Graphical modeling and simulation for design and c

tele-operated clinker clearing robot

J.W.Seo a,⁎,Carl Haas b ,Kamel Saidi c

a

Department of Civil Engineering,Hanyang University,Seoul,133-791,Korea b

Department of Civil Engineering,University of Waterloo,Waterloo,Canada N2L 3G1

c

Construction Metrology and Automation Group,National Institute of Standard and Technology,Gaithersburg,MD 209,USA

Accepted 5December 2005

Abstract

Graphical modeling and simulation to analyze and visualize geometry and dynamic processes can improve construction and maintenance in many ways.This paper presents the usefulness of graphical modeling and simulation for the development of construction/maintenance equipment that requires extensive geometry analysis for machine design and remote-operation for control purpose.The development of a tele-operated clinker clearing robot for the maintenance of lignite-fired power facilities is described.The geometric and kinematic design of the robot was performed based on graphical modeling and simulation through interference analysis and visualization.The use of this graphics technique was further extended to the control of the robot from a remote distance because of the hazardous work environment condition.Graphical representation of the equipment and the work environment enhanced equipment control by providing better spatial perception to the operator.Real-time simulation with graphical models also ensured safe and reliable operation of equipment.It is believed that the workflow and the techniques discussed in this paper can also be successfully employed for the development of new equipment that needs to deal with geometrically complex and hazardous work environments.©2006Published by Elsevier B.V .

Keywords:Graphical simulation;Geometric modeling;Graphical control interface;Tele-operation;Robot design

1.Introduction

Three-dimensional graphical modeling and simulation technique is a valuable tool to improve construction and maintenance processes.Graphical simulation can be achieved by changing position,orientation,and geometry of graphical models and by changing viewing or zooming patterns.Graphical simulation in the context of construction is defined as,“The use of high-powered computer graphics to visualize geometry and animate motion in order to simulate a dynamic process ”[1].With the increased computing power of the personal computers and the decreased expenses for graphics hardware and software platforms,the technique is now used in AEC industry in many ways.Firstly,it enhances the communication between different project participants with the intuitive visualization capability.Design and construction

details can be dynamically visualized with walkthrough functionality.Collisions are checked and eliminated for dynamic elements of the project to improve constructability and maintainability by simulating the construction and maintenance process within the graphically generated virtual world [3,19].In addition,construction schedules can be developed and reviewed by visualizing activity sequences [12].Construction equipment selection can also be optimized based on the performance check with the models [1,8].

The use of this graphics technique is further extended to design and control of construction and maintenance equipment and/or robot.Graphical modeling and simulation in this sense in comparison with the above described construction applications would mean 3-D modeling of equipment and work environment followed by simulation and/or visualization of equipment operation with the purpose of geometric and kinematic design as well as remote-control of equipment.

The benefits of using graphical modeling and simulation for design of equipment/machinery are clearly evidenced by the

Automation in Construction 16(2007)96–

106

⁎Corresponding author.Tel.:+82222201482;fax:+82222939977.E-mail address: jseo@hanyang.ac.kr (J.W. Seo).

0926-5805/$-see front matter ©2006Published by Elsevier B.V .doi:10.1016/j.autcon.2005.12.002

The graphical modeling and simulation technique for the purpose of the control of construction or maintenance equipment is also emerging.In space applications or hazardous operations such as nuclear waste remediation,control techniques based on dynamic graphical feedback updated in real-time have been employed for safe and remote control of equipment[2,6].This technique,often referred as graphical control,graphical control interface or graphical programming,however,depends on the ability to generate accurate work environment model.

Despite the difficult technical issue,dynamically updated graphical models have been successfully employed to control several types of construction and maintenance equipment.One good example would be the joint effort between Caterpillar and Trimble that developed a3-D graphical model based earthmoving equipment control systems[4,17].Other examples are the graphical model-based real-time operator interfaces for excavators, and compactors[7,10,11].Further examples can be found from a survey on graphical control interfaces for construction and maintenance equipment[15].Other benefits of using the simulated or animated graphics technique include the operator training capability in a simulated environment.

This paper presents an exemplary practice on using the3-D graphical modeling and simulation technique for design and control of equipment in hazardous,large-scale,and unstructured environments with complex geometry.The equipment discussed in this paper is a robot developed to break and clear“clinkers”in fossil power plant as an automated maintenance effort.As other construction and maintenance operation,clinker clearing is a very labor intensive,dirty,requiring heavy forces and,most of all,very dangerous.This paper describes the development process of this robot,which was based on the3-D graphical modeling and simulation technique.The focus of this paper is on the3-D graphics technique in relation to the development of geometric design and the control system of the robot.The mechanical system design of the robot along with the test results of the robots mechanical performance such as the robots clinker breaking capability is discussed in Saidi et al.[14].It is believed that the development of new equipment that needs to be tele-operated in hazardous,unstructured and complex environment can benefit from the tools and procedures presented in this paper.

2.Graphical modeling and simulation for equipment design and remote control

This section describes the workflow of graphical modeling and simulation for the development of new construction/maintenance equipment that requires intensive geometric analysis and remote operation.Construction and maintenance equipment that should deal with the existing geometry of constructed facilities carefully for the equipment operation including bridge girder maintenance robots,pipe manipulators, concrete surface finishing robots,and building maintenance equipment could be good examples of equipment that can employ the work flow shown in Fig.1.

The models of the work environment and equipment should be prepared initially at the design stage as shown in Fig.1.To develop the work environment model,available design CAD data of the work environment such as geometry of constructed facilities including plant structures,bridges,buildings,etc.,can be utilized.If the CAD data of the work environment are not available or the task of the equipment involves accurate as-built data of specific work environment component,as-built model should be developed following the necessary measurement activities.

The equipment model should include a kinematic model that specifies the movement of equipment based on joint configu-ration to simulate the motion of equipment.Therefore, conceptual models that include the rough geometry and the motion definition through kinematics are needed.With the3-D graphical models of work environment and equipment as well as the kinematic model,the equipment operation can then be simulated and the necessary geometric and performance analysis for the equipment design can be evolved through the iterative design and analysis(through simulation)process.This process could be done successfully with commercial mechan-ical/civil3-D simulation software.The conventional alternative of physically fabricating models and prototypes can be avoided with this approach,which can save the significant amount of cost and time for the development considering the iterative nature of the design process.

The next step is the development of the graphical control interface for remote operation.At this stage,the developed graphical and kinematic models during the design stage may be utilized.However,the graphical models should represent the real-time status of equipment and work environment correctly and efficiently for the tele-operation purpose.Therefore,the graphical models used for the design of the equipment could be modified and/or adjusted to better aid the equipment control.The design issues related to the graphical control interface at this stage are the graphical representation scheme and the graphics update scheme.For the graphical representation scheme, abstraction including the level of detail for the models,viewing schemes and the effective use of colors should be determined for better decision making for equipment operation.For the graphics update scheme,the selection of sensors and model update scheme such as localization,parameter modeling and occupancy array should be performed.Another issue to be considered at this stage is to determine the type of control architecture.Details on these design issues and guidelines for the design of graphical control interface for construction and maintenance equipment in general can be found in Seo et al.[15].

The following sections describe the development of the tele-operated power plant maintenance robot that performs the

97

J.W.Seo et al./Automation in Construction16(2007)96–106clearing operation of clinkers in the boiler hoppers of furnaces. The workflow described in this section was employed for the design of the robot and for the development of the graphical control system for the tele-operation of the robot.

3.Tele-operated clinker clearing robot

Three-dimensional graphical modeling and simulation techniques explained in the previous selections were applied to the development of a tele-operated clinker clearing robot.In this section,the conventional clinker clearing operation at the lignite-fired electric power facilities is introduced,and the conceptual design of the clinker clearing robot is described.

3.1.Clinker clearing operation

Lignite-fired electric power facilities produce a byproduct of the combustion process commonly referred to as clinkers.The clinkers accumulate along the boiler hopper walls and continuously drop to the bottom into a cooling pool of water. The cooling pool serves to fragment the clinkers and to flush them out of the hopper[9].As shown in Fig.2(a),the hopper is composed of a hopper main structure that contains cooling water and a gate housing structure with a grinder and a hatch opening.A sluice gate separates the hopper main structure from the gate housing structure.This gate is opened to allow the flushing of the hopper contents.However,some clinkers get stuck before they reach the grinder and others are simply too large to be handled by the grinder.These clinkers must be dislodged and broken into small pieces either to be processed by the grinder or to be manually removed.

Clearing clinkers is a dangerous operation.Workers are required to wear cumbersome,hot suits and manipulate a long heavy steel rod connected to a jackhammer.Controlling the jackhammer and steel rod combination with their hands while in heavy protection suits exposes the workers to several safety hazards.This conventional clinker clearing operation is described in Fig.2(b).The vibrations from the jackhammer can cause severe fatigue and internal damage to the workers.The hot and humid environment easily induces worker fatigue.Also,the end of the rod can swing upward and impact a worker if hit by a falling clinker.Clearing while the furnace remains in operation,even at reduced output,increases the danger substantially.The work envi-ronment of the clinker clearing operation is very unstruc-tured,unpleasant,and dirty.Many similarities also exist between typical construction or maintenance work and the clinker clearing operation.The conventional clinker clearing operation is laborious,physically dangerous,and requires large forces.

Fig.1.Graphical modeling and simulation for equipment design and remote control. 98J.W.Seo et al./Automation in Construction16(2007)96–106

3.2.Robot specifications and conceptual design

To improve the dangerous and labor intensive clinker clearing process,a tele-operated clinker clearing robot was developed by the research team.The specifications included in Table 1were considered for the design of the robot.

Considering above design specifications,a customized three-degree-of-freedom manipulator with an pneumatic hammer at

the front end of the robot arm was conceptually designed.As shown in Fig.3,the degree of freedom (DOF)of the robot includes two rotational degrees of freedom and one translational degree of freedom.

4.Graphical modeling and simulation for robot geometry design

This section describes the three-dimensional modeling and graphical simulation technique for the geometric design of the clinker clearing robot.The robot must work in a constrained space,and many possible interferences with the existing structure exist.The robot design was improved and verified through three-dimensional graphical simulation.The robot behaviors could be simulated within the gra-phical work environment,and every step of the operating sequence could be verified with graphical models.The models that were developed at this stage naturally migrated to those used for graphical control for remote operation of the robot.

Table 1

Robot specification Specification

Justification

Tele-operation from a distance of 80ft from the hopper

To avoid the exposure of human labor to the harsh and dangerous environment (hot temperature and flushed water).

Limitation on the set-up time (20min)and the working time (30min)Plant needs to reduce the output of furnace during the clearing operation.

Therefore,the limitation on operation time was needed.

Ability to break clinkers into small pieces Reducing the size of the clinkers was required so that clinkers can be

handled by the grinder.

Operating temperature (140F to as high as 1200F for short periods of time)

Robot needs to withstand the high temperature environment.Ability to handle the flushing operation Flushing is required before the clearing operation.Stable platform Exact positioning of the robot was

expected to be difficult with a mobile platform.

Satisfactory reachability Needs to reach the entire hopper area

avoiding the interference with other components in and out of the hopper structure.

Low-cost components inside the hopper to be sacrificed in possible impacts Molten ashes dropping from the furnace can easily harm the sensitive components inside the hopper.Visual feedback and sensing Requirement for tele-operation.

Fig.3.Three DOF manipulator for conceptual design.

a) Bottm hopper

b) Clinker clearing operation

Steel Rod

Laborer

Jack Hammer

Clinker

Hatch Grinders

Sluice Gate

Boiler Center Line

Gate Housing Structure

Hopper Main Structure

Hopper Structure (Hatch)

Cooling Water Before Flushing

Fig.2.(a)Hopper structure of power plant furnace and (b)clinker clearing operation.

99

J.W.Seo et al./Automation in Construction 16(2007)96–106

4.1.Geometric modeling

The hopper structure and the outer environment that could affect the robot manipulation were modeled.Fig.4shows the photo of the actual hopper structure and its outer environment,and Fig.5shows an isometric view of the work environment model.The photo in Fig.4corresponds to the lower-left portion of the work environment model in Fig.5.Complex pipelines and other outer structures were simplified to virtual ceilings and walls that are not visible in the Fig.5.Fig.6.shows a wireframe representation of these virtual ceilings and walls.Once the work environment model was completed,the various ideas of the robot geometry could be experimented within the graphical work environment to embody the conceptual design.4.2.Robot geometry model

For the platform of the robot,the following two options were considered:1)forklift and 2)attachment frame.The attachment frame option was selected because the forklift-mounted system had problems associated with the exact positioning of the forklift relative to the hopper.Access difficulties for the forklift after the flushing operation also discouraged the option.Therefore,the robot was designed to be anchored to the hopper structure so that it can perform the clearing operation through the access door of the hopper.

The attachment frame and other geometry of the robot was designed with the aid of three-dimensional graphical simula-tion.Fig.7shows the completed geometry model of the robot.Basically,the robot has three main components:1)an arm with a pneumatic hammer attached at the front end,2)an arm insertion mechanism and main cylinders,and 3)an attachment frame.As represented in Fig.7,the arm is inserted into the insertion mechanism.The arm is inserted and retracted by six hydraulic cylinders installed inside of the insertion mechanism for the prismatic motion specified in the conceptual design.For the two rotational degree-of-freedom of the robot,two main hydraulic cylinders and the gimbal structure are used.The two main cylinders extend and retract to rotate the insertion mechanism that contains the arm about the gimbal pivot point.After the pneumatic hammer attached at the end of the

arm is positioned near clinkers,the hammer is activated and breaks clinkers so that they can be handled by the grinder.The analysis of this robot geometry model is further explained in the following sections.

4.3.Operating sequence and interference analysis

As explained in Section 3.1,the hopper main structure is filled with cooling water,and it needs to be flushed before the clinker clearing operation.This flushing situation produced a unique operation sequence for the robot.During flushing,it was desirable to keep the critical components of the robot out of the way of hot water flushing from the hopper.Therefore,the robot was designed to swing open with a hinge mechanism installed on the left side of the robot.Fig.8shows the configuration of the robot when the actuator frame is open.An actuator frame that is a portion of the attachment frame swings open during flushing to keep the insertion mechanism and the main cylinders out of the way of flushing water.The other part of the attachment frame,the mating frame,remains to be attached to the hopper structure.After flushing,the actuator frame is positioned to its working location.Fig.9shows the configu-ration of the robot when the actuator frame is closed to its working position.

With the developed robot design,the following operating sequence was required:1)Winch and clamp the robot to the hopper;2)Open the actuator frame;3)Open the access door;

4)Flushing by opening the sluice gate;5)Close the actuator frame;7)Insert the robot arm;8)

Start operation.

The operation sequence involves possible interference between the robot and the work environment.Any modification of the existing structure was not desirable,so the interferences were carefully checked for the geometric design of the

robot.

Fig.4.Hopper structure and outer

environment.

Fig.5.Work environment model.

100J.W.Seo et al./Automation in Construction 16(2007)96–106

The following interferences were avoided through graphical simulation.

4.3.1.Access door and attachment frame –robot in working position

After the flushing operation,the actuator frame gets into the working position.At that time,the access door should be open and positioned between the actuator frame and the mating frame.The collision between the attachment frame and the access door should be avoided.The right side of the actuator frame and the mating frame was designed to accommodate the door safely.Fig.9shows the access door accommodated by the attachment frame.

4.3.2.Insertion mechanism and actuator frame

The robot arm is a simple pole,and the orientation of the arm is achieved by rotating the insertion mechanism about the gimbal pivot point.This operation involves the interference between the insertion mechanism and the attachment frames.The limitation of the rotation of the insertion mechanism was identified with the interference analysis.Fig.10(a)shows the robot and the hopper structure,and Fig.10(b)shows the collision-free movement of the insertion mechanism.4.3.3.Camera enclosure and the opening

The CCTV camera for the visual feedback is protected with the camera enclosure from the heat of the bottom hopper.It was

designed to rotate with the robot arm.The size and shape of the camera enclosure was designed to avoid interferences with the access opening and the actuator frame.Fig.10shows the collision-free camera movement.4.4.Workspace analysis

It was important to analyze the workspace of the robot arm and to determine the proper length of the arm.The inner and outer interferences limit the workspace of the robot,so the workspace was determined considering the interferences.The whole arm has to move back and forth for insertion and extraction because expensive components such as telescopic arm were not allowed in the hopper as explained in Section 4.1.Therefore,the determination of the arm length was important.The arm should be long enough to be able to reach far away from the access door.However,a long arm cannot operate in the area closer to the access door because of interference with outer structures.It was decided to use two types of arms (long and short).Each arm was designed to compensate for each other's workspace limitation.Fig.11shows a good example.In Fig.11(a),the long arm can reach to P 2,but it has interference with the floor to reach to P 1.On the contrary in Fig.11(b),the short pole can not reach to P 2,because of the short length,but it has no interference to reach to P 1.

The work space analysis has been done with an AUTOLISP routine.AUTOLISP is an interface language for AutoCAD ™.The routine iterates interference checking between the arm and the existing structures by changing the insertion length and

the

Fig.6.Virtual ceiling and wall.

Attachment Frame

Arm

Main Cylinders

Insertion

Camera Enclosure

Gimbal Pivot

Fig.7.Robot model.Hinge

Actuator Frame

Frame

Fig.8.Actuator frame open for flushing operation.

101

J.W.Seo et al./Automation in Construction 16(2007)96–106

rotation angles of the arm incrementally.The result of the routine is a representation of the reachable volume with a given length of the robot arm.An example of the result of the workspace analysis is presented in Fig.12.The figure shows the top view of the hopper structure,and the horizontally reachable area of two arms at a fixed vertical rotation configuration was represented.The gray-colored area shows the reachable area with the short arm where the long arm could not access because of the outer interferences.The workspace of the robot arm was visually verified by running the routine with the variation in vertical rotation angles.4.5.Animation

The robot was intended to replace the workers in hazardous areas.The communication between the robot designer and the field workers who have been doing the clearing operation manually was critical to check the design of the robot.The robot design and its operation was presented to the field workers with animated graphics for better communication.Animation software 3D Studio ™was used to animate the static CAD models.3D Studio ™is not a robot simulation program,but the robot installation process and the operation could be clearly animated with rendered images.It helped the design process tremendously with the feedback from the field workers.

5.Graphical modeling and simulation for robot control The robot has a CCTV camera for tele-operation as shown in Fig.7.However,it was anticipated that the visual feedback from CCTV may be limited for the tele-operation if the ash obscures the vision.In addition,tele-operation with CCTV feedback has an inherent depth perception problem.Unexpected collisions during the robot operation were also a risk if the CCTV were the only source of the visual feedback.A control interface based on graphical modeling and simulation (graphical control interface)was required to overcome the problems in the remote control of the robot as explained above.The robot is a new device,so the operator should be trained before actual execution of the robot operation.The graphical control interface was also required for safe operator training because the operator can be trained by

running the graphical model of the robot instead of running the actual robot.

5.1.Overall control system architecture

Fig.13shows the control architecture of the clinker clearing robot.The graphical model of the clinker clearing robot is updated based on the real motion of the robot.The sensor data from the robot's actuators give the configuration of the robot in real-time.The clinker model is also updated and represented along with the static CAD models of the hopper environment.The operator gets enhanced visual feedback by combining the live view from the CCTV camera and the real-time updated graphics,but she directly interacts with the robot for the motion control with a joystick.During the motion control,collisions between the robot and the work environment are avoided with the real-time analysis of the graphical models.Fig.14shows the control station with two monitors (CCTV monitor and Graphics Monitor).The graphical interface was also used for safe operator training.

Frame

Access Door

Mating Frame

Fig.9.Robot in working

position.

Camera Enclosure

a)

b)

Fig.10.(a)Hopper wall,insertion mechanism,and camera enclosure.(b)Collision-free movement of insertion mechanism.

The operator could be trained by simulating the robot operation with the graphical model instead of running the actual robot.5.2.Graphics module

A C++based graphics library,OpenInventor ™,and Microsoft Visual C++™compiler were used to develop a customized graphics program.Fig.15shows a view of the graphical control interface screen.For the robot components,only the pole is shown to the operator because the operator does not normally need visual feedback on the attachment frame and the insertion mechanism.A transparent rendering scheme was used for the walls of the main hopper structure and the gate housing structure.With the transparent rendering scheme,the operator can see the pole through the hopper wall,yet the hopper structure is still identifiable.The operator can change the viewpoint easily,and multiple windows can provide different views simultaneously.The operator can also return to the reference views quickly after viewpoint changes.The details on the graphical representation design including abstraction,coloring and viewing schemes can be found in Seo et al.[15].

5.3.Sensing and position kinematics

Sensing is required to update the graphical models in real time.The main cylinders are equipped with linear transducers which report the cylinder length information in real time.A kinematic model was required to calculate the gimbal angles based on the cylinder lengths.Fig.16shows the kinematic structure of the robot.A and B are ball joints that connect the gimbal extension and the cylinders.

A set of equations defining forward position kinematics was prepared to compute the robot's configuration (gimbal angle θand ϕ)based on the data obtained from the transducers (p and q ):

p 2¼c 2þa 2þr 2x þr 2y þr 2

z À2r x s h ðc /c þas /Þ

þ2r y c h ðc /c þas /ÞÀ2r z ðac /Àcs /Þð5:1Þq 2¼c 2þb 2þs 2x þs 2y þs 2

z À2c /c ðs x s h Às y c h Þ

þ2b ðs x c h þs y s h Þþ2c ðs z s /Þ:

ð5:2Þ

The equations can have up to sixteen distinct solutions,and reasonably accurate solutions of the above high degree

a) Long Arm

b) Short Arm

P 2

Interference

P 1

P 2

P 1

Fig.11.Two arms of the robot:(a)long arm;(b)short

arm.

Fig.12.Workspace analysis.

Fig.13.Control system architecture.

polynomials in real-time are difficult to obtain.Instead,a look-up table with pre-calculated θand ϕvalues for 625sets of p and q values was prepared and incorporated into the graphical control face program.To detect the pole insertion length,two potentiometers were installed on the insertion and extraction cylinders.

A dual laser triangulation method was used to update the graphical model of the clinkers.This sensing scheme provides the depth information from the CCTV camera to the clinker surface pointed by a laser beam.Two laser pointers were installed underneath the camera enclose as shown in Fig.17(a).The range information from the camera image plane to the clinker surface (d)is obtained from the geometric relationship between the camera focal length,and the physical laser off-set,and the laser off-set on the image plane as shown in Fig.17(b)and Eq.(5.3).

To effectively update the clinker model by incorporating the sensor data,the occupancy array method was used.The 3-D space of the inside of the hopper was divided into 6-in.cube cells which are not visible to the operator.If a point on the clinker surface is detected by the laser triangulation,the cube within which the point is located is considered occupied,and the cube becomes visible to the operator.Considering the 3-in.

discrete movement of the insertion motion of the robot,a 6-in.cube cell to represent occupied space was used.Fig.15shows the graphical control interface screen with clinker models.d ¼f ðy þh Þ=y ð5:3Þ

where,h Distance between Laser 1and Laser 2f CCTV camera focal length

y

Distance between two laser dots on the image plane.

6.Tests and evaluation

Tests on the performance of the developed tele-operated clinker clearing robot were conducted.The test results of the mechanical performance such as the clinker breaking capability of the robot are discussed in Saidi et al.[14].This section describes the graphical model accuracy test,the collision avoidance test,and the operator's performance test with the developed graphical control interface.6.1.Accuracy tests

The geometric model was proven to be correct because the robot could be installed and operated as planned as shown in Fig.18.To check the accuracy of the graphical model of the robot during the robot operation ,the position of the end point of the pole was measured.The average offset (distance between the measured data and the calculated data from kinematics)was 0.28in.,which is sufficient for the clinker clearing operation.The graphical model based collision avoidance functionality tested with the mock-up structure was also successful.The pole was stopped before actual collisions occurred,and then allowed to move only in the safe direction as shown in Fig.19.

The

Fig.14.Robot control

station.

Fig.15.Graphical control interface screen.

Fig.16.Kinematic structure of the robot.

accuracy of the laser triangulation range sensing was also tested.The average error was 2.16in.with decreasing accuracy as the distance to the target increases.6.2.Operator's performance tests

The operator's performance with the developed graphical control interface system was tested by comparing the operator's performance of the CCTV-based operation with that of the graphical control interface and CCTV combined operation.Two targets were placed with different depths on the mock-up structure shown in Fig.19(a).The subjects'task was to hit the targets with the end point of the pole.The test subjects were asked to hit the front face of the targets with the end point of the pole,and not to hit the targets from the side with another part of the pole.The side hits with other parts of the pole were considered as errors,because they mean the subjects mis-

interpreted the depth information.The subjects were also instructed to hit the closer target first to reduce the operation time.Therefore,the three operational performances observed to measure the speed and the quality of the robot operation are:1)

d

f y

h

Camera Image Plane

Top View

Laser 1Laser 2

Rear View

Laser 1Laser 2

Camera Lens Camera Enclosure

(a) Laser set-up (b) Range calculation

Fig.17.Dual laser triangulation for range sensing:(a)laser set-up;(b)range

calculation.

Fig.18.Robot installed at the power

plant.Fig.19.Collision avoidance test result.

The subjects finished the operation5s faster on average when they used the combined system(Graphics and CCTV). The average finishing time of the CCTV-based operation and the combined system were2min15s and2min10s, respectively.The speed of the operation with the graphical control interface could be improved by reducing the time required for the laser triangulation to model the clinkers.Further investigation on the clinker sensing system will be valuable. The subjects showed significant improvements in the quality of the work when they used the graphical control interface and CCTV together.Nine out of eleven subjects made mistakes,at least once,in detecting the closer target in the CCTV-based operation.More importantly,five subjects hit the targets from the side when they used CCTV only.This is clear indication of the depth perception problem which could induce damage or undesirable results during the robot operation.Obscured vision caused by the smoke in the real furnace hopper structure would make the graphical interface more valuable.

The operator training was also possible with the developed graphical control interface simply connecting the control joystick to the graphical models only.A selected plant maintenance worker without previous experience on the robot control could successfully finish a clearing operation with a test setup after a3-h training session.

7.Conclusions

This paper presented a development of a tele-operated robotic device for clearing bottom ash clinkers from lignite-fired power plant furnaces.Graphical modeling and simulation approach was utilized for the whole development process.The geometric analysis of the robot and the work environment and the visualization of the operating sequence with3-D graphical models very much improved the robot design.Tele-operation of the robot was also much improved with the real-time graphical feedback of the robot and the work environment.The experimental results showed that the developed robot is fully functional and able to remove human workers from the dangerous conditions of the manual clinker clearing process.

The presented approach can be employed for the development of new equipment that needs to deal with geometrically complex and hazardous work environments.The usefulness of3-D graphical modeling and simulation for the design of such equipment is clear.The cost and time for developing such equipment can be saved significantly because of the iterative nature of design process.Graphical modeling and simulation for equipment control,however,depends on the ability to generate accurate work environment model. While the dynamic nature of many construction operations may make the use of graphical modeling and simulation for equipment control difficult,some types of equipment operated in work environments that can be modeled in real time, especially many of the survey and maintenance equipment, can benefit from this approach.Various research efforts on the real-time modeling of construction work site that are currently underway[5,13]would make the use of graphical models for equipment control possible for many other construction operations.

Acknowledgements

The authors wish to thank Houston Lighting and Power and Electrical Power Research Institute for their financial and technical support for this project.The technical contributions of Mr.R.Greer and Dr.S.V.Srinivasan are greatly appreciated. References

[1]D.Alciatore,J.O'Connor,P.Dharwadkar,A survey of graphical

simulation in construction:software,usage,and applications,Report to the Advanced Technological Systems Task Force of The Construction Industry Institute,1991(June).

[2]A.K.Bejczy,W.S.Kim,P.S.Schenker,The role of computer graphics in

teleoperation,Proc.of the ASCE Specialty Conference,Albuquerque,NM, 1994,pp.1–9.

[3]Y.J.Beliveau,T.Dal,Dynamic-behavior for material handling in

construction,Journal of Computing in Civil Engineering8(3)(1994 (July))269–285.

[4]Caterpillar,company homepages,WWW-pages,Url:http://www.cat.com,

2005.

[5]Y.Cho,C.Haas,K.Laiapi,S.Sreenivasan,A framework for rapid local

area modeling for construction automation,Automation in Construction11 (2002)629–1.

[6]B.K.Christensen,Virtual environments for telerobotic shared control.

Proc.of the Int.Society for Optical Engineering,Telemaniplator Technology and Space Telerobotics,1993,pp.74–83.

[7]M.Froumentin,F.Peyret,An operator aiding system for compactors,

Proc.,13th Int.Symp.On Automation and Robotics in Contr.,Tokyo, 1996,pp.359–368.

[8]L.Gary,J.Gee,Introduction to Pro/ENGINEER Wildfire2.0,Schroff

Development Corp,2004.

[9]C.Haas,Feasibility study of alternative clinker clearing solutions.Tech.

Rep.,Dept.of Civ.Engrg.,University of Texas at Austin,1995.

[10]M.Harada,K.Kawamoto,T.Sakuma,Ground-level remote control system

for pneumatic caisson,Proc.,7th Int.Symp.On Automation and Robotics in Constr.,Bristol,England,1990,pp.166–173.

[11]X.Huang,L.E.Bernold,CAD-integrated excavation and pipe laying,

ASCE Journal of Construction Engineering and Management123(3) (1997)318–323.

[12]B.Koo,M.Fischer,Feasibility study of4D CAD in commercial

construction,ASCE Journal of Construction Engineering and Management 126(4)(2000)251–260.

[13]A.Lytle,K.Saidi,R.Bostleman,W.Stone,N.Scott,Adapting a

teleoperated device for autonomous control using three-dimensional positioning sensors:experiences with the NIST RoboCrane,Automation in Construction13(2004)101–118.

[14]K.Saidi,J.Seo,S.Sreenivasan, C.Haas, A.Traver,Design of a

tele-operated robot for the maintenance of boiler hoppers in electric power plants,ASME Journal of Mechanical Design121(4)(1999) 7–651.

[15]J.Seo,C.Haas,S.Sreenivasan,S.Saidi,Graphical control interface for

construction and maintenance equipment,ASCE Journal of Construction Engineering and Management126(3)(2000)210–218.

[16]S.Tickoo,CATIA for Designers V5R13,Cadcim Technologies,2004.

[17]Trimble,company home pages,WWW-pages,URL:http://www.trimble.

com,2005.

[18]A.Warszawski,R.Navon,Robot for interior-finishing works,Journal of

Construction Engineering and Management117(1191)(1991)402–422.

[19]M.Williams,C.Bennett,ALPS:the automated lift planning system,

Proceedings of the Third Congress on Computing in Civil Engineering, Anaheim,California,June17–19,1996,812–817.