively constant throughout the drilling process, perhaps these ill effects could be averaged out. It may be unwise to pursue a potential solution based on that assumption however. More details must also be known about the environment inside the drill umbilical - any dust trapped in the gap between the umbilical wall and the pneumatic/hydraulic service conduits may interfere with the light beam and the PSD's ability to accurately detect it. In addition, it is not known how flexible the umbilical material will be, which has direct, and possibly deleterious implications for this approach. In total, there are too many unknowns concerning the drill design to confidently pursue this method. Many clinometer designs feature capabilities well in excess of what would be required for the robotic drill, however high levels of performance are also accompanied by unwanted physical bulk. Smaller sensors, while not quite as capable, suffer from relative intolerance to vibration. This technology should be kept in mind for future iterations of the robotic drill since it is likely to benefit from continued miniaturization, however at present this technology is not suitable. Due to their versatility, accelerometers offer several interesting perspectives on monitoring drill orientation, however, as with most other types of sensors, small form factors typically harbour lower performance. By virtue of their inclusion in automobile air-bag systems, consumer video recorders, computer hard drive protection systems and many more applications, small, low-cost accelerometers are among the most widely used sensors in the world. As such, they're undergoing tremendous development and within a very short period of time may provide a clearly superior approach for tracking the end effector's 42 movements [46], [47], [48]. At this time however, it seems as though this technology is perched on the cusp between superiority and relative inadequacy and will therefore not be pursued presently. Gyroscopes also offer a great deal of versatility and may be employed in two primary configurations: angular rate tracking from either a 'global' or 'local' perspective. In the global perspective, the drill's angular rate of speed is tracked while it traverses a large circle defining the deviation path. These angular rates are typically very low due to the slow progression of the drill though the rock, and the immense radius of curvature of the deviation. The use of high-end fiber optic or ring laser gyros would be required to detect these faint angular rates, however they are not appropriate for this application primarily due to their physical size, prohibitive cost and the stringent trade restrictions governing their use. Alternatively, the local perspective for angular rate tracking may hold significant promise. As opposed to measuring very small angles of end effector tilt directly -as would be the case for clinometers and accelerometers - low-cost MEMS gyros could be used to detect the rate of end effector tilt. This means that even if the tilt angles to be detected are very small, as long as they are induced relatively quickly, consumer-grade sensors capable of reasonable performance could be used to detect them. Although potential drawbacks could include achieving the required level of performance in a vibration-rich environment, it is believed that this approach holds the most promise of all the options considered at the present time. The remainder of this thesis will detail further conceptualization and testing of a PSS configuration employing MEMS gyroscopes using the local tilt rate approach to measurement. It should be noted that none of the concepts presented in this section would be particularly adept at detecting lateral end effector translations in the borehole. 43 However, it is thought that such translations, occurring without any rotation, would be highly unlikely. Until further end effector prototype testing can be carried out, the drill's exact behaviour in its operational setting will not be known with any level of certainty. However, it is likely safe to assume that the end effector will experience tilt, particularly due to the motions of the rear feet and their contact with the borehole wall. Therefore, the approaches investigated over the course of this research have ignored the possibility of end effector translation and have instead focused on detecting changes in the end effector's orientation. 44 Concept Comments Clinometers Generally too large for the application; smaller sensors typically feature poor vibration tolerance. Magnetometers Magnetic fields generated by the local rock being drilled through could obfuscate the Earth's magnetic field, rendering these sensors ineffective. Accelerometers While not required for detecting the underground distance traveled by the drill, these sensors could be used to detect end effector tilt. However, at present, accelerometers that conform to the volume constraints are not sensitive enough for this application. Gyroscopes MEMS gyroscopes could be used to detect the rate of end effector tilt induced by the drill's rear feet. This ability is an advantage over other methods since the tilt rate is a controlled parameter. This approach to PSS design will be investigated further. External Cameras This technique would be rendered ineffective due to the cuttings being blown out of the borehole, which would obscure the borehole wall from view. External Rollers The uneven nature of the borehole wall combined with the likelihood of roller slippage would render this approach problematic at best. Biomimetic Solutions Several configurations were investigated, however not enough is currently known about the drill/umbilical design to confidendy undertake any particular approach. Fluid Level Vibration and shock would render this method prohibitively difficult to implement. Laser Line-of-Sight Vibration would induce a significant amount of signal noise, and false readings may occur due to umbilical sagging or other flexures not related to end effector orientation. Non-Contact Fiber Optic Probe The equipment required for this approach is much too large for the available volume. Capacitive Sensing This method is particularly sensitive to changes in temperature and humidity, and even slight plate misalignments. Optical Shape Sensing While perhaps one of the most interesting approaches to PSS design for future iterations of the robotic drill, this technology is not yet mature enough to produce the required level of accuracy. Table 2 - Concept Summaries 45 4.0 MEMS Gyro-Based Position Sensing System This section presents a more detailed overview of how the gyroscope-based PSS could be configured and how it would function. As outlined in the previous section, several gyros would be used in conjunction with each other to measure the rate of end effector tilt. This data would be combined with information concerning the underground distance traveled by the drill, as measured by umbilical deployment from the aboveground spool, to determine both the position and heading of the end effector. In the event that a deviation is detected, a feedback control system would be provided with this information to determine the magnitude of corrective action required to alter the drill's heading back toward the intended path. As currently envisioned, the gyros would be mounted near the interface between the end effector and the umbilical, inside the gap between the umbilical's interior wall and the pneumatic/hydraulic service conduits. To fully characterize the drill's orientation (pitch and yaw, looking down the length of the end effector toward the rock face being drilled), three gyros would be required with the possible inclusion of a fourth to track roll if desired. To determine specific sensor configuration geometry, it may prove insightful to examine how and when the end effector tilting could be expected to occur. There are two major factors that lead to borehole deviation in traditional long-hole drilling operations: drill setup inaccuracies and the geology encountered while drilling [2]. However, a third deviation-inducing factor exists that is unique to the robotic drill - the end effector feet. Throughout the various phases of the end effector's operation, the front and rear feet repeatedly come into contact with and retract from the borehole wall. Assuming that the wall's surface will seldom be smooth and uniform, it is unlikely that these feet will all come into contact with the wall simultaneously. If there is even a slight discrepancy 46 between w h e n the va r ious feet touch or let go of the w a l l , a ro ta t ion about the d r i l l head w i l l be p roduced . In add i t i on to the feet act ing as a potent ial source of dev ia t ion , they are also the mechan i sm t h r o u g h w h i c h d e v i a t i o n s c a n be corrected. Since tilt (both unwan ted and corrective) seems to depend so heav i ly on the actions of the end effector feet, i t m a y make sense for the PSS to inc lude three gyros , one or iented to detect the angular rates p r o d u c e d by each of the three feet, as depic ted i n figure 14. Since the end effector w i l l be stationary throughout most of the d r i l l i n g operat ion (supported by the rear feet), i t m a y not be necessary to moni tor the e n d effector's tilt con t inuous ly throughout d r i l l i n g . In other words , the gyros w o u l d on ly need to be ' o n ' and t ak ing measurements at those times w h e n t i l t ing is most l i k e l y to occur - w h e n the rear feet engage or disengage the borehole w a l l . W h e n angular rate measu remen t s are t aken , e n d effector t i l t w o u l d be d e t e r m i n e d by compar i ng the current set of values against those recorded d u r i n g the p rev ious r o u n d of measurements . The chief advantage of this approach is that errors i n the s ignal w o u l d not accumulate as qu i ck ly , since the gyros w o u l d n ' t be t ak ing measurements d u r i n g those times w h e n the p robab i l i ty of exper ienc ing a tilt is ve ry l o w , but v ibra t ions are on-going. The assumpt ion that t i l t ing w i l l not occur w h i l e the rear feet are d e p l o y e d is cons ide red v a l i d s ince i f there is any appreciable s l ippage of the feet w i t h respect to the borehole w a l l , the down-force requi red for d r i l l i n g c o u l d not be achieved and d r i l l i n g w o u l d cease. Successful operat ion of the d r i l l depends o n the fact that the end effector's rear feet h o l d the d r i l l f i rmly i n posi t ion. 47 J i 1 • b • ] C / \ \ >v_ Umbilical / End Effector Interface - Top V i e w Figure 14 - Gyro Mount ing . Each gyro is oriented such that its sensing axis is perpendicular to the direction of motion produced by its corresponding end effector foot. 4.1 Gyro-based PSS Method of Operation F r o m the perspective of the PSS, the general descr ip t ion of the d r i l l i n g procedure out l ined i n Section 1.3 can n o w be adapted to inc lude the measurement and data a c q u i s i t i o n p rocesses . H o w e v e r , i t s h o u l d be n o t e d tha t before the commencemen t of d r i l l i n g , a d d i t i o n a l procedures s h o u l d be i m p l e m e n t e d to ensure that the end effector is a l i g n e d w i t h the in t ended pa th as c lose ly as possible, perhaps u s ing h igh-qua l i ty cl inometers . A s s u m i n g a l inear path, even an in i t i a l offset of O. ldeg can lead to a 35cm (14in) dev ia t ion at a dep th of 200m (656ft). Referr ing to figure 15 (next page; a copy of f igure 4 i nc luded here for clarity), it is assumed that d r i l l i n g begins w i t h the end effector rear feet h e l d f i r m l y against the borehole w a l l to p r o v i d e stabil i ty (figure 15A) w h i l e the midsec t ion s l o w l y expands to ma in ta in the requi red down-force (figure 15B). If s l ippage of the rear feet t ru ly were negl ig ib le (to be de te rmined w i t h add i t i ona l p ro to type testing), tilt rate measurements w o u l d not have to be recorded d u r i n g this t ime. Once the midsec t ion is fu l ly extended, the front feet extend to s tabi l ize the d r i l l (figure 15C) so that the rear feet can disengage f rom the borehole w a l l (figure 15D) . A t this po in t , the e n d effector m i d s e c t i o n can retract, p u l l i n g m o r e u m b i l i c a l off the aboveground spoo l and d o w n the hole (figure 15E). The length of u m b i l i c a l p u l l e d off the spool indicates the u n d e r g r o u n d distance traversed by the end effector. D u r i n g this midsec t ion retraction process and before the rear feet re-engage the borehole w a l l (figure 15D & E), s ignificant end effector t i l t ing c o u l d occur due to the considerable amount of we igh t bearing d o w n o n the d r i l l head, w h i c h is on ly par t ia l ly suppor ted by the front s tab i l iz ing feet. Before re -dep loy ing the rear feet, the gyros w o u l d be ' sw i t ched o n ' to detect the rotations i n d u c e d by each of the 48 il Iii - P u Figure 15 - Robotic Dr i l l ing Sequence, Redux. three feet as they come into contact w i t h the borehole w a l l (figure 15F). The gyros c o u l d cont inue to take measurements u n t i l the front feet retract (figure 15A) to ensure that any res idual t i l t ing is detected. The ti l t rates measured by the three gyros w o u l d be combined to p roduce a net end effector tilt, ind ica t ing the amount of p i t ch and y a w experienced by the back end of the end effector. Since these measurements w o u l d be taken re la t ive to w h e n the sensors were ' t u r n e d on ' , p i t c h a n d y a w data w o u l d have to be c o m p a r e d to p rev ious rounds of t i l t ing to determine a head ing relat ive to the in tended orientat ion. W h e n this in format ion is a d d e d to the length of u m b i l i c a l dep loyed f rom the aboveground spool , both the pos i t ion and head ing of the end effector w o u l d be k n o w n . If head ing errors were to accumulate beyond the po in t of manageabi l i ty , this approach to detecting tilt c o u l d also be used to recalibrate the PSS. B y retracting one end effector rear foot and extending the other two, the d r i l l c o u l d p u s h itself u p against the borehole w a l l to measure the amount of gap between the d r i l l and w a l l . If this process were repeated for each of the three feet - es tabl ishing h o w 49 much gap exists between the drill and borehole wall on all sides - the drill's angle of tilt could be determined irrespective of previous measurements. Of course, a system would be required to determine when the two feet should stop pushing, however if the angular rate as measured by the gyros came to zero, this would indicate that the drill was in all likelihood being pushed up against the borehole wall. In theory, this process could be carried out without having to stop the drill, maintaining the continuous-operation philosophy of the robotic drill concept. 4.2 Test Apparatus Conception and Development Of the three main thrusts of this research, the first - detailing a reasonable position sensing solution - is now complete. However, it has yet to be demonstrated that this approach is technically feasible. Since the actual prototype end effector was not available for testing PSS arrangements, a test apparatus capable of exhibiting analogous behaviour had to be designed and constructed. Perhaps the most important factor to keep in mind concerning the design of the test apparatus is the objective for building the system in the first place. The primary function of the test apparatus is to validate the 'tilt rate' measurement approach to PSS design, and as such there are at least two primary systems that must be accurately represented: a center of rotation to simulate the drill head, and a linear actuator to imitate the motion of the rear feet. As an example of how other environmental factors could be included in the testing process, it was decided to incorporate a proof-of-concept vibration source in the design. Of course, in addition to vibration there are many other environmental factors that could be taken into account such as temperature variations and electromagnetic interference. However, at this stage it was decided that these issues were considerably less important for validating the overall gyro-based PSS approach. 50 P r o v i d e d the gyro-based des ign su rv ives the i n i t i a l e v a l u a t i o n stage, these functionalities can be added at a later date. A s w e l l as being able to p roduce the above-ment ioned mot ion , there are certain k e y d imens ions that mus t be m i r r o r e d i n the test apparatus . The rad ius of r o t a t i o n (i.e. the d i s t ance b e t w e e n the g y r o m o u n t a n d d r i l l head) is a p p r o x i m a t e l y 152.4cm (5ft), a s s u m i n g that the gyros are m o u n t e d at the umbi l i ca l - end effector interface. In addi t ion , the l inear actuator mus t be capable of p r o d u c i n g d i sp lacements of u p to 2.54cm ( l i n ) , w h i c h r ep re sen t s the m a x i m u m p o s s i b l e gap b e t w e e n the u m b i l i c a l ex te r io r a n d the b o r e h o l e w a l l . A l l o f the pert inent d imens ions used i n the d e s i g n o f the tes t apparatus were gleaned f rom patents that cover the device and pr iva te c o m m u n i c a t i o n s [6], a n d unfor tuna te ly m o r e accurate d imens ions are not available. F i g u r e 16 dep ic t s the test a p p a r a t u s t h a t w a s c o n s t r u c t e d based o n the above-described criteria (refer Figure 16 - Test Apparatus Setup. A) end effector proxy B) universal joint C) linear actuator D) frame E) detachable gyro mount F) vibration control and data acquisition system to A p p e n d i x A for d i m e n s i o n e d schematics of the appara tus) . The sys tem consists of the d r i l l end effector p r o x y (figure 16, i t em A ) s u p p o r t e d by a un iversa l jo int (item B), w h i c h serves as the center of rotat ion. The end effector 51 proxy can be displaced by the linear actuator (item C), which in turn is supported by the frame (item D). The gyros can be fixed to a detachable mount (item E) that also houses a vibration-inducing mechanism. The data acquisition and analysis system is shown as item F. The linear actuator was constructed using a print head carriage assembly from an inkjet printer, complete with stepper motor (7.5deg/step), a drive belt and gears. This assembly was mounted to the test apparatus frame. As can be seen in figure 17, as the actuator moves left to right, the end effector proxy is tilted about the pivot. A slot in the pivot bar accommodates the small vertical displacements that occur as the proxy is titled. Figure 17 - Linear Actuator Assembly. As the actuator moves from left to right, the end effector proxy rotates about the pivot (red dot), and the universal joint below (not shown). Although a three-degree-of-freedom (3DOF) test apparatus would have yielded a system much truer to the actual behaviour of the drill end effector, a single-degree-of-freedom (1DOF) system is just as effective for evaluating the general feasibility of the gyro-based PSS. In addition, it is believed that this simpler 52 design is m u c h more versatile i n terms of absorbing manufac tur ing inaccuracies. For instance, w i t h three l inear actuators any mi sa l i gnmen t be tween the paths w o u l d manifest itself as an output error. W i t h a s ingle l inear actuator, the on ly significant requirement is that the actuator itself is not bent appreciably. In fact, great care was taken to ensure that desp i te the m a n y po ten t i a l sources of er ror i n t roduced i n the test apparatus ' construction, that none of these factors w o u l d p l a y a l a rge ro le i n the s y s t e m ' s o v e r a l l p e r f o r m a n c e . A s a n example , f igure 18 shows a figure i g _ Universal Joint. This enables the end effector d e t a i l e d v i e w o f t h e proxy to rotate about two perpendicular axes. universa l jo int that supports the end effector p r o x y 8 . E m p l o y i n g an au tomot ive u n i v e r s a l j o in t as a s t and- in for the d r i l l head was d e e m e d s i m p l e r than des ign ing a n d m a c h i n i n g a g i m b a l f rom scratch. A l s o , the fact that this g i m b a l supports rotations about two perpendicu la r axes as opposed to just the one axis under inves t iga t ion means that the l inear actuator and g i m b a l do not have to be a l igned w i t h respect to each other. R e t u r n i n g the d i s c u s s i o n aga in to the top of the e n d effector p r o x y , the detachable v ib ra t ion hous ing supports the sensor m o u n t and v ib ra t ion assembly (figure 19, next page). The gyros can be arranged 120 degrees f rom each other, as they w o u l d be on the actual d r i l l end effector. E v e n though the test apparatus is on ly capable of i n d u c i n g rotations about a s ingle axis, the sensor m o u n t can be 8 This particular automotive universal joint was taken from a 1972 M G B . To ensure min imal friction, the joint was cleaned and lubricated. 53 t u r n e d m a n u a l l y 360 degrees to s imula te end effector rotat ions about t w o p e r p e n d i c u l a r axes (figure 20). T h e sensor m o u n t is a c t u a l l y a laser l e v e l t r i p o d p la t fo rm, a n d as such features a n u m b e r of desi rable capabi l i t ies i n c l u d i n g l e v e l i n g Figure 19 - Vibrat ion Housing. This detachable component houses the vibration assembly and sensor mount. screws, and a graduated s w i v e l mount . The s w i v e l is accurate to 0.5 degrees and can be l ocked i n place. Th i s feature is useful for accurately k e e p i n g track of different sensor or ienta t ions that m a y be used d u r i n g test ing. The l e v e l i n g screws and o i l bubble ind ica tor help ensure that the sensor m o u n t can be kept reasonably leve l at an in i t i a l posi t ion, negat ing any potent ial sources of error due to manufac tur ing inaccuracies i n the rest of the setup. N o t e that since the gyros *t\ 1 m Linear Displacement 1 Gyro 1DOF Testing 1 1 Top-Down View Simulated 3DOF Testing Figure 20 - Simulating 3DOF Behaviour. The illustration on the left depicts the init ial setup. The illustration on the right demonstrates that by rotating the sensor mount, contributions from each of the end effector rear feet can be simulated. 54 detect t i l t rates that are not dependant on m o n i t o r i n g external references, the inc l ina t ion of the sensor m o u n t w i l l not affect i n i t i a l readings as i t w o u l d an externally referenced tilt magnitude-based sensor system. T h e sensor m o u n t is f i x e d to a v i b r a t i o n pla te , w h i c h i n t u r n is moun ted to the v ib ra t ion hous ing by s p r i n g s ( f igure 21). T h e sp r ings , loca ted above the v i b r a t i o n plate , p u s h the p la te d o w n agains t t w o a l u m i n u m s t o p p e r s . A s t r i k e r connected to a rota t ing shaft strikes the bot tom of the v ib ra t ion plate once per r e v o l u t i o n , p u s h i n g u p against the spr ings , a n d then a l l o w i n g the plate to re turn to its i n i t i a l pos i t ion , impac t i ng against the stoppers. This m o t i o n is i n t ended to s imula te the vibra t ions i nduced by the percussive d r i l l h e a d . S i n c e the v i b r a t i o n c h a r a c t e r i s t i c s of the p r o t o t y p e r o b o t i c d r i l l h a v e n o t b e e n es t ab l i shed , this test sy s t em w a s cons t ruc t ed as a p roof -of -concep t Figure 21 - Vibra t ion Assembly. Top: The striker rotates on the motor axel, coming into contact w i t h the vibrat ion plate once per revolut ion, pushing it up against springs. Shock is induced when the vibrat ion plate snaps back into contact wi th the two stoppers. Bottom: Springs hold the vibration plate down against the stoppers. and can suppor t a range of output . C h a n g i n g the length of the str iker alters the a m p l i t u d e of v i b r a t i o n w h i l e v a r y i n g the a m o u n t of vol tage s u p p l i e d to the d r i v i n g motor changes the frequency. F igure 22 (next page) depicts the va r ious elements of the test apparatus con t ro l system and data acquis i t ion hardware . V i b r a t i o n cont ro l is carr ied out by means 55 of a control panel (figure 22B), i n c l u d i n g torque control , a master o n / off swi tch , frequency cont ro l , and a mul t imete r . A l l systems are t ied into the v i b r a t i o n motor , w h i c h is m o u n t e d to the rear of the v ib ra t i on hous ing . The frequency control ler is s i m p l y a househo ld l igh t ing d i m m e r swi tch , and the mul t imete r is used to moni to r the A C voltage supp l i ed to the motor so that motor frequencies can be rel iably repeated from test to test. Figure 22 - Vibrat ion Control and Data Acquisi t ion Hardware. A) an overview of the setup. The D A Q system can be seen attached to the test apparatus frame, and the stepper controller is mounted to the left of the computer screen. B) Vibration motor control panel (torque control, on/off, frequency control), and multimeter for monitoring the voltage supplied to the vibration motor (not pictured). C) Stepper motor controller for alt control. D) National Instruments UBS-6009 D A Q system. Ti l t cont ro l is carr ied out v i a the stepper moto r interface (figure 22C) . This off-the-shelf 6-pin un i t is manufactured by K e m o G e r m a n y (mode l M109) [49], and offers a s imp le means of con t ro l l i ng the stepper moto r (and therefore the end effector p roxy tilt) th rough a P C para l le l port. U s i n g the accompany ing software, 5 6 short routines can be programmed that involve changes in stepper direction and speed. These programs can be run in a loop or as one-off events. The software also allows for manual control of the stepper where direction can be operated using the computer keyboard arrows. The data acquisition system is shown in figure 22D. A National Instruments USB-6009 [50] unit was employed, featuring 13-bit resolution (single-ended), a 42kHz maximum analog input sampling rate, eight analog input channels, 12 digital input/ output channels and a USB interface. This hardware was used in conjunction with LabVIEW for initial data analysis and recording. At the heart of the testing system lie the gyroscopes themselves. Citing considerably better performance specifications than 'hobby-grade' sensors, the CRS03 MEMS gyro produced by Silicon Sensing Systems is available at very reasonable per unit costs ($150USD) and in a suitably small form factor (figure 23, next page). The CRS03 makes use of a vibrating silicon ring structure, as opposed to forks or other shapes, to detect rotation and is therefore reputed to be much less susceptible to vibration. These units feature a scale factor of 20mV/deg/sec and have come to be relied upon in many applications, such as the Segway Human Transporter device. Operation of the CRS03 is very simple. Each unit has three pins that are connected to the data acquisition hardware: driving voltage (5V max), ground (0V) and an analog output that relays the angular rate as a voltage. The CRS03 has a typical bias of 50% of the driving voltage, meaning that when the sensor is at rest, it will output 2.5V. When the sensor is rotated, it will output a voltage according to: 57 v V d d n? o r V d d ^ Vo = + (Ra xSFx ) 2 5 [Eq. 4] where Vo is the output voltage, Vdd is the d r i v i n g voltage ( typical ly 5V) , Ra is the rate of rotation (deg/sec), and SF is the scale factor ( 2 0 m V / d e g / s e c ) . • .... • BP* Figure 23 - CRS03 gyro. Each gyro measures 29mm (1.14in) x 29mm (1.14in) x 18.4mm (0.72in) and weighs less than 18g (0.63oz). The CRS03 features 3 pins - driving voltage, ground, and analog output voltage. It shou ld be noted that success of the t i l t rate approach to PSS des ign does not depend on the CRS03 gyro specif ical ly . This un i t was chosen for its reasonable cost and performance, however other gyros exist, perhaps more expensive, that c o u l d y i e l d super ior results. F r o m the po in t of v i e w of d e v e l o p i n g a proof-of-concept system, the CRS03 was deemed adequate for the task. Moreover , it is not necessary that the test apparatus e m p l o y gyros as the sensing devices. Since the design ph i losophy beh ind the test apparatus was to m i m i c the d r i l l end effector's b e h a v i o u r , any t i l t - ba sed P S S a p p r o a c h , i n c l u d i n g those that feature accelerometers or cl inometers, c o u l d be tested. A g a i n re la t ing back to the three centra l themes of this research, the second objective - to deve lop a system for testing the pos i t ion sensing so lu t ion - has been a c c o m p l i s h e d . N o w that b o t h a reasonable P S S s o l u t i o n a n d test 58 methodology are in hand, determining the general feasibility of the proposed solution based on initial testing is all that remains. 59 5.0 Testing the Gyro-Based Position Sensing System To place the details concerning the gyro-based PSS and test apparatus design into proper perspective, the overall intent of testing these systems bears mentioning again. This research has yielded two proof-of-concept systems: one for solving the issue of drill end effector tracking, and the other for testing this and other tilt-based solutions. The specific hardware employed is not the message of this work, but rather the configurations and methodologies behind their use. Having said that, specific hardware had to be chosen and tested to illustrate these methodologies and to shed light on the first-order practical issues associated with their use. This section will first discuss the test apparatus and its performance, followed by initial results from testing the CRS03 gyro. A procedure for interpreting the data will be presented, providing a means to evaluate the soundness of the gyro-based PSS. 5.1 Test Apparatus Specifications and Performance As stated earlier, the dimensions of the test apparatus were matched to those of the actual robotic drill as closely as possible given the information available (Appendix A includes dimensioned schematics of the test apparatus). As such, measured from its mounted position, the gyro's sensing element is 158.7cm (62.5in) from the universal joint's center of rotation. Also, although the linear actuator is capable of producing an overall displacement of 35.2cm (13.9in), only 2.5cm (lin) is required to simulate the motion induced by the end effector's rear feet. In terms of angular displacement, at 1.25cm (0.5in) of linear displacement (the approximate gap between the end effector when vertical and the borehole wall), the geometry of the test system yields approximately 0.45 degrees of tilt. 60 The stepper motor controls the magnitude of, and rate at which, t i l t ing is induced. The step frequency can be set from lHz-2000Hz , w i th each step producing approximately 0.9mm (0.04in) of linear displacement (14 steps corresponds to 1.25cm (0.5in)). It should be noted that wi th the current setup, step sequences are not continuous. For example, when implementing a series of 10 steps at 1Hz, 10 discontinuous steps w i l l occur, one each second - this particular hardware is not capable of producing smooth displacements. Higher frequencies lend themselves to rough approximations of continuous behaviour, however these speeds are much greater than what the dr i l l is thought to be capable of producing. It is assumed that the end effector rear feet are move at a max imum of 1.25cm/sec (0.5in/sec), resulting in a max imum angular displacement of 0.45deg/sec, which works out to 14Hz for the test apparatus stepper motor. Few specifics are known about the robotic dr i l l , and even fewer concerning its vibration characteristics. A s such, the test apparatus' vibration mount was included in the design to demonstrate one approach for implementing this functionality. The prototype vibrat ion mount is capable of displacing the vibrat ion plate vertically by a maximum of approximately 0.5cm (0.2in), although wi th continued refinements it is expected that this distance could be increased if required. Currently, the mount does not support direct measurement of vibration frequencies. A multimeter is used to gauge how much A C voltage is applied to the vibration motor, however subjective observations indicate that frequencies range from less than 1Hz to approximately 10Hz. In future iterations of the test apparatus, an accelerometer could be used to monitor the output of the vibration assembly. There are a number of different approaches to implementing vibration, however more must be known about how the d r i l l vibrates before real progress can be made. For instance, the gyros could be set atop speakers that emit low-frequency ; 6 1 sound waves. Vibration would be transferred to the sensors, however the real-world setting could involve both vibration and shock - behaviour that speakers may not be capable of producing. Depending on the exact nature of the drill's vibration characteristics, producing analogous vibrations with the test apparatus may not be a trivial matter. From the point of view of implementing a PSS design, vibration testing is a significant topic requiring continued attention. 5.2 Initial CRS03 Test Results Although the test apparatus is capable of accommodating three gyros and can submit them to vibration and shock, the tests in this work focused on evaluating a single gyro that was not subject to vibration. If using gyros as discussed in previous sections is deemed generally suitable for this application, additional layers of complexity (more gyros, vibration, etc.) could be added to more accurately reflect the demands of the robotic drill in its operational environment. Initial testing consisted of rotating the gyros at approximately 0.45deg/sec for 1 second. This corresponds to a linear displacement of 1.25cm (0.5in), the size of the gap between the end effector when vertical and the borehole wall. The angular rate was achieved by programming the stepper motor to produce 14 steps per second, and trials were conducted with sampling rates of 1kHz, 2kHz, 2.8kHz, 5.6kHz and 11.2kHz. To reduce signal noise, samples were averaged in groups of 200, resulting in 'averaged' sampling rates of 5Hz (half the sensor bandwidth), 10Hz (the sensor bandwidth), 14Hz (one averaged sample per motor step), 28Hz (two averaged samples per motor step), and 56Hz (4 averaged samples per motor step). In other words, each data point in these averaged sets represents the average of 200 points sampled over that same period of time. Plots for each trial can be found in Appendix B, while Appendix C includes sample data recorded without any averaging, for comparison. 62 Figure 24 shows a plot where data is sampled at 2kHz (i.e. 10 averaged samples per second). The gyro is maintained at rest for approximately 10 seconds, and then rotated at a rate 0.45deg/sec for one second. The discontinuous nature of the stepper motor is clear in the output signal between 10 and 11 seconds on the plot, in addition to the effect of acceleration caused by the abrupt commencement and cessation of end effector proxy rotation. Gyro Output: Ra=Q.45deg/secf S - 2kHz 2.575 2.57 O) 2.565 I 2.56 2.555 2.55 10 16 Time (sec) Figure 24 - Sample Gyro Output. This trial featured an angular displacement of 0.45deg/sec, over a linear distance of 1.25cm (0.5in). Samples were recorded at 2kHz, and averaged in groups of 200. The rotation was induced approximately 10 seconds into the trial. The intent of each trial was to reliably rotate the end effector proxy by 0.45 degrees and to compare the gyro output against this actual displacement9. Due to slippage of the linear actuator belt on the stepper motor gear, the desired displacement was not always achieved. In all cases, the actual linear displacement was measured manually and the gyro output was compared to this 9 The actual linear displacement was measured on the test apparatus with a set of calipers. 63 measurement. All linear displacements for this first round of trials fell within the range between l.lcm-1.3cm (0.43in-0.51in). Once a trial was run and the actual displacement measured, the gyro output would be analyzed - a procedure that involves four main steps. First, the dataset is plotted and those portions pertaining to the rotation are extracted to form a separate dataset. Next, an average is calculated for the initial dataset (from t=0 seconds to where the rotation begins). This average is to provide a baseline reading that reflects the gyro's environment prior to experiencing the rotation. In the third step, this baseline average is plotted with the rotation data to establish a cutoff level that helps to identify the precise start and end points of the rotation data. Finally, the average of the rotation signal is compared to the baseline pre-rotation average. This offset, measured in millivolts, divided by the sensor's scale factor yields an angular rate, in degrees per second. This value is then multiplied by the amount of time the sensor was rotated for - calculated from the dataset -to yield the angular displacement. The angular displacement measured by the gyro is then compared to the calculated angular displacement (based on the actual movement of the test apparatus). The difference between these values is expressed as a percentage error. Figure 25 (next page) depicts the offset between the pre-rotation average and rotation signal average to help obviate this signal analysis process as described above. Since the intent of these tests is to provide an initial evaluation of the overall gyro-based PSS, the ability to support real-time signal processing wasn't perceived as critical at this stage. As previously mentioned, tests were run using a variety of sampling rates, and some yielded more favourable results than others. Table 3 (next page) includes the percentage errors for 5 tests run at 5 different sampling rates each. For each test, the intent was to produce 1.25cm (0.5in) of linear displacement in one second. 64 Gyro Output: Ra=0.45deg[/sec. S = 2kHz 2.58 n , , , , , , , , , , , , , , , , , , , , , 2.575 2.555 » t | I I I I | I I I I | I I I I | I I I I | I I I I | I I I I | I I M | I I I M 0 2 4 6 8 10 12 14 16 Time (sec) Figure 25 - Data Analysis Methodology. The pre-rotation average is compared to the average obtained over the course of the rotation. This offset indicates the angular rate measured by the gyro. Multiplying this rate by the amount of time the rate was sustained for yields the angular displacement. Stepper Rate = 14Hz Sampling Rale (Hz) % Error Average StDev 4.56% 4.78% 1000 7.20% 6.26% 1.78% 5.91% 8.84% 0.75% 5.08% 2000 3.22% 1.90% 2.17% . 0.12% 0.34% 1.83% 1.28% 2800 6.88% 5.48% 4.40% 5.27% 12.16% 9.60% 6.03% 5600 2.12% 6.63% 2.83% 7.42% 8.00% 4.74% 3.64% 11200 8.32% 6.54% 2.58% 9.94% "6.05% ~~ Table 3 - Percentage Errors for Trial 1 65 Tests sampled at 2kHz (and averaged in groups of 200 samples) produced significantly less error than trials sampled at other rates, with the best trial resulting in a percentage error of just 0.12%. The average percentage error for the 2kHz trials is 1.90%, corresponding to an average angular difference of just 0.0082deg (table 4). This result is encouraging since the maximum tilt to be expected from a 0.15%-of-depth deviation at 200m (656ft) is 0.172deg. Put another way, the observed error is 20 times smaller than this maximum allowable tilt. Stepper Rate = 14Hz Sampling Rate (Hz) Difference (deg) Average SlDev 0.01576 0.02031 2000 0.00339 0.00822 0.00916 0.00092 0.00073 Table 4 - Offsets from Actual Angular Displacements for Trial 1 In the real-world setting, it will often be the case where linear displacements much smaller than 1.25cm (0.5in) are produced. In fact, the maximum allowable tilt of 0.172deg corresponds to a linear displacement of just 0.5cm (0.2in), according to the drill and test apparatus dimensions. As such, a second round of tests were run at the same angular rate (14Hz), but for half the previous displacement [0.6cm (0.25in), or 7 motor steps]. The results are summarized in table 5 (next page). The output for this second round of tests is considerably less accurate than for the first round, with an average percentage error of 11.22%. Still, this value represents an offset of just 0.0315deg from the actual orientation, an error 5.5 times smaller than the maximum allowable tilt. 66 Stepper Rale - 14Hz, Disl = 7 steps Sampling Rate (Hz) % Error Average SlDev 2000 2034% 11.12% 13.37% 0.95 10.31% 11.22% 6.97% Difference (cleg) Average SlDev 0.05569 0.03053 0.00685 0.02850 0.0315 0.0175 Table 5 - Trial 2 Summary It may also be that the end effector feet move at a different rate than the one assumed thus far. Although it is unlikely that the feet will move more quickly than 1.25cm/sec (0.5in/sec), should they move more slowly, an increase in output error would likely result; this is due to the fact that smaller signals would be more prone to obfuscation from background noise. Therefore the third trial employed a reduced angular rate of half the previous speed, equaling 0.6cm/sec (0.25in/sec), (i.e. a stepper motor frequency of 7Hz). Results of the third trial are summarized in table 6. Stepper Rate = 7Hz Sampling Rale (Hz) % Error Average SlDev 530% 4.13% 3.53 'V 2.92% 2.27% 3.43% 0.70% 2000 Difference (cleg) Average SlDev 0.02611 0.01859 0.00238 0.0131 0.0102 0.01545 0.00316 Table 6 - Trial 3 Summary 67 As expected, performance is less than what was achieved in the first round of tests, although not significantly so. This third trial features an average percentage error of 2.92%, or 0.0131deg of offset from the actual orientation - 13 times smaller than the maximum allowable tilt. 5.3 Drift Test One issue that may be of concern with prolonged use of the CRS03 is signal drift. Over time, many sensors exhibit the tendency to output signals that veer away from an expected value. This drift can be caused by a variety of factors inherent to the sensor design, including limitations intrinsic to certain manufacturing techniques or materials properties, and may alter measurements appreciably if not given proper consideration. The drift tests also answer the question of what would happen if the gyros were simply 'left on' throughout the entire drilling operation (neglecting vibration and other sources of error). Two trials were run to determine the general drift characteristics of the CRS03 gyro. The first test ran 12 hours, keeping the gyro reasonably stable and motionless while sampling data at 2kHz and averaging data points in groups of 200. Figure 26 (next page) shows the results of this test10. This plot can be divided into three general regions based on the magnitude of trend line slope. The drift tends to rise relatively sharply during the first 60 minutes of the test, most likely due to initial warming up of the electronics, followed by an 8-hour period of diminished but appreciable increases. Finally, drift reaches a plateau at the 9-hour mark that is maintained through to test completion. Overall, this drift corresponds to an increase of approximately 2mV, 1 0 It is thought that the abundance of noise in the drift plots are due to samples being recorded at too high a rate. 68 w h i c h can be interpreted as an error of O. ldeg/sec . The drift error incur red over the first hour of the test is rough ly 0.05deg/sec. CRSQ3 Drift. S=2kHz 2.564 2.562 2.552 2.55 I—>—I—Li—I—I—I— I | I I I I I I I I I | I II I | I I I | I I II I | I I I I | 0 5000 10000 15000 20000 25000 30000 35000 40000 Time (sec) Figure 26 - CRS03 Drift Test. This trial was run for 12 hours, keeping the gyro as stable and motionless as possible. Three distinct regions of drift trend line slope can be identified, however the most radical shift occurs over the first hour of the trial. O n e of the se l l ing points of the C R S 0 3 is that it features repeatable drift. To pu t this c l a im to the test, a second test was carr ied out to verify the gyro behav iour d u r i n g the first two hours of operat ion, where the b u l k of the drift took place i n the first test. The results of this second test are s h o w n i n figure 27 (next page). A g a i n , a re la t ively sharp rise i n drift is seen d u r i n g the first 40 minutes of the test, fo l lowed by a pe r iod of less drastic increases. It shou ld be kept i n m i n d that a l though it appears as though the magni tude of this drift is significant, the t rend l ine slopes are i n fact ve ry smal l . For example , d u r i n g the first 60 minutes , the steepest drif t segment, both tests y i e l d e d trend 69 lines with slopes of 4x10^ mV/sec. In addition, since the gyro-based PSS design calls for the sensors to be used continuously only over very short periods of time (i.e. a minute or less), drift should not be a significant issue in processing the gyro's output. CRS03 Drift. 2hrs. S=2kHz 2.563 2.562 2 561 5? 1—. I 2,56 2.559 i 1 | III J 1 ; 1 i , I  H : , | 1 1 i 1, }l 1 j | il i l J il | 1 ij i i L, .1 i! * J Ui 1 I ]i Li mm mm MM • ( • • m i ' i m i i i r i i • i p w i i i i n ' i i ' i i i i i i i II TI •IIIIIIIIV 1 1 1 1 II'! 1 w • » i m i i m r » i i i u m i n m n m n u i i i n p u w n i i vw n i v i ' i n IHI IIr ir PH ' ' ' r r r ' I I " IIIIIIH i n ni'" ']•» i i i i n n i i i i ' i i i ' i r i i i u i i i iiiwni1 f • m i l " i " i • » • PIP r i i ' i i ! n n m i M n i n i n i HI' 1 1 ' . i i " 'ii 11 I •in '"ITH1 ' | l l l l l ! V Villi r I  | 2.557 2.556 2.555 m i i n i im n I 'i i i m n i i " • r -r . 1 , 1 , 1 , . , , , . .mm.. ,i. i ,i, i .uiiiji,. iii ,,..iu i . i i i i . i . ii 11, . i i i II,1 i i , i i i u i ,ii. iiIIII,iiiIUIII.JL i , iiiiiiii,iiiiiiuiiiii.iiiiiii,iiy.jiiiiiiiiu i i i.i in,. . . i , . i iiiiiii,iiiiiii.iiwiii..iii.iiiii.,iiLjiiiiiiiiiiaiiiiiMiiiiiiiui i i y i u n i i j i i g a i i i n i i i i i i i i i b i y i i i i t i i i i i I. J iml lllllllilllMIIIIIIHII l l . j l . l l l l l l l Itl.llil JlllllllHIWUIIIIlliai.HI.MIIIIUIIIllBIUIIUUilllJIIU iMiniiui Lii,iii,Hi»iiiyi,mAiiiMyiaiuw i B t i J i i i B i » H f t « i i i , M i a i i H t B a i i M H i a H i i i i H i H I I 1VIUM IUHJll l l lH mtm I M I H • l U ' H M H M U a M H B H M I B M H H m i H H I H 111 I i . i m\mi i i w H i i a a H H ^ • ^ • H . H mm H • • • • • • • • • • • • • I IIJ„iillllilMIIIIMMMMM 1000 2000 3000 4000 Time [sec] 5000 6000 7000 Figure 27 - Second CRS03 Drift Test. This trial was run for 2 hours, and the magnitude of signal drift over this period matches that from the first test. 5.4 Interpretation of Results No quantitative conclusions should be drawn from these results for speculating on expected PSS performance in the real-world setting. In the absence of vibration and 3DOF testing, and based solely on these results, it is impossible to claim that the CRS03 is fully up to the task of providing the required level of accuracy. However, these tests have fulfilled their purpose by determining whether or not there is merit to the idea of implementing a gyro-based PSS design. Demonstrating offset errors up to 20 times smaller than expected drill tilt 70 magnitudes is ample proof that this approach to PSS design is deserving of further investigation. While the real-world application would introduce significant sources of signal noise, further refinement of the test setup coupled with more sophisticated data analysis procedures could drastically improve already very promising performance and deliver results more amenable to the drill's actual operational environment. Stepper motor discontinuities and slippage, reference measurements of the actual linear displacements, and the data analysis process employed are all prone to propagating some degree of inaccuracy, and therefore a more refined testing procedure must be implemented before any firm conclusions concerning gyro performance can be drawn. Despite these inaccuracies however, percentage errors of 1% or less have already been observed in many trials. As the testing methodology is improved upon, the precision of these figures will also increase. 71 6.0 Conclusions The three p r i m a r y goals of this project were to: 1) propose a feasible so lu t ion for t rack ing the robotic l i m b u n d e r g r o u n d , 2) to deve lop a m e t h o d for testing this po ten t ia l so lu t ion , a n d 3) to evaluate the general feasibi l i ty of the p r o p o s e d so lu t ion based o n in i t i a l testing. In p r inc ip le , the gyro-based approach to PSS des ign is s o u n d - were a gyro able to detect the rotat ion rate about the d r i l l head p r o d u c e d b y the e n d effector rear feet, the d r i l l ' s o r i e n t a t i o n c o u l d be de t e rmined . Th i s s o l u t i o n offers a u n i q u e advantage i n that i t is based o n m e a s u r i n g tilt-rate - an oftentimes cont ro l lab le characteristic of the d r i l l - as opposed to t i l t magni tude . W h i l e a set of gyros c o u l d p r o v i d e data concern ing or ienta t ion , the a m o u n t of u m b i l i c a l d e p l o y e d f rom the a b o v e g r o u n d s p o o l w o u l d indicate the u n d e r g r o u n d distance t raveled by the d r i l l . C o m b i n i n g these t w o pieces of i n f o r m a t i o n v i a the n a v i g a t i o n a l t echn ique k n o w n as dead r e c k o n i n g w o u l d y i e l d the end effector's u n d e r g r o u n d pos i t i on . Since t i l t i ng w o u l d l i k e l y on ly occur as the end effector rear feet engage the borehole w a l l , the gyros w o u l d on ly be required to take measurements d u r i n g this t ime, i n an effort to reduce s ignal noise. In theory this s o l u t i o n seems reasonable , h o w e v e r quan t i t a t ive tes t ing to investigate the gyro 's sensi t iv i ty as w e l l as first-order issues concern ing its use w o u l d have to be carr ied out before any significant deve lopment c o u l d begin. A s such, the second goal of this research had to be met - to deve lop a test system for e v a l u a t i n g the gyro-based approach . D r a w i n g o n detai ls f o u n d i n patents concerning the robotic d r i l l , a test apparatus was des igned a n d constructed w i t h the intent to m i m i c the behav iour or the d r i l l end effector. The system that was deve loped features a stepper m o t o r - d r i v e n l inear actuator that is connected to the rear of an end effector p roxy . A s the l inear actuator moves f rom left to r igh t s imula t ing the extension and retraction of the dr i l l ' s rear feet, a rotat ion about the d r i l l head (a un iversa l joint, i n the case of the test apparatus) is p roduced . 72 With both a proposed solution and a means to test it in place, the third and final goal of evaluating the gyro-based PSS could be worked toward. Using a single gyro, several tests were performed at speeds similar to what is expected of the drill's rear feet, and using a range of sampling rates to eliminate as much signal noise as possible. After each rotation, the gyro output was compared to the actual angular displacement produced by the test apparatus. Results were very favourable, with the best set of tests yielding an average offset from the actual angular displacement of just 0.0082 degrees. Reducing the displacement magnitude by half while maintaining the angular rate resulted in an average offset of 0.032 degrees. Tests conducted at half the expected drill foot displacement rate and to 1.25cm (0.5in) of displacement resulted in an average angular offset of 0.013 degrees. At the very least, these results demonstrate that the low-cost, M E M S gyro-based approach to PSS design has merit and is worth continued investigation. If the same level of performance could be obtained in the three-dimensional case and in the presence of all real-world environmental factors, the stated performance target of producing a borehole with a 0.15%-of-depth deviation may very well be within reach using the approach outlined in this work. Furthermore, it may be possible to adapt this design for use in any kind of long-hole drilling system. Although other drills don't necessarily include stabilizing feet, existing designs are still prone to tilting. It is possible that a set of gyroscopes could be configured to detect the tilt rate in a traditional, rigid drill string-based operation. Although very little could be done in the way of corrective action during drilling, having access to real-time underground position data would eliminate having to determine the borehole profile with a separate sensor package (eg: IMU, clinometers, etc.) once drilling was complete. 73 6.1 Future Work The gyro-based PSS may hold significant potential, however much work remains in taking the project from its current state to the point where that potential may be realized in a real-world setting. To continue development, the project can be broken up into four main branches of research: • Test Apparatus Refinements: One branch of development should concentrate on improving the test hardware. Even if the drill prototype is readied for testing PSS configurations in the near term, the need to provide a controlled environment for sensor testing will still exist. Initial improvements should include the employment of a linear actuator capable of outputting smooth displacements at a range of low speeds. In addition, an improved scheme for determining the actual angular displacement would be very beneficial. It is imagined that this new measurement system could either be developed in conjunction with a more sophisticated linear actuator assembly, or be included as a high-precision clinometer mounted on the end effector proxy. Once these initial issues are worked out, or perhaps in conjunction with these developments, a true 3DOF linear actuator should be designed and constructed to allow for a more realistic testing environment. Although the 1DOF test apparatus is adequate for general testing of multi-gyroscope configurations, a 3DOF system would also allow for the testing of feedback control system designs. • Environmental Factors: For any test to be considered truly realistic, several additional factors should be included in the test setup. Chief among these are vibration and shock. The first concern of this research branch should be to obtain more detailed information regarding the drill's vibration characteristics, and to determine whether stoppages in the 74 drilling procedure are feasible for collecting vibration-free data. If it is found that work stoppages are not permissible, and with the drill's vibration characteristics in hand, work can begin on refining the current vibration mount design, or developing a new approach from scratch. The vibration mount design should be carried out in parallel with any proposed changes to the linear actuator since these two systems are very much interrelated. Additionally, the output from many different types of sensors (including the CRS03) varies with temperature, and it would be desirable to include a mechanism for controlling this environmental parameter to more fully understand gyro behaviour over the course of a drilling operation. Furthermore, although the umbilical will be sealed from the external environment, the potential exists for small amounts of debris to enter the region where the sensors will be mounted. Care must be taken to ensure that the sensors are as ruggadized as possible to combat any contaminating material that may be present. Multi-gyro Configuration: Whether testing is conducted on a 1DOF or 3DOF platform, there are several issues that remain to be worked out in terms of using several gyros together in concert. First and foremost would be developing the capability to interpret the data from three gyroscopes in order to determine the actual direction and magnitude of tilting. Since all gyros feature some degree of off-axis sensitivity, even if a rotation occurs about a single axis, all three gyroscopes would produce some kind of output. The task would be to cut through the unwanted components of this output to discern the true signal. The work done in this branch could also lead into the design of the self-recalibration protocol, where the drill pushes itself against the borehole wall using each of the drill's three rear feet, sequentially. If successful, this self-recalibration process will depend heavily on the interrelationship between the three gyros. 75 • Signal Processing: The procedure for analyzing the gyro output described in this work is an initial stab at what would ultimately have to become a much more sophisticated process. Methods adept at filtering out signal noise will be required to improve upon the gyro's accuracy, particularly in the presence of vibration and changes in temperature. Ultimately, real-time processing of the sensor's output would be the goal of this branch of research. Once sufficient headway had been made on this front, a real-time feedback control system could be simulated in conjunction with the advances made in the other branches. It would not be necessary to pursue each of these branches sequentially - each branch could be performed concurrently with the others, as there is some overlap between them. Also, advances in any one of these areas would most likely lead to benefits in the others, so although development across all branches does not necessarily have to be conducted in parallel, communication across all branches, including development of the other drill systems, should be as fluid as possible. Once solutions for all of these issues have been developed to a certain level of maturity, the only significant hurdle remaining would be integration of the PSS with the actual end effector prototype. Key design specifications for the drill must first be established and proven before issues such as sensor positioning and mounting can be tackled. Until then, the knowledge gained from lab testing the PSS while the end effector prototype is being developed could immediately be put to use once the overall system is ready for integration. 76 Bibliography [I] Wikipedia, the free encyclopedia, "Proprioception," Apr i l 2006, [2] F. Karlsson, "What makes holes deviate?," Geodrilling International, vol. 5, no. 4, p. 1, May 1997. [3] G. Baiden, D. Young, F. Delabbio, J. Pascoli and P. DeVlugt, "Tractor for remote movement and pressurization of a rock drill," U.S. Patent 5,752,572, May 19,1998. [4] Pajari Instruments Ltd. , "Why Bore Holes Deviate," Apr i l 2006, http:/ / why_bore_holes_deviate.htm. [5] O. Issakainen, "Feed storage for drill rods for long-hole drilling apparatus," U.S. Patent 4,632,618, December 30,1986. [6] G. Baiden (private communication), 2005. [7] G. Baiden, D. Young, L, Van Berkel, D. Hoover and P. DeVlugt, "Guided drilling system with shock absorber," U.S. Patent 5,778,987, July 14,1998. [8] G. Baiden and D. Young, "Hydraulic drive for rotation of a rock drill," U.S. Patent 5,853,052, December 29,1998. [9] W. S. Dunbar (private communication), 2005. [10] Laurent ian U n i v e r s i t y , " G u i d e d D r i l l i n g , " A p r i l 2006, [II] Schlumberger, "Oilfield Glossary: Geothermal Gradient," Apr i l 2006, http: / / / Display.cfm?Term=geothermal% 2 Ogradient. [12] J. McGillis, "Method of accurate trenchless installation of underground pipe," U.S. Patent 6,953,306, October 11, 2005. [13] K. Alft, G. Draper and H. Kelpe, "Horizontal directional drilling machine and method employing configurable tracking system interface," U.S. Patent 6,484,818, November 26, 2002. 77 K. Alft, G . Draper and H. Kelpe, "Underground boring machine employing navigation sensor and adjustable steering," U.S. Patent 6,719,069, Apri l 13, 2004. T. Dallas, G. Uttecht, E. Wright, G. Neubauer and J. Brosnahan, "Method for drilling under rivers and other obstacles," U.S. Patent 6,588,116, July 8, 2003. M. Mizuno, "Bore location system and method of calibration," U.S. Patent 6,854,535, February 15, 2005. T, Stockton, "Disconnect information and monitoring system for dynamically positioned offshore drilling rigs," U.S. Patent 5,978,739, November 2,1999. A . Noureldin, H. Tabler, W. Smith, D. Irvine-Halliday and M. Mintchev, "Continuous measurement-while-drilling surveying," U.S. Patent 6,823,602, November 30, 2004. A . Kuckes, "Alternating and static magnetic field gradient measurements for distance and direction determination," U.S. Patent 5,305,212, Apri l 19, 1994. A . Kuckes, J. Gaenger and H. Bayer, "Method and apparatus for producing parallel boreholes," U.S. Patent 5,725,059, March 10,1998 I. Shirasaka, K. Schrader and J. Hache, "Method and apparatus for borehole surveying," U.S. Patent 6,453,239, September 17, 2002. M. Wassell, "Magnetorheological fluid apparatus, especially adapted for use in a steerable drill string, and a method of using same," U.S. Patent 6,257,356, July 10, 2001. G . McElhinney, "Supplemental referencing techniques in borehole surveying," U.S. Patent 7,002,484, February 21,2006. Appl ied Geomechanics Inc., "Clinometer Pak 420," Apr i l 2006, http:/ / Wikipedia, the free encyclopedia, "Dead reckoning," Apr i l 2006, http: / / 78 [26] Columbia Research Laboratories, "SI-701FNE Spec Sheet," Apr i l 2006, http:/ /, % 20SI-701FNE. [27] US Digital Corporation, "T2 Incremental Inclinometer," Apr i l 2006, http:/ / [28] N. Hall, "Long range electronic guidance system for locating a discrete in-ground boring device," U.S. Patent 6,543,550, Apri l 8, 2003. [29] Texas Instruments, "Accelerometer Guide," April 2006, http:/ / www2.usfirst. org/2005comp / Manuals / Accelerl .pdf. [30] Crossbow Technology Inc., "IMU300," Apri l 2006, http:/ / [31] Analog Devices Inc., "ADXL202," April 2006, [32] H. Weinberg and C. Lemaire, "Using the ADXL202 Accelerometer as a Multifunction Sensor (Tilt, Vibration and Shock) in Car Alarms," Analog Devices, August 1998. [33] B. Tryggvason (private communication), 2005. [34] "Developing Science & Technologies List, Section 16: Positioning, Navigation and Time Technology," U.S. Department of Defense, pp. 15-18, January 2006. [35] Honeywell Aerospace, "Satellite Guidance & Attitude Control," Apri l 2006, http:/ / [36] Silicon Sensing, "It is Rocket Science - GYRING: How it works," Apri l 2006, [37] A . Kingsbury (private communication), 2003. [38] Northrop G r u m m a n , " F O G Family Overv iew, ' A p r i l 2006, http:/ / x.htm. 79 [39] Silicon Sensing, "CRS03," April 2006, http:/ / [40] How Stuff Works, "How Do Optical Mice Work?," Apr i l 2006, [41] University of Bath, "Picking things up," April 2006, [42] Hamamatsu, "Position Sensitive Detectors: Two-dimensional," April 2006, [43] MTI Instruments Inc., "MTI-2100 Fotonic Sensor," Apr i l 2006, [44], "A l l you need to know about capacitive sensors," Apri l 2006, [45] Measurand, "Shape Tape," Apri l 2006, [46] B. Bais and B. Y. Majlis, "Mechanical Sensitivity Enhancement of an Area-Changed Capacitive Accelerometer by Optimization of the Device Geometry," Analog Integrated Circuits and Signal Processing, vol. 44, no. 2, pp. 175 - 183, August 2005 [47] Z. Yufeng, T. Xiaoyun, C. Weiping, Z. Guowei and L. Xiaowei, "Study of MEMS packaging technology," in 6th International Conference on Electronic Packaging Technology, 2005, pp. 1 -4 . [48] S. Su, H. Yang and A . Agogino, " A resonant accelerometer with two-stage microleverage mechanisms fabricated by SOI-MEMS technology," Sensors Journal, IEEE, vol. 5, issue 6, pp. 1214 - 1223, December 2005. [49] Kemo-Electronic G m b H , "M109 Step-Motor-Interface 6 Pin," Apri l 2006, [50] National Instruments, "NI USB-6009 Multifunction Data Acquisition for USB," April 2006, http:/ / .eom/nips/cds/view/p/lang/en/nid/14605. 80 Appendix A - Test Apparatus Schematics Figure 29 - Gyro Mount Detail. Three gyros are spaced 120 degrees from each other, atop the vibration assembly. Figure 30 - Vibrat ion Assembly Detail. The vibration plate is pressed against the two stoppers by springs until displaced vertically by the striker. Appendix B - Trial Data Figures 31 through 37 represent the data that was taken during each gyro trial, upon which the conclusions of this work are based. Gvro Outout: Ra=0.45dBa/s6c. S=1kHz. Triad I L r \ — 0 i t e in Timo [tec] 1* 16 Gyro Output: Ra=0.45deg/BBC. S A V * • Ii 1* « Gyro Outout: Ra-Q.45daa/tMC 5^1 kHz. Trial 3 Gwio Outout: Ra=0.45dea/sec. S--1kHz. Trial 4 J II \ rfvy Tlmofsecl 1. 1* is U I . CI Hi 1» 16 IB Tunalaac] Gi/ro OutDUt Ra=0.45dea/sec. S-=7Wfe. 7r/a/5 i C . 2.5S6 V J 4 Time (bg 3 10 12 14 Figure 31 - Data, Ra = 0.45deg/sec, d = 1.25cm, S = 1kHz 83 Gyro Output: Ra=0.45deg/sac. S=2kHz. Trial 1 I III Gyro Output- Ra=0.45deq/sec. S=2kHz. Trial 2 1 aA A*. . -AA . A rt_ L™"v u™"^^V"*V V i Gyro Output: Ra=0.45deg/sec. S=2kHz. Trial 3 £ 2 » 1 II Gyro Output: Ra=a.45deg/sec S=2kHz. Trial 4 (111 1. Gyro Output- Ra=0.45deq/secr S=3kHz, Trial 5 2 $7 i 1 J 256 1 2.S55 • 2 55 Ifww 6 8 10 12 14 16 18 20 Time [sec) Figure 32 - Data, Ra = 0.45deg/sec, d = 1.25cm, S = 2kHz 84 Gyro Output Ra=0.43deg/sec. S=2.8kHz. Trial 7 ll i 1 Tima (see) Gyro Output: Ra=0.45deg/sec. S=2.8kHz. Trial 2 1 1 i m, 1 P 1 * B B 'i; ? 1* Time (sec} G)g-p Output: Ra=0.45dBq/sec. S=2.BkHz. Trials 1 i 1 1 I I i, p GKHO Output Ra=0.45dBg/sec. S=2.8kHz. Trial 4 L 1 Ai, ft Gyro Output: Ra=Q.45deg/sec. S=2.8kHz. Trial t 2.57 2565 § j 2.56 1 2.S5S 2.55 2 545 ll f J , 1 1 '* l rl| I 'IT*-1 2 4 6 8 10 12 14 16 77ms {Bee} Figure 33 - Data, Ra = 0.45 deg/sec, d = 1.25cm, S = 2.8kHz 85 Gyro Output- Ra=D.45deg;/sec. S=S.GkHr. Trial 1 Gyro Output ffa=P 45rfeg/sec. S=5.GkHz. Trial 2 JIJIL il 4 i<4 J " I " lilbi til m Gyro Output-Ra=g.45deg[/sRr., S=S.6kHz. Trial 3 I I I , . I r ' I Gyro Output: Ra=0.45de%/sBc. S=5.6kHz. Trial 4 | *% * I r | r f P Tima [mec) Gyro Output- Ra=OA5de&/sec. S=5.6kHzr Trial 5 2.565 s-O, 2.55 1 2 555 10 12 . H 16 Time (sec) Figure 34 - Data, Ra = 0.45 deg/sec, d = 1.25cm, S = 5.6kHz 86 Gyro Output Ra=0.45deg/sec. S= 11£kHz. Trial 1 Time face) Gyro Output Ra*a.4Sdeg/sei:. S=11.2kHz. Trial 2 III 1 1 r 1' '' ' '1 1 1 Vmo feoo) Gyro Output: Ba=0.45deQ/sec. 5=11.8kHz. Trial 3 1 J 1 1 1 ' 1 1 1 Gym Output Ra=0.4SdBg./sec. S=11.2kHz. Trial 4 i 1 ukiiiiiiiiikiJi^ iiiiiiAiiiiikiiiiji I I S ' i ' 'I | " " " Gyro Output: Ra=0.45deg/sec. S=11.2kHz. Trial 5 Time (sec} Figure 35 - Data, Ra = 0.45 deg/sec, d = 1.25cm, S = 11.2kHz 87 Gyro Output- tla=a.45deq/sec. d=O.Bcm. S=2kHz. Triall 1 I ^ AtfA 1 e ID i 1* i6 TcTte/eocj Gvro Output- Ra=0.45deq/sec. d=0.6cm. S=2khtz. Trial 2 8,Z5I I C 2 < 6 a ;o 11 r* It Gyro Output Ra=Q.45deg/sec. d=D.Ecm. S=2kHz. Trial 3 p Gyro Ot/tpuc r)a=a.45dBq/sac. d=0.6cm. S=2kHz, Trial 4 1 ± k Gyro Output Ra=0.45deq/sec. d=O.Bcm. S=2kHz. Trial 5 ft x n 1. w sty. 3 2 * 6 - 8 10 12 W 1 77me fsac/ Figure 36 - Data, Ra = 0.45 deg/sec, d=0.6cm, S = 2kHz 88 Gyro Output- Ra=0.23deg/sec. d=1.25cm. S=2kHz. Trial 1 iset 1 A U HWy f I * V"-"* j » F e tc -3 M is Timo(OBe) Gyro Output Ra=0.23deq/sec. d-1.25cm. S=2kHz. Irial2 r 1 MA/V 1 ' i 1 «5 ll M H Timefaecl Gyro Output Ra=0.23deg/aec, d=1.25cm, S=2kHz, Trial 3 In/Id Jlr'KI M M A i ft, I M All !U II VWvV o 2 * e » io ii io Tone [sec} Gyro Output: Ra=0.23deg/sec. d= 1.25cm. S=2kHz. Trial 4 2 6W . . . ' . . . . . AH/IN 1 i MM •MM r Timo(tca) Gyro Output Ra=0.23deg[/sec. d= 1.25cm. S=2kHz. Trial 5 4 J U I / 0 r I Mr, M 1/ 1 |v»| p y 0 2 4 6 . 8 10 12 14 16 Time [sec) Figure 37 - Data, Ra = 0.23 deg/sec, d = 1.25cm, S = 2kHz 8 9 Appendix C - Raw Data Trials Unaveraged Gyro Output. Ra=0.45deg/sec. d=1.25cm. S=2kHz Unaveraged Gyro Output: Ra=0.45deg/sec, d=1.25cm. S=10Hz The top panel of figure 38 represents data sampled at 2kHz, without datapoint averaging. As can be clearly seen, the signal is significantly more noisy than similar trials that featured averaging in groups of 200 datapoints, as performed in the other trials. The bottom panel of figure 38 represents unaveraged data sampled at 10Hz, which corresponds to the C R S 0 3 ' s b a n d w i d t h . A l t h o u g h the t r ia l followed the same general pattern of inducing rotation approximately 10 seconds into the trial, the signal is so noisy that it is dificult to pointpoint where the rotation actually occurs. Tuns fsecj Figure 38 - Unaveraged Gyro Output 9 0 . 


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items