FC95-230

Automated Robot Inspection of Spray Painted Surfaces authors

a U U

ANN T. WILKEY Senior Memberflechnical Staff Sandia National Laboratories Albuquerque, New Mexico

JOHN W. SCHWARZ Senior Memberflechnical Staff Sandia National Laboratories Albuquerque, New Mexico *

abstract Applications of paints and coatings can be greatly enhanced by using robotic spraying. Benefits include increased application rates with tighter control over coating finish, quality, and costs. The use of robotics can reduce human exposure to hazardous materials and dangerous spray environments. Successful deployment of a robotic system requires process feedback, automated task and path planning, and teleoperation. This paper describes a measurement system consisting of an inspection sensor and automated path planner for automated robot inspection of painted surfaces. The inspection sensor includes ultrasonic and eddy current components configured in a mechanism providing both compliant force and compliant motion.

conference FINISHING '95 September 18-21, 1995 Cincinnati, Ohio .

terms Robots Robotics Sensors Painting

Measuring Instruments

I

Society of Manufacturing - Engineers

1995

0 ALL RIGHTS RESERI

Society of Manufacturing Engineers One SME Drive. P.O.Box 930 *Deafborn, MI 48121 Phone (313) 271-1500

FC95-2 30

DISCLAIMER This report was prepared as an account of work sponsored by an agency of the United States

.-

Government. Neither the United States Government nor any agency thereof, nor any of their employees, makes any warranty, express or implied, or assumes any legal liability or responsibility for the accuracy, completeness. or usefulness of any information, apparatus, product, or process disclosed, or represents that its use would not infringe privately owned rights. Reference herein to any specific commercial product, process, or service by trade name, trademark, manufacturer, or otherwise does not necessarily constitute or imply its endorsement, recommendation, or favoring by the United States Government or any agency thereof. The views and opinions of authors expressed herein do not necessarily state or reflect those of the United States Government or any agency thereof. ~

~~~

~~~

~

.

~~

~~~~~

__

-

-

INTRODUCTION The ability to inspect the painted surface is essential to developing a fully telerobotic painting system. Inspection prior to painting. can confirm the absence of d a c e contaminaton, the presence of primers, and the readiness of the Surface for painting. Inspection after the painting process can provide information on paint deposition, quality, and adhesion. The Intelligent Systems and Robotic Center at Sandia National Laboratories has recently developed and tested a fully telerobotic painting system. This system consists of a robot system and mobile control room. The robot system includes a hydraulically driven, six degree-of-fteedom manipulator with three quick interchange tools: Inspection Sensor, Surface Finishing Tool, and spray gun. The control room is located at a distance 30 meters from the robot. It contains the computer work stations, --based controller, and video interface. The operator controls the robot fiom a graphical user interface (GUI).This GUI enables the operator to execute the robot subsystems without previous knowledge of computer languages, robot instruction langages, or data acquisition. The robot subsystems include the Video Calibration and Targeting System (VCTS), SMART0 . *This work was performed at Sandia National Laboratories operated for the US Department of Energy under contract number DE-AC04-94AL.85000.

FC95-230-2

control system [l], Sancho GUI, Robot Independent Programming Environment (RIPE/RTPL& and Automated Path Planner software. Operation of the telerobotic system requires no prior experience or knowledge of robot operation, robot programming, or computer lanwues. After a five minute tutorial, new operators have successfully completed robotic painting, inspection, and finishing tasks. From the control room, the robot operator instructs the system to plan the paths for painting and inspection. Then, the operator is prompted to select the object to be painted” from the graphic.d mdel-oof the robot workcell. The operator can specify painting and inspection parameters or select the default parameters which are based on the kinematics of the robot and geometry of the object. After specifying process parameters, the operator can either simulate the results of the Automated Path Planner using the graphical model or proceed with the painting and inspection processes. In the latter case, the operator invokes the VCTS to determine the placement of the object with. respect to the robot. Using stereo triangulation, the VCTS determines the location and orientation of the object. Afterwards, the coordinate system of the object to be painted is referenced to the robot’s base coordinate system and the graphical workcell is updated with this information. At this point, the operator can choose to inspect the object with the Inspection Sensor or paint the object. The Inspection Sensor is capable of measuring and displaying the deposition, quality, finish, and proper adhesion of paints and coatings (Fi-gxe 1 and 2). The process of inspecting the painted surface is automated through the use of the Automated Path Planner, which generates the robot path required to inspect the surface based on CAD model information and user-specified process parameters.

Figure 1. Inspection Sensor with Quick Change Adaptor.

I

FC95-230-3

Figure 2. Inspection Sensor Subassemblies. *

.

* .

The Inspection Sensor includes both eddy current and ultrasonic components configured in both a compliant force and compliant motion mechanism. Industry has previously used eddy current sensors for measuring the deposition of paints and coatings on metallic surfaces. Rather than iterative or sampled eddy current sensors available on the market, the Inspection Sensor offers continuous or sweeping measurements. The ultrasonic sensor component of the Inspection Sensor permits deposition measurements on both metallic and non-metallic surfaces. In addition, the ultrasonic component provides information on surface contamination, paint adhesion, and paint quality when coupled with the deposition measurement provided by the eddy current sensor. _ . - ... . +.. This paper will discuss the Inspection Sensor and Automated Path Planner. First, the requirements for the Inspection System are -presented. Second, the mechanical configuration and sensor subsystem’&e presented. Third, the automated generation of inspection paths is presented. Finally, benefits and limitations of the inspection process are discussed.

INSPECTION SYSTEM REQUIREMENTS There are two categories of requirements imposed on the inspection system. First, there‘is the requirement to perform the inspection process robotically. Second is the requirement for the Inspection Sensor to be adaptable, flexible, and reusable. The constraints imposed on the requirement to perform the inspection process robotically include those imposed by the accuracy limitations of the robot, compliant motion, compliant force, automated generation of inspection points and paths, remote

FC95-230-4

measurements for teleoperation, data collection and presentation, automated tool exchange, and spray paint environments. The deployed robot, a Schilling Titan II, is a hydraulic manipulator by design. With the addition of the SMART0 control system, the manipulator is transformed into a robot capable of perfbrming various painting, inspection, and finishing tasks. Although hydraulically driven robots exhibit smooth motion profiles at slow speeds, they can exhibit undesired features during starts and stops, compliant motion maneuvers, and compliant force operations. Compliant motion is required to engage the sensor normal to the surface being inspected. This results in the sensors laying flush on the surface.. A small error in any joint of the robot can result in the sensor being off-normal with the surface. In addition, such an error wilI result in the achieved inspection point to deviate &om the desired location. For highly attenuating coatings, the ultrasonic sensor must come into contact with the surface. Accuracy and repeatability of contact measurements require a precise and consistent force during the measurement process. Although real-time force feedback can be intergrated into the software control of the robot, the integrity of the surface being measured can be preserved, without incident, through the use of passive compliant force implemented in the tool design. Generation of inspection points and the paths required to reach these points should occur automatically, without additional operator input. Yet, the operator should be able to select points on the surface for inspection without having to generate trajectories or paths to reach these points. The second requriements category is concerned with the flexibility and accuracy of the Inspection Sensor. The Inspection Sensor was designed to measure latex and enamel paints, urethane coatings, polycarbonate materials, and fiberglass moldings. Commercially available ‘thickness sensors” require either the paint or the substrate to have a metallic composition. By incorporating two complimentary sensing techniques, the Inspection Sensor is compatible with both metallic and non-metallic paints and substrates.

In addition to deposition measurements, the Inspection Sensor was designed to measure the quality and consistency of paint applied, presence of surface contaminations, and proper adhesion. These measurements enhance the autonomy of the robotic system by closing the loop on the painting process without operator intervention. The deposition measurements &om the Inspection Sensor must be accurate and repeatable. The required accuracy is dependent on the application or objective of the painting task. The sensor elements of the Inspection Sensor can be easily changed to meet the specific requirements of the application. In most cases, highly accurate paint depositions are not required. However, the Inspection Sensor can be used to diagnosis

,

I

FC95-230-5

and refine the robotic paint operation, leading to reductions in application time, application costs, applied paint, and generated wastes.

INSPECTION SYSTEM CONFIGURATION Mechanical Configuration The mechanical configuration of the Inspection Sensor is shown in Figure 3. This design provides passive-compliant force; compliant motion, a sensor alignment in a light-weight and,.lowcost arrhgem of Eve d can approach the surface' &th orientation-error - - positional error of three inches in the z-axis direZion whilg main%ining accurate, reliable measurements. The major mechanical components of ttie Inspection Sensor are the sensor housing (Figure 4), flex coupling, Neglator springs, adjustable offset, and Quick Change Adaptor[Z].

P

- -. .~ I .

i

-

.*

.

1. Flex Coupling . 2. Eddy Current Ring 3. Ultrasonic Sensor 4. Neg'ator Spring 5. Tool Offset Adjustment 6. Cabling Access Hole 7. EOA Plate

Figure 3. Inspection Sensor Mechanical Configuration.

____. ..

FC95-230

Automated Robot Inspection of Spray Painted Surfaces authors ANN T. WILKEY Senior Memberflechnical Staff Sandia National Laboratories Albuquerque, New Mexico

JOHN W. SCHWARZ Senior Membernechnical Staff Sandia National Laboratories Albuquerque, New Mexico *

abstract Applications of paints and coatings can be greatly enhanced by using robotic spraying. Benefits include increased application rates with tighter control over coating finish, quality, and costs. The use of robotics can reduce human exposure to hazardous materials and dangerous spray environments. Successful deployment of a robotic system requires process feedback, automated task and path planning, and teleoperation. This paper describes a measurement system consisting of an inspection sensor and automated path planner for automated robot inspection of painted surfaces. The inspection sensor includes ultrasonic and eddy current components configured in a mechanism providing both compliant force and compliant motion.

conference FINISHING '95 September 18-21, 1995 Cincinnati, Ohio .

terms Robots Robotics Sensors Painting Measuring Instruments Society of Manufacturing - Engineers

1995

0 ALL RIGHTS RESERVED

Societv of Manufacturing Engineers One SME Drive. PO. Box 930 .Dea;born, MI 48 Phone (313) 271-1500

FC95-230

DISCLAIMER This report was prepared as an account of work sponsored by an agency of the United States Government. Neither the United States Government nor any agency thcreof. nor any of their employees, makes any warranty, express or implied, or assumes any legal liability or responsibility for the accuracy, completeness, or usefulness of any information, apparatus, product, or process disclosed, or represents that its use would not infringe privately owned rights. Reference herein to any specific commercial product, process, or service by trade name, trademark, manufacturer, or otherwise does not necessarily constitute or imply its endorsement, recammendation, or favoring by the United States Government or any agency thereof. The views and opinions of authors expressed herein do not necessarily state or reflect those of the United States Government or any agency thereof. ~

~

____

-

~

-

~- ~- -

~

~

-~ _

_

INTRODUCTION The ability to inspect the painted surface is essential to developing a fully telerobotic painting system. Inspection prior to painting. can confirm the absence of Surface contamination, the presence of primers, and the readiness of the surface for painting. Inspection after the painting process can provide information on paint deposition, quality, and adhesion. The Intelligent Systems and Robotic Center at Sandia National Laboratories has recently developed and tested a fully telerobotic painting system. This system consists of a robot system and mobile control room. The robot system includes a hydraulically driven, six degee-of-fieedom manipulator with three quick interchange tools: Inspection Sensor, Surface Finishing Tool, and spray gun. The control room is located at a distance 30 meters fiom the robot. It contains the computer work stations, --based controller, and video interface. The operator controls the robot fiom a graphical user interface (GUI). This GUI enables the operator to execute the robot subsystems without previous knowledge of computer languages, robot instruction languages, or data acquisition. The robot subsystems include the Video Calibration and Targeting System (VCTS), SMART0 *This work was performed at Sandia National Laboratories operated for the US Department of Energy under contract number DE-ACO4-94AL.85000.

FC95-230-2

control system [l], Sancho GUI, Robot Independent Progmmmiug Environment (RIPELRIPL~),and Automated Path Planner software. Operation of the telerobotic system requires no prior experience or knowledge of robot operation, robot programming, or computer l a n ~ ~ e After s . a five minute tutorial, new operators have successfully completed robotic painting, inspection, and finishing tasks. From the control room, the robot operator instructs the system to plan the paths for painting and inspection. Then, the operator is prompted to select the object to be painted-fiom the graphical medebof the robot workcell. The operator can specify painting and inspection parameters or select the default parameters which are based on the kinematics of the robot and geometry of the object. After specifying process parameters, the operator can either simulate the results of the Automated Path Planner using the graphical model or proceed with the painting and inspection processes. In the latter case, the operator invokes the VCTS to determine the placement of the object with. respect to the robot. Using stereo trian,oulation, the VCTS detennines the location and orientation of the object. Afterwards, the coordinate system of the object to be painted is referenced to the robot's base coordinate system and the graphical workcell is updated with this information. At this point, the operator can choose to inspect the object with the Inspection Sensor or paint the object. The Inspection Sensor is capable of measuring and displaying the deposition, quality, finish, and proper adhesion of paints and coatings (Figure 1 and 2). The process of inspecting the painted surface is automated through the use of the Automated Path Planner, which generates the robot path required to inspect the d a c e based on CAD model information and user-specified process parameters.

Figure 1. Inspection Sensor with Quick Change Adaptor.

I

FC95-2 30-3

. .

Figure 2. Inspection Sensor Subassemblies. - ,

The Inspection Sensor includes both eddy current and ultrasonic components configured in both a compliant force and compliant motion mechanism. Industry has previously used eddy current sensors for measuring the deposition of paints and coatings on metallic surfaces. Rather than iterative or sampled eddy current sensors available on the market, the Inspection Sensor offers continuous or sweeping measurements. The ultrasonic sensor component of the Tnspection Sensor permits deposition measurements on both metallic and nonmetallic surfaces. In addition, the ult~asoniccomponent provides information on Surface contamination, paint adhesion, and paint quality when coupled with the deposition measurement provided by the eddy current sensor. . . - *.. - .. . This paper will discuss the Inspection Sensor-and Automated Path Planner. First, the requirements for the Inspection System are "presented. Second, the mechanical configuration and sensor subsystem-are presented. Third, the automated generation of inspection paths is presented. Finally, benefits and limitations of the inspection process are discussed. I

INSPECTION SYSTEM REQulREMENTS There are two categories of requirements imposed on the inspection system. First, there'is the requirement to perform the inspection process robotically. Second is the requirement for the Inspection Sensor to be adaptable, flexible, and reusable. The constraints imposed on the requirement to perform the inspection process robotically include those imposed by the accuracy limitations of the robot, compliant motion, compliant force, automated generation of inspection points and paths, remote

FC95- 23 0- 4

measurements for teleoperation, data collection and presentation, automated tool exchange, and spray paint environments. The deployed robot, a Schilling Titan 11, is a hydraulic manipulator by design. With the addition of the SMART@control system, the manipulator is transformed into a robot capable of performing various painting, inspection, and finishing tasks. Although hydraulically driven robots exhibit smooth motion profiles at slow speeds, they can exhibit undesired features during starts and stops, compliant motion maneuvers, and compliant force operations. Compliant motion is required to engage the sensor normal to the surface being inspected. This results in the sensors laying flush on the surface.. A small error in any joint of the robot can result in the sensor being off-normal with the sufiace. In addition, such an error will result in the achieved inspection point to deviate from the desired location. For highly attenuating coatings, the ultrasonic sensor must come into contact with the surface. Accuracy and repeatability of contact measurements require a precise and consistent force during the measurement process. Although real-time force feedback can be intergrated into the software control of the robot, the integrity of the surface being measured can be preserved, without incident, through the use of passive compliant force implemented in the tool design. Generation of inspection points and the paths required to reach these points should occur automatically, without additional operator input. Yet, the operator should be able to select points on the surface for inspection without having to generate trajectories or paths to reach these points. . . The second requriements category is concerned with the flexibility and accuracy of the Inspection Sensor. The Inspection Sensor was designed to measure latex and enamel paints, urethane coatings, polycarbonate materials, ind fiberglass moldings. Commercidy available ‘thickness sensors”require either the paint or the substrate to have a metallic composition. By incorporating two complimentary sensing techniques, the Inspection Sensor is compatible with both metallic and non-metallic paints and substrates.

In addition to deposition measurements, the Inspection Sensor was designed to measure the quality and consistency of paint applied, presence of surface contaminations, and proper adhesion. These measurements enhance the autonomy of the robotic system by closing the loop on the painting process without operator intervention. The deposition measurements from the Inspection Sensor must be accurate and reperttable. The required accuracy is dependent on the application or objective of the painting task. The sensor elements of the Inspection Sensor can be easily changed to meet the specific requirements of the application. In most cases, highly accurate paint depositions are not required. However, the Inspection Sensor can be used to diagnosis

FC95-230-5

and refine the robotic paint operation, leading to reductions in application time, application costs, applied paint, and generated wastes.

INSPECTION SYSmM CONFIGUXATION Mechanical Contiguration The mechanical configuration of the Inspection Sensor is shown in Figure 3. This design provides passive-compliant force; compliant motion, a sensor & p n e n t in a light-weight and..low cost arrkgement; can approach the surface’ &th orientation.error - - of dg& positional error of three inches in the z-axis direZion whild mainhning accurate, reliable measurements. The major mechanical components of tfie Inspection Sensor are the sensor housing (Figure 4), flex coupling, Negyator springs, adjustable offset, and Quick Change Adaptor[2].

P

i

‘

I

.

”

*

*

.-.

1. Flex Coupling . 2. Eddy Current Ring 3. Ultrasonic Sensor 4. Neg’ator Spring 5. Tool Offset Adjustment 6. Cabling Access Hole 7. EOA Plate

Figure 3. Inspection Sensor Mechanical Configuration.

FC95-230- 6

1. Flex Coupling (1.25 Dia X 2.00 Long)

2.3.0 Inch Dia X -75 Inch Long Eddy Current Ring 3.1.5 Inch DiaX 1.5 Inch Long Ultrasonic Sensor

Figure 4. Sensor Housing of the Inspection Sensor.

The potential for robot approach and placement errors motivated the inclusion of compliant motion as a requirement for the Inspection Sensor. Compliant motion is achieved through the use of a flexible shaft coupling. The coupling corrects for placement errors in both x and y axes. Initially, an aluminum coupling with +/-5" flexibility was implemented. The inspection path specification and accuracy of the robot permit this part to be replaced with a stiffer +/-lo shaft coupling. , This part can be excluded for more accurate robots. The ultrasonic and eddy current sensors are attached to Delrin mounting rings which are connected to the flexible shaR coupling. The ultrasor& sensor is directly attached to the coupling and the eddy current sensor is attached with spring shims and projects onetenth of an inch beyond the ultrasonic sensor face. This enables the eddy current sensor to float and align with the surface independent of the ultrasonic sensor. With the force applied by the robot through the extension of the Neg'ator springs, the sensors align with a semi-flat surface. The applied force is determined by the selection of Neg'ator constant force springs. A precise and repeatable eight pound force on the ultrasonic sensor is accomplished with two five pound rated Neg'ator springs. Losses in the spring load occur in both the sliding Delrin shaft and spring shims supporting the eddy current sensor. The

FC95-230-7

configuration of the Neg’ator springs and shaft offer a three inch stroke which more than compensates for robot inaccuracies occuring normal to the surface. The selection of constant force extension springs in fulfillment of the compliant force requirement is based on the cost, availability, ff exibility, and precision. Fabricated of Type 301 stainless, Neg’ator springs with 40,000 cycle fatigue life can be purchased for less than ten dollars. Available in a wide range of loads, it is possible to apply the ultrasonic sensor with a precise, repeatable force in the range of 1-10 pounds. Another option for integrating passive-compliant force into the Inspection Sensor is through the use of a compressed air cylinder. The Inspection Sensor housing was designed to accommodate a 1/2” air cylinder capable of providing a regulated force in the range of 1-20 pounds. One of the benefits of using an air cylinder includes a highly selectable and precise force achieved without changing springs. Using a well sealed, pressure reservoir system the need for a compressed air supply can be eliminated. Due t o the moderate increase in cost and complexity, the air cylinder was not used in the initial configuration. The adjustable offset permits six inches of extension and results in a total tool length of 24 inches. Although the overall length of the Inspection Sensor could be substantially smaller, the additional length permits the robot and workpiece to remain fixed for painting, inspection, and finishing tasks. This reduces the amount of operator interaction, fixturing, repositioning, and associated costs. Autonomous, remote tool exchanges can occur when instructed by the robot operator through the use of a Quick Change Adaptor. Using predetermined tool paths, the operator can command the robot to perform a tool exchange.

INSPECTION SYSTEM CONFIGURATION Sensor Subsystem The Inspection Sensor includes both ultrasonic and eddy current components. These well known, non-destructive measuring techniques can be deployed in either contact or non-contact measurements [3,4]. Contact measurements are required to measure a variety of coatings, wide range of depositions, and non-metallic substrates. Since deposition measurements are dependent on the acoustic, electrical, and magnetic properties of both the material and substrate, a combination of ultrasonics and eddy current sensors were incorporated into the Inspection Sensor. The use of an eddy current sensor is predicated by the fact that many substrates are metallic. The measurement process does not require a couplant or application of pressure. In addition, the hardware and instrumentation are simple, inexpensive, and fast. However, eddy current sensors are not compatible with all paints and substrates, limited in accuracy, and unreliable near edges and comers.

FC95-230-8

The use of ultrasonic sensors permits measurement of both conductive and nonconductive coatings, resolution on the order of lmil, and fast readings. However, the use of ultrasonic sensors require either a coupIant or combination of dry couplant and pressure. The Inspection Sensor relies on a dry couplant of cured RTV. There are two primary configurations for ultrasonic thickness measurements: pulseecho and pitch-catch. In the latter case, the two transducer elements are slightly angled towards each other in the sensor housing (Figure 5). The sound pulse is initiated by one transducer element and propagates into the material in a V-shaped path. The second transducer measures the sound wave pulse reflected from the substrate surface. Knowing the sound velocity of the material being measured, the time interval between the initial pulse and the returned pulse provides the thickness information. Because the pitch-&tch or dual sensing technique is useful in measuring highly attenuating or rough textured surfaces, greater flexibility is achieved by implementing this configuration into the Inspection Sensor. - -

Figure 5. Pitch-catch configuration of ultrasonic sensor.

FC95-230-9

The operating frequency of the ultrasonic sensor must be determined. Low fiequency transducers are required to compensate for the loss of signal in highly attenuating materials or rough surfaces. High resolution of the thickness measurements requires a higher sensor fiequency. Hence, a trade-off exists between measurement resolution and the type and total thickness of the coatings. The optimal sensor fiequency for a group of materials is determined through laboratory experimentation with representative samples of desired thickness. Although sound attenuation, scattering, material variations, surface finish, and surface curvature adversely affect the accuracy of ultrasonic thickness measurements, calibration of the ultrasonic measurement system prior to use can greatly improve the measurement performance and accuracy. Proper calibration requires material samples within the desired measurement range. Initial calibration of the ultrasonic sensors can occur using a witness calibration plate configured with material having similar sound velocity and thickness. For test purposes, a calibration plate has been constructed with four samples of lucite measuring 0.125’: 0.18V7 0.200’: and 0.275’: Several measurements are taken for each sample and the data is tabulated (Figure 6). Using a linear regression model, a relationship between sensor reading and material thickness is determined (Figure 7). After performing such calibrations accuracies in the range of 1 4 % have been measured for coating -. -- - -.thickness in the range of 50-75Omils.

Calibration Data -~

0.300 0.280 0.260 v) .. 20 0.240 .f 0.220 2.. 0.200 0.180 Y 2 0.160 0.140 0.120 0.1 00 650

Y

-

700

750 rnV

800

850

Figure 6. Ultrasonic calibration data for Lucite plates:

FC95-230-10

Linear Regressive Fit 0.4

0.35

--r u) u)

2

x

z u I-

0.3

0.25 0.2 0.15 0.1

8 8 Z O 0 w w . w E e

2

P g 0

0

. mV

0

0

s 8

3 Q

:

3

Figure 7. Linear regression relationship of calibration data.

INSPECTION SYSTEM CONFIGURATION Automated Path Planner In typical robotic spray applications, the robot’s path is determined by the generation of pre-computed trajectories using information gathered during off-line programming. Off-line programming is traditionally accomplished with a human operator using a teach pendant. This approach encompasses human enor, human fatigue, non-optimal robot kinematics, and can require the human operator to perform work in dangerous environments.

It has been previously shown that the costs associated with motion planning of painting and inspection tasks make a significant contribution to the overall cost of painting an object robotically. The costs associated with off-line programming of paths is amortized for large lot sizes. The viability of robotic painting for small lot sizes or single, large objects can be achieved with the Automated Path PIanner[S, 61 which reduces the time and costs associated with motion planning activities. The primary goals of the Automated Path Planner are the minimization or elimination of manual painting activities and their associated costs. These processes include the determination of the robot’s motions, obstacle avoidance, and tool interfacing. The current state of the Automated Path Planner permits automated generation of robot

FC95-2 30-11

painting paths, automated generation of robot inspection paths, Inspection Sensor operation, and spray gun operation. The Automated Path Planner (APP) is a stand-alone, rule-based software package. Using parameters specified by the operator, the APP determines the paths a robot’s spray gun should follow to paint a designated surface and the robot’s Inspection Sensor should follow to properly inspect the painted surface. The Automated Path Planner requires a CAD model of the object or surface to be inspected. To prevent possible interference and collisions with objects in the workcell, CAD models of adjacent or nearby objects are also needed. The APP uses model information in the form of IGFUPm part files. These part files define the basic geometric descriptions of physical entities within the robot’s workcell. The APP requires all surfaces to be represented as coUections of planar polygons. Disregarding coordinate system information, the APP considers only vertex and polygon data. This ficilitates the use of visual targeting for object registration. e

In addition to 1G-a part files, the operator of the system can supply input parameters prior to execution of the APP. In the case of surf& inspection, the operator can speci€y the x-cccis inspection ciisance, y-axis inspection dsimce, inspection tool sf and'^ and initiaI speed (Figure 8). These parameters are specified by the robot operator through the Sandia-developed SANCHO graphical user interfitce [7]. Within this graphical environment, the current or default parameters are displayed. The robot operator can readily change these values by toggling the up or down arrows located next to each individual parameter (Figure 9).

Figure 8. Inspection tag points.

FC95-230-12

Using information determined during the generation of painting paths and infomation specified by the operator, the APP determines the path required to inspect the painted surface. The path for each inspection point consists of three tag points (Fi-we 10). A tag point is three-dimensional point in space with orientation information. Graphically, a tagpoint is represented by a Cartesian coordinate h e . Mathematically, a tag point is represented by the six-tuple, (x,y,z,yaw, pitch, roll). The first tag point, TI,^, is perpendicular to the inspection point and a distance equal to the inspection stand-offiom the surface. The inspection point, T~J,is located on the surface. The last tag point, T1,3 is coincident with TI,^. The speed at which the robot approaches the surface is defined to be one-third of the initia2 speed. After the robot reaches the inspection point, T1,2it pauses one second while measurements are taken. The robot retracts from the surface at one-third of the initial speed to prevent marring the surface. M e r reaching Tis, the robot moves to Tq1 with a speed equal to the initia2 speed and the process is repeated.

Figure 9. Sancho Graphical User Interface

FC95-230-13

’ -

, *

Figure 10. Inspection Goints generated by Automated Path Planner

.

-

-

CONCLUSION

-

c

.- . .C

I

-

. -. 1-

*

The use of the Inspection Sensor improves the autonomy of the robot in painting and coating processes. U G g feedback fiom t&e inspection process,-the’robot performed an‘ order of magnitude better- than- manual phting with respect to uniform material application, surface f i s h , and waste reduction. il

Although the Inspection Sensor performed well in actual tests, there are some inherent limitations in the measurement process. The most serious limitations occur when measuring near inside comers or next to protruding objects. Although eddy current sensors are unreliable near edges, the ultrasonic sensor can provide measurements within the radius of the sensor head. Another limitation arises from the inverse relationship of the curvature of the surface and the thickness ofthe coatings. The Inspection Sensor was designed to be as ffexibile and reusable as possible. However, no single combination of eddy current and ultrasonic sensors will result in a

FC95-230-14

measurement system capable of determining the thichess of the multitude of coatings used in industrial processes. With this in mind, the Inspection Sensor was designed to facilitate quick interchanges of sensor heads to meet the specific application. %

In addition to the ro8ot-operated Inspection Sensor, these sensors have been redesigned for manual use. The Hand-held Inspection Sensor weighs less than two pounds and is battery operated. This sensor has undergone extensive. operation and testing over the past six months. ACKNOWLEDGMENTS The design, development, and deployment of the Inspection Sensor was quickly accomplished by an enthusiastic and talented team at Sandia National Laboratories. The authors thank Stan Dains, John Gieske, David Miller, Colin Selleck, and Peter Watterberg for their participation.

REFERENCES [l] Anderson, R., “SMART: A Modular Control Architecture for Telerobotics,” EEE Robotics and Automation Society, Accepted for Publication in 1995. [2] “Quick Change Adaptor: Operations Manual”, EOA Systems, Inc., November 1992. [3] Hull B;, and John, V., Non-Destructive Testing, Springer-Verlag, New York, W , 1988. [4] Kuttnrff, H., Ultrasonics: Fundamentals and Applications, Elsevier Science Publishers, New York, NY,1991. [5] Wilkey, A., Internal Document: ‘%AutomatedPath Planning for Robotic Application of Faint Coatings, Sandia National Laboratories, April 1995. [6] Chen, P. And Hwang, Y., “SANDROS: A Motion Planner with Performance Proportional to Task Difficulty,” Proc. of B E E ConJ on Robotics and Automation, 1992. [7] McDonald, M. And Palmquist, R., “Graphical Programming: On-Line Simulation for Telerobotic Control,” Proc. of Robots and VisionAutomation, 1993.