- Email: [email protected]
Analysis and optimization on in-vessel inspection robotic system for EAST
G Model
ARTICLE IN PRESS
FUSION-8239; No. of Pages 5
Fusion Engineering and Design xxx (2015) xxx–xxx
Contents lists available at ScienceDirect
Fusion Engineering and Design journal homepage: www.elsevier.com/locate/fusengdes
Analysis and optimization on in-vessel inspection robotic system for EAST Weijun Zhang ∗ , Zeyu Zhou, Jianjun Yuan, Liang Du, Ziming Mao Robotics Institute, Shanghai Jiao Tong University, Shanghai, PR China
a r t i c l e
i n f o
Article history: Received 18 September 2014 Received in revised form 15 July 2015 Accepted 24 July 2015 Available online xxx Keywords: Inspection Manipulator Remote handling In-vessel operation EAST
a b s t r a c t Since China has successfully built her first Experimental Advanced Superconducting TOKAMAK (EAST) several years ago, great interest and demand have been increasing in robotic in-vessel inspection/operation systems, by which an observation of in-vessel physical phenomenon, collection of visual information, 3D mapping and localization, even maintenance are to be possible. However, it has been raising many challenges to implement a practical and robust robotic system, due to a lot of complex constraints and expectations, e.g., high remanent working temperature (100 ◦ C) and vacuum (10−3 pa) environment even in the rest interval between plasma discharge experiments, close-up and precise inspection, operation efficiency, besides a general kinematic requirement of D shape irregular vessel. In this paper we propose an upgraded robotic system with redundant degrees of freedom (DOF) manipulator combined with a binocular vision system at the tip and a virtual reality system. A comprehensive comparison and discussion are given on the necessity and main function of the binocular vision system, path planning for inspection, fast localization, inspection efficiency and success rate in time, optimization of kinematic configuration, and the possibility of underactuated mechanism. A detailed design, implementation, and experiments of the binocular vision system together with the recent development progress of the whole robotic system are reported in the later part of the paper, while, future work and expectation are described in the end. © 2015 Elsevier B.V. All rights reserved.
1. Introduction The first Experimental Advanced Superconducting TOKAMAK (EAST) was built in Hefei city of China in 2006 [1], and has been keeping in experimental operation for years. It was reported that the plasma facing components (PFCs) [2], namely, the main parts of first wall in vessel, were partially damaged in some local areas due to the high energy particles and electromagnetic force, and had been optimized and upgraded to full graphite tiles bolted to copper alloy heat sinks from 2008 [1]. The maintenance was a long term and hard task, nevertheless, has to be carried out cyclically in future. Robotics and remote handling technologies therefore have been considered as great challenges for such routine maintenance [3], or for an observation of in-vessel physical phenomenon after the experiments, or for a collection of visual information, etc., since their robust automated/tele-operated features are capable of doing multiple in-vessel operation/inspection tasks without converting or breaking so much the initial environment.
∗ Corresponding author. E-mail address: [email protected] (W. Zhang).
We have proposed a conceptual robotic system composed of base, big arm, and small arm with 10 degrees of freedom in total for EAST inspection, as shown in Fig. 1. The base has one translation and two rotation joints, the big arm takes three sliding DOFs and is responsible to move along toroidal direction in an equatorial plane of the TOKAMAK, and the small arm is for the motion in each vertical section. More detailed description of the structure could be found in literature [4]. So far the situation and requirements for robots have been changing a lot under a long term investigation. An updated discussion on the in-vessel inspection features is given in order to make the later optimization and development easy to be understood.
1.1. A fast setup The main body of robot is stored in an extended equatorial port (W × H, 528 mm × 970 mm) directly connected to the TOKAMAK with an identical vacuum pressure of 10−4 –10−5 pa during plasma discharge, and is driven in/out through the port between the experimental intervals. A thermal shield may be placed to protect the robot.
http://dx.doi.org/10.1016/j.fusengdes.2015.07.024 0920-3796/© 2015 Elsevier B.V. All rights reserved.
Please cite this article in press as: W. Zhang, et al., Analysis and optimization on in-vessel inspection robotic system for EAST, Fusion Eng. Des. (2015), http://dx.doi.org/10.1016/j.fusengdes.2015.07.024
G Model FUSION-8239; No. of Pages 5
ARTICLE IN PRESS W. Zhang et al. / Fusion Engineering and Design xxx (2015) xxx–xxx
2
Fig. 1. The inspection robot system.
A fast setup means that the robot is capable to be used in the first couple of days of experiment intervals while the in-vessel temperature is still high (100 ◦ C level). It helps people to understand early and well on: (1) what have happened in-vessel, especially for physical phenomenon, (2) whether to perform maintenance or to continue scheduled experiments directly, and (3) precise location of the place needs to repair/replace. It enhances an efficient way to fusion research; the intervals could also be shortened consequently. The Articulated Inspection Arm robot (AIA) [3] was designed for ITER under similar considerations. Recent experimental reports [5] show good function under high vacuum and high temperature conditions. 1.2. A close-up observation In order to collect sufficient information of the first wall from scheduled inspection, both accurate motion and precise images are indispensable. The former is ensured by high quality sensors and components mounted into the robot; but the latter must be realized by high definition CCD cameras, and a close-up, preferably, normal observation to the main surfaces of the PFCs. However, performing a very close-up observation takes long time for a whole in-vessel scanning, which not only is inefficient but also enlarges risks of unexpected collision, for instance. Therefore, a switch function of control software is superior to make a optimal effects for a fast scanning routine and a necessary close-up observation. 1.3. Toward a practical and robust implementation As described before, first of all, an automated in-vessel inspection is highly expected for EAST to speed up the research and development of fusion technologies. Meanwhile the scientific research has to meet the requirement toward a practical and robust implementation as close as possible. A combination of high quality commercial components and highly developed techniques may make sense. 2. An upgraded proposal of robotic system 2.1. A binocular vision system We propose a binocular vision system attached to the tip of the robot manipulator, and basically is comprised of two different spec CCD cameras. One camera is in charge of routine but fast scanning task with auto focusing spec, keeping a close distance (250 mm level) from the camera front to main surface of the PFCs. The other camera is with manually adjustable focus lens but higher resolution spec, and is particularly used for more close-up inspection, the distance from the lens front to PFCs is 100 mm level. In order to enhance the inspection efficiency, we install the two cameras very closed to each other but vertically distributed up and down, as shown in Fig. 2. When any suspect image is found during a routine vertical scanning, the system is able to response directly, by interrupting the scanning task and giving an instant close-up inspection at the spot. The high resolution camera then sends back
Fig. 2. The objective of binocular vision system.
Fig. 3. In-vessel dimension unit of mm.
high quality images for further diagnosis. An example is given in the right side of Fig. 2. 2.2. Structural optimization Fig. 3 shows a schematic in-vessel dimension of vertical section of EAST. A highlighted workspace of the camera/lens fronts is plotted between fast scanning (250 mm) and precise close-up inspection (100 mm). The robot, particularly, the small arm for vertical section motion must be capable of carrying the vision system to traverse the whole workspace, and contracting its pose into the equatorial port. It is easy to find a maximum and a minimum concentric radiuses (namely, rmax and rmin ) from the workspace. While, the distance from polar-axis to the common center is defined as equatorial radius Re , which is 1940 mm in EAST. With a detailed examination in Fig. 4, we consider a serially linked configuration with three revolute joints for the small arm, and set the first joint S-I locate at the common center. Define a and b as joint to joint lengths for each arm frame, while, c as joint to camera front distance and partially determined by dimension of the vision system. The three parameters basically keep following geometric/kinematic constraints. • Maximum reachable constraint: a + b + c ≥ rmax • Minimum reachable constraint: c + (a − b) ≤ rmin , or, c − (a − b) ≤ rmin • Safety constraint: max(a, b, c) ≤ rs • Normal observation constraint. Computational optimal solutions of a, b, and c are derivable by dividing the linearized vertical section boundaries into
Please cite this article in press as: W. Zhang, et al., Analysis and optimization on in-vessel inspection robotic system for EAST, Fusion Eng. Des. (2015), http://dx.doi.org/10.1016/j.fusengdes.2015.07.024
G Model
ARTICLE IN PRESS
FUSION-8239; No. of Pages 5
W. Zhang et al. / Fusion Engineering and Design xxx (2015) xxx–xxx
3
Fig. 6. Underactuated mechanism.
Fig. 4. Detailed configuration and optimization.
Fig. 5. The big arm.
several segments. Other constraints, component dimensions, wiring design, for instance, are taken into consideration in the assembly and implementation stage. A final spec of the small arm is reported in Section 4. 2.3. Underactuated mechanism As shown in Fig. 5, the big arm provides three curved sliding DOFs, namely B-I, B-II, and B-III, with four hollow but different frames. It is limited by control to be capable of stretching along a toroidal direction in the equatorial plane only. Therefore, the big arm is devoted to a horizontal inspection by designing following constraints into configuration: (1) the three curved sliding motions are concentric; (2) axis of the motions is identical to polar-axis of the TOKAMAK; and (3) give appropriate radiuses to the motions to fulfill the requirement in Fig. 4. However, the big arm is designed only in charge of a half circle inspection (±90◦ ) due to the dimension limitation, so that we need two sets of that arms in total. In fact, the exact tip position/pose of the big arm is necessary but sufficient info to figure out precise localization of the camera images. We do not need to know which frame is in motion. That is, the three DOFs could be driven independently by different
actuators, or could even be driven by one. And, the latter may lead to simpler implementation in actuator distribution and control, etc. However, there are some difficulties in finding a practical and robust underactuated mechanism. The curved sliding structure is one challenging issue. Attaching tendons (steel cables, for instance) into a telescopic antenna like mechanism is one of the considerable solutions, but still not good for curved sliding, besides its stiffness problem. And, there is few commercial pneumatic or hydraulic products providing curved translation, in addition, they may not be allowable in vacuum even high temperature environment. We propose an electrical actuated conical winding mechanism in Fig. 6. The four hollow frames are simply constrained by commercial curved linear motion guide (e.g., the R Guide Model of THK) so that they can only slide among each other relatively. A conical drum is utilized and connected to servo motor. A thin elastic but curved stainless steel sheet is fixed at one end to the very left frame, and is wound around the conical drum. Constraints are used as support/limit to avoid unexpected sheet bending in motion. It is well known that steel has good properties of both tensile and compressive strengths. Therefore, unlike the tendons, it is possible to employ the winding mechanism to pull or push the frames under well designed constraints. The winding mechanism is aligned inclined to the arm frames with an angle (A-A view) which is identical to the semi conical angle of the drum. Now, define an equivalent radius Rd to represent the drum and formulate the following equation that: ωh =
R d
Re
× ωmotor
(1)
where ωmotor represents angular velocity of the drum and motor, and ωh means the tip sliding speed of the big arm in angular velocity form. The equivalent radius Rd of the drum is basically determined by yield strength of the steel sheet, and gets slight bigger when it winds more portion of that. A simple result is derivable that the actual maximum length of the steel sheet in use is Re × 4/, and 2470 mm in EAST, which is less than 3.3 turns of winding if Rd ≥120 mm. Therefore, we can ignore the nonlinearity in sliding motion since the sheet is thin within several millimeters. 3. Estimation of the inspection 3.1. Inspection efficiency Efficiency here means the duration of a scheduled inspection. This chapter discusses its optimization and some related issues, but only half of the in-vessel inspection is taken into consideration, since two robotic systems could work parallel to save time. And, the structure of robot having big and small arm has successfully
Please cite this article in press as: W. Zhang, et al., Analysis and optimization on in-vessel inspection robotic system for EAST, Fusion Eng. Des. (2015), http://dx.doi.org/10.1016/j.fusengdes.2015.07.024
G Model
ARTICLE IN PRESS
FUSION-8239; No. of Pages 5
W. Zhang et al. / Fusion Engineering and Design xxx (2015) xxx–xxx
4
Table 1 Estimation of inspection efficiency by 0–90◦ angle range in equatorial plane. Plans
VH plan
HV plan
Small arm: continuous motion
Big arm: stepping mode
Small arm: stepping mode
Big arm: continuous motion
Closeness: 250 mm
One circle length: 3850 mm In 65 s 2200 s in total
One step length: 100 mm In 6 s
One step length: 120 mm In 3 s 3460 s in total
One circle length: 3050 mm In 105 s
Closeness: 100 mm
One circle length: 4875 mm In 82 s 3830 s in total
One step length: 70 mm In 5 s
One step length: 80 mm In 2 s 6530 s in total
One circle length: 3050 mm In 105 s
decoupled the whole in-vessel surface inspection into vertical section and horizontal inspections independently. An inspection plan is then creatable by setting a horizontal stepping mode and performing a circular vertical section scanning (VH plan), or by setting a vertical section stepping mode and performing a horizontal scanning reversely (HV plan), or even others. The efficiency varies by different plans, but also depends on the moving speeds of the robot, and the closeness the vision system to PFCs surface. Table 1 shows an estimation of the efficiency under different plans (VH VS HV) and different closeness (250 mm of routine scanning VS 100 mm of close-up inspection). The tip speeds of the vision system are limited to 60 mm/s for small arm vertical section motion and 30 mm/s for big arm horizontal sliding motion, considering the demands for stable and smooth image processing and the limitation of robot joint-velocities. Detailed derivation of the VH plan efficiency is explained below. As shown in Table 1, in the VH plan, the small arm performs a continuous vertical section motion when the big arm slides forward every step. Since the whole sliding stroke for big arm is Re × /2, therefore, the total inspection time is estimated as follows: TVH =
Re · /2 · (circle time + step time) step length
(2)
where the step length is determined by camera spec, and circle time/step time is estimated by the motion speeds and accelerations. Table 1 shows that the VH plan is faster than HV plan, and the efficiency of whole vessel inspection with 250 mm closeness is less than 1.5 h, which is double of 2200 s plus extra time for robot transformation. However, if a close-up inspection is demanded, the duration will extend.
Fig. 7. Inclined observation.
into finite number (N) of unit areas.
N
Sr =
ın · r rn
n=1 N
ı n=1 n
N
× 100% =
ı n=1 n
· cos ˇn · cos n
N
ı n=1 n
× 100% (4)
where ın is defined as importance weights of every unit area according to the maintenance requirements of the relative in-vessel components, ranging from 0 to 1, a bigger value of which means the area is more important for inspection. In the early study on the ITER remote handling, K. Shibanuma and other scholars have classified all ITER components into Class 1–4 [6], from highly dependent on scheduled remote maintenance or replacement to never. Similar methods are expectable to quantify the importance weight of every unit area taken by cameras. Computational approach is capable of detailed derivation of the success rate. A simple result for VH plan with 250 mm closeness shows that the success rate is about 91%, where ın is set to be constant 1.0 for simplicity meaning that every in-vessel surface is same important.
3.2. Success rate 3.3. Fast localization As mentioned above, precise images by normal observation are expected in order to get sufficient information for whole vessel diagnosis. However, as shown in Fig. 4, the normal pose of vision system is not fully achievable during the vertical section scanning, due to the irregular in-vessel contour also the dimensional limitation of the vision system. In addition, as shown in Fig. 5, it also has to inspect inclined when the big arm slides near to its base. It is necessary to find a mathematical approach to quantify the level of normal observation, and then evaluate the level of successful inspection. Note that the resolution of unit area drops if an image is taken inclined. By defining two inclined angles, ˇ in vertical section and in equatorial plane in Fig. 7, it becomes possible to measure a ratio of the actual resolution over ideal resolution of unit area in Eq. (3). r r=
resolution ideal · cos ˇ · cos × 100% resolution ideal
(3)
Define success rate Sr in time in Eq. (4) to represent how well a whole in-vessel inspection is achieved by dividing all of the images
All of the in-vessel images together with the relative coordinate information by encoders are merged into remote virtual reality system in real time. If any suspect image at any position is found by comparing with standard ones, either a marker is recorded or a command is sent directly to robot side for instant close-up inspection. The robot returns back to routine inspection later. This fast localization technique enhances the inspection efficiency with less cost. 4. Implementation and experiments 4.1. Implementation of the binocular vision system The design of binocular vision system is shown in Fig. 8 in detail. We utilize two digital cameras with Ethernet interface vertically up and down, one is 2448 × 2050 resolution CCD with manual focus lens in charge of close-up inspection, and the other is 1600 × 1200 resolution CCD for routine scanning. A LED array circuit board is
Please cite this article in press as: W. Zhang, et al., Analysis and optimization on in-vessel inspection robotic system for EAST, Fusion Eng. Des. (2015), http://dx.doi.org/10.1016/j.fusengdes.2015.07.024
G Model
ARTICLE IN PRESS
FUSION-8239; No. of Pages 5
W. Zhang et al. / Fusion Engineering and Design xxx (2015) xxx–xxx
5
Fig. 10. The thermal test of vision system.
Fig. 8. The structure of binocular vision system.
4.3. Tests of the vision system Basic tests on the vision system were carried out in high temp box furnace in Fig. 10. It was carefully set up from room temperature to 120 ◦ C by steps. Each step kept 1–2 h long. Images (in the right side of Fig. 10) recorded by the remote system showed a good performance of the vision system. 5. Conclusion
Fig. 9. The small arm.
attached in front as light source. All the electrical components are protected by copper tubing wound aluminum alloy outer case. Cooling liquid flowing through the copper tubing takes away the heat from the environment. With vacuum sealing components (such as Viton O-rings) and inert-gas pumped into the outer case, the vision system is capable of working in EAST. The dimension of the developed vision system is 308 mm × 80 mm × 166 mm (L × W × H), weight is about 6.5 kg, including a part of the fourth revolute axis of small arm. While, the joint to camera front distance, namely the parameter c in Section 2.2, is 253 mm. 4.2. Configuration of the small arm
Acknowledgements
As shown in Fig. 9, all of the mechanical and electrical components for the four axes, including DC motors, gear boxes, encoders, controllers, are designed with a nearby distribution to their relative axes and protected by thermal shields. The joint to joint lengths (a and b in Section 2.2) are designed 350 mm identically. Table 2 shows the spec in detail. All the controllers are connected serially with each other through Ethercat interface, and controlled by a real time control system.
Table 2 Detailed spec of small arm.
Motor Gearbox Encoder Brake Stroke Max speed
In this paper, a detailed analysis and optimization of a 10 DOFs robotic system is discussed, which is particularly developed for the use in EAST with many challenging constraints and expectations, such as high remanent working temperature (100 ◦ C) and vacuum (10−3 pa) environment, close-up and precise inspection demand, also irregular D-shape in-vessel structure. An upgraded robotic system with a binocular vision system, an underactuated mechanism in big arm, and a virtual reality system is described. The reason of binocular vision system and close-up inspection is explained. Path planning, inspection efficiency, success rate in time, and fast localization are defined and discussed. The results shown in the paper give a good estimation and pre-evaluation of the developed robotic system. A detailed design, implementation, and experiments of the binocular vision system and part of the robotic system are stated. Now, the whole system including the big arm and base is under development. In future, reports will be given on the implementation and performance tests of the entire robot system, actual experiments and in-vessel task in EAST are expected.
Axis 1
Axis 2
Axis 3
Axis 4
250 W DC Harmonic Absolute 16-bit Yes ±170◦ 60◦ /s
200 W DC Harmonic Absolute 16-bit Yes −15◦ to 30◦ 80◦ /s
200 W DC Harmonic Absolute 16-bit Yes ±170◦ 80◦ /s
60 W DC Harmonic Absolute 16-bit No ±120◦ 360◦ /s
This work is supported by the China Domestic Research Project for the International Thermonuclear Experimental Reactor (ITER) under Grant 2012GB102001, also is partially supported by National Natural Science Foundation of China (NSFC), No. 51275286, and No. 51175324. References [1] X. Ji, Y. Song, et al., Optimization and update of EAST in-vessel components in 2011, Plasma Sci. Technol. 15 (3) (2013) 277–281. [2] D. Yao, L. Bao, et al., Design, analysis and R&D of the EAST in-vessel components, Plasma Sci. Technol. 10 (3) (2008) 367–372. [3] L. Gargiulo, P. Bayetti, et al., Operation of an ITER relevant inspection robot on Tore Supra tokamak, Fusion Eng. Des. 84 (2009) 220–223. [4] X. Peng, J. Yuan, et al., Kinematic and dynamic analysis of a serial-link robot for inspection process in EAST vacuum vessel, Fusion Eng. Des. 87 (2012) 905–909. [5] L. Gargiulo, J.J. Cordier, et al., Towards operations on Tore Supra of an ITER relevant inspection robot and associated processes, Fusion Eng. Des. 82 (2007) 1996–2000. [6] K. Shibanuma, T. Burgess, et al., Overview of ITER remote handling, in: Fusion Engineering, 16th IEEE/NPSS Symposium, vol. 1, 1995, pp. 280–283.
Please cite this article in press as: W. Zhang, et al., Analysis and optimization on in-vessel inspection robotic system for EAST, Fusion Eng. Des. (2015), http://dx.doi.org/10.1016/j.fusengdes.2015.07.024
ARTICLE IN PRESS
FUSION-8239; No. of Pages 5
Fusion Engineering and Design xxx (2015) xxx–xxx
Contents lists available at ScienceDirect
Fusion Engineering and Design journal homepage: www.elsevier.com/locate/fusengdes
Analysis and optimization on in-vessel inspection robotic system for EAST Weijun Zhang ∗ , Zeyu Zhou, Jianjun Yuan, Liang Du, Ziming Mao Robotics Institute, Shanghai Jiao Tong University, Shanghai, PR China
a r t i c l e
i n f o
Article history: Received 18 September 2014 Received in revised form 15 July 2015 Accepted 24 July 2015 Available online xxx Keywords: Inspection Manipulator Remote handling In-vessel operation EAST
a b s t r a c t Since China has successfully built her first Experimental Advanced Superconducting TOKAMAK (EAST) several years ago, great interest and demand have been increasing in robotic in-vessel inspection/operation systems, by which an observation of in-vessel physical phenomenon, collection of visual information, 3D mapping and localization, even maintenance are to be possible. However, it has been raising many challenges to implement a practical and robust robotic system, due to a lot of complex constraints and expectations, e.g., high remanent working temperature (100 ◦ C) and vacuum (10−3 pa) environment even in the rest interval between plasma discharge experiments, close-up and precise inspection, operation efficiency, besides a general kinematic requirement of D shape irregular vessel. In this paper we propose an upgraded robotic system with redundant degrees of freedom (DOF) manipulator combined with a binocular vision system at the tip and a virtual reality system. A comprehensive comparison and discussion are given on the necessity and main function of the binocular vision system, path planning for inspection, fast localization, inspection efficiency and success rate in time, optimization of kinematic configuration, and the possibility of underactuated mechanism. A detailed design, implementation, and experiments of the binocular vision system together with the recent development progress of the whole robotic system are reported in the later part of the paper, while, future work and expectation are described in the end. © 2015 Elsevier B.V. All rights reserved.
1. Introduction The first Experimental Advanced Superconducting TOKAMAK (EAST) was built in Hefei city of China in 2006 [1], and has been keeping in experimental operation for years. It was reported that the plasma facing components (PFCs) [2], namely, the main parts of first wall in vessel, were partially damaged in some local areas due to the high energy particles and electromagnetic force, and had been optimized and upgraded to full graphite tiles bolted to copper alloy heat sinks from 2008 [1]. The maintenance was a long term and hard task, nevertheless, has to be carried out cyclically in future. Robotics and remote handling technologies therefore have been considered as great challenges for such routine maintenance [3], or for an observation of in-vessel physical phenomenon after the experiments, or for a collection of visual information, etc., since their robust automated/tele-operated features are capable of doing multiple in-vessel operation/inspection tasks without converting or breaking so much the initial environment.
∗ Corresponding author. E-mail address: [email protected] (W. Zhang).
We have proposed a conceptual robotic system composed of base, big arm, and small arm with 10 degrees of freedom in total for EAST inspection, as shown in Fig. 1. The base has one translation and two rotation joints, the big arm takes three sliding DOFs and is responsible to move along toroidal direction in an equatorial plane of the TOKAMAK, and the small arm is for the motion in each vertical section. More detailed description of the structure could be found in literature [4]. So far the situation and requirements for robots have been changing a lot under a long term investigation. An updated discussion on the in-vessel inspection features is given in order to make the later optimization and development easy to be understood.
1.1. A fast setup The main body of robot is stored in an extended equatorial port (W × H, 528 mm × 970 mm) directly connected to the TOKAMAK with an identical vacuum pressure of 10−4 –10−5 pa during plasma discharge, and is driven in/out through the port between the experimental intervals. A thermal shield may be placed to protect the robot.
http://dx.doi.org/10.1016/j.fusengdes.2015.07.024 0920-3796/© 2015 Elsevier B.V. All rights reserved.
Please cite this article in press as: W. Zhang, et al., Analysis and optimization on in-vessel inspection robotic system for EAST, Fusion Eng. Des. (2015), http://dx.doi.org/10.1016/j.fusengdes.2015.07.024
G Model FUSION-8239; No. of Pages 5
ARTICLE IN PRESS W. Zhang et al. / Fusion Engineering and Design xxx (2015) xxx–xxx
2
Fig. 1. The inspection robot system.
A fast setup means that the robot is capable to be used in the first couple of days of experiment intervals while the in-vessel temperature is still high (100 ◦ C level). It helps people to understand early and well on: (1) what have happened in-vessel, especially for physical phenomenon, (2) whether to perform maintenance or to continue scheduled experiments directly, and (3) precise location of the place needs to repair/replace. It enhances an efficient way to fusion research; the intervals could also be shortened consequently. The Articulated Inspection Arm robot (AIA) [3] was designed for ITER under similar considerations. Recent experimental reports [5] show good function under high vacuum and high temperature conditions. 1.2. A close-up observation In order to collect sufficient information of the first wall from scheduled inspection, both accurate motion and precise images are indispensable. The former is ensured by high quality sensors and components mounted into the robot; but the latter must be realized by high definition CCD cameras, and a close-up, preferably, normal observation to the main surfaces of the PFCs. However, performing a very close-up observation takes long time for a whole in-vessel scanning, which not only is inefficient but also enlarges risks of unexpected collision, for instance. Therefore, a switch function of control software is superior to make a optimal effects for a fast scanning routine and a necessary close-up observation. 1.3. Toward a practical and robust implementation As described before, first of all, an automated in-vessel inspection is highly expected for EAST to speed up the research and development of fusion technologies. Meanwhile the scientific research has to meet the requirement toward a practical and robust implementation as close as possible. A combination of high quality commercial components and highly developed techniques may make sense. 2. An upgraded proposal of robotic system 2.1. A binocular vision system We propose a binocular vision system attached to the tip of the robot manipulator, and basically is comprised of two different spec CCD cameras. One camera is in charge of routine but fast scanning task with auto focusing spec, keeping a close distance (250 mm level) from the camera front to main surface of the PFCs. The other camera is with manually adjustable focus lens but higher resolution spec, and is particularly used for more close-up inspection, the distance from the lens front to PFCs is 100 mm level. In order to enhance the inspection efficiency, we install the two cameras very closed to each other but vertically distributed up and down, as shown in Fig. 2. When any suspect image is found during a routine vertical scanning, the system is able to response directly, by interrupting the scanning task and giving an instant close-up inspection at the spot. The high resolution camera then sends back
Fig. 2. The objective of binocular vision system.
Fig. 3. In-vessel dimension unit of mm.
high quality images for further diagnosis. An example is given in the right side of Fig. 2. 2.2. Structural optimization Fig. 3 shows a schematic in-vessel dimension of vertical section of EAST. A highlighted workspace of the camera/lens fronts is plotted between fast scanning (250 mm) and precise close-up inspection (100 mm). The robot, particularly, the small arm for vertical section motion must be capable of carrying the vision system to traverse the whole workspace, and contracting its pose into the equatorial port. It is easy to find a maximum and a minimum concentric radiuses (namely, rmax and rmin ) from the workspace. While, the distance from polar-axis to the common center is defined as equatorial radius Re , which is 1940 mm in EAST. With a detailed examination in Fig. 4, we consider a serially linked configuration with three revolute joints for the small arm, and set the first joint S-I locate at the common center. Define a and b as joint to joint lengths for each arm frame, while, c as joint to camera front distance and partially determined by dimension of the vision system. The three parameters basically keep following geometric/kinematic constraints. • Maximum reachable constraint: a + b + c ≥ rmax • Minimum reachable constraint: c + (a − b) ≤ rmin , or, c − (a − b) ≤ rmin • Safety constraint: max(a, b, c) ≤ rs • Normal observation constraint. Computational optimal solutions of a, b, and c are derivable by dividing the linearized vertical section boundaries into
Please cite this article in press as: W. Zhang, et al., Analysis and optimization on in-vessel inspection robotic system for EAST, Fusion Eng. Des. (2015), http://dx.doi.org/10.1016/j.fusengdes.2015.07.024
G Model
ARTICLE IN PRESS
FUSION-8239; No. of Pages 5
W. Zhang et al. / Fusion Engineering and Design xxx (2015) xxx–xxx
3
Fig. 6. Underactuated mechanism.
Fig. 4. Detailed configuration and optimization.
Fig. 5. The big arm.
several segments. Other constraints, component dimensions, wiring design, for instance, are taken into consideration in the assembly and implementation stage. A final spec of the small arm is reported in Section 4. 2.3. Underactuated mechanism As shown in Fig. 5, the big arm provides three curved sliding DOFs, namely B-I, B-II, and B-III, with four hollow but different frames. It is limited by control to be capable of stretching along a toroidal direction in the equatorial plane only. Therefore, the big arm is devoted to a horizontal inspection by designing following constraints into configuration: (1) the three curved sliding motions are concentric; (2) axis of the motions is identical to polar-axis of the TOKAMAK; and (3) give appropriate radiuses to the motions to fulfill the requirement in Fig. 4. However, the big arm is designed only in charge of a half circle inspection (±90◦ ) due to the dimension limitation, so that we need two sets of that arms in total. In fact, the exact tip position/pose of the big arm is necessary but sufficient info to figure out precise localization of the camera images. We do not need to know which frame is in motion. That is, the three DOFs could be driven independently by different
actuators, or could even be driven by one. And, the latter may lead to simpler implementation in actuator distribution and control, etc. However, there are some difficulties in finding a practical and robust underactuated mechanism. The curved sliding structure is one challenging issue. Attaching tendons (steel cables, for instance) into a telescopic antenna like mechanism is one of the considerable solutions, but still not good for curved sliding, besides its stiffness problem. And, there is few commercial pneumatic or hydraulic products providing curved translation, in addition, they may not be allowable in vacuum even high temperature environment. We propose an electrical actuated conical winding mechanism in Fig. 6. The four hollow frames are simply constrained by commercial curved linear motion guide (e.g., the R Guide Model of THK) so that they can only slide among each other relatively. A conical drum is utilized and connected to servo motor. A thin elastic but curved stainless steel sheet is fixed at one end to the very left frame, and is wound around the conical drum. Constraints are used as support/limit to avoid unexpected sheet bending in motion. It is well known that steel has good properties of both tensile and compressive strengths. Therefore, unlike the tendons, it is possible to employ the winding mechanism to pull or push the frames under well designed constraints. The winding mechanism is aligned inclined to the arm frames with an angle (A-A view) which is identical to the semi conical angle of the drum. Now, define an equivalent radius Rd to represent the drum and formulate the following equation that: ωh =
R d
Re
× ωmotor
(1)
where ωmotor represents angular velocity of the drum and motor, and ωh means the tip sliding speed of the big arm in angular velocity form. The equivalent radius Rd of the drum is basically determined by yield strength of the steel sheet, and gets slight bigger when it winds more portion of that. A simple result is derivable that the actual maximum length of the steel sheet in use is Re × 4/, and 2470 mm in EAST, which is less than 3.3 turns of winding if Rd ≥120 mm. Therefore, we can ignore the nonlinearity in sliding motion since the sheet is thin within several millimeters. 3. Estimation of the inspection 3.1. Inspection efficiency Efficiency here means the duration of a scheduled inspection. This chapter discusses its optimization and some related issues, but only half of the in-vessel inspection is taken into consideration, since two robotic systems could work parallel to save time. And, the structure of robot having big and small arm has successfully
Please cite this article in press as: W. Zhang, et al., Analysis and optimization on in-vessel inspection robotic system for EAST, Fusion Eng. Des. (2015), http://dx.doi.org/10.1016/j.fusengdes.2015.07.024
G Model
ARTICLE IN PRESS
FUSION-8239; No. of Pages 5
W. Zhang et al. / Fusion Engineering and Design xxx (2015) xxx–xxx
4
Table 1 Estimation of inspection efficiency by 0–90◦ angle range in equatorial plane. Plans
VH plan
HV plan
Small arm: continuous motion
Big arm: stepping mode
Small arm: stepping mode
Big arm: continuous motion
Closeness: 250 mm
One circle length: 3850 mm In 65 s 2200 s in total
One step length: 100 mm In 6 s
One step length: 120 mm In 3 s 3460 s in total
One circle length: 3050 mm In 105 s
Closeness: 100 mm
One circle length: 4875 mm In 82 s 3830 s in total
One step length: 70 mm In 5 s
One step length: 80 mm In 2 s 6530 s in total
One circle length: 3050 mm In 105 s
decoupled the whole in-vessel surface inspection into vertical section and horizontal inspections independently. An inspection plan is then creatable by setting a horizontal stepping mode and performing a circular vertical section scanning (VH plan), or by setting a vertical section stepping mode and performing a horizontal scanning reversely (HV plan), or even others. The efficiency varies by different plans, but also depends on the moving speeds of the robot, and the closeness the vision system to PFCs surface. Table 1 shows an estimation of the efficiency under different plans (VH VS HV) and different closeness (250 mm of routine scanning VS 100 mm of close-up inspection). The tip speeds of the vision system are limited to 60 mm/s for small arm vertical section motion and 30 mm/s for big arm horizontal sliding motion, considering the demands for stable and smooth image processing and the limitation of robot joint-velocities. Detailed derivation of the VH plan efficiency is explained below. As shown in Table 1, in the VH plan, the small arm performs a continuous vertical section motion when the big arm slides forward every step. Since the whole sliding stroke for big arm is Re × /2, therefore, the total inspection time is estimated as follows: TVH =
Re · /2 · (circle time + step time) step length
(2)
where the step length is determined by camera spec, and circle time/step time is estimated by the motion speeds and accelerations. Table 1 shows that the VH plan is faster than HV plan, and the efficiency of whole vessel inspection with 250 mm closeness is less than 1.5 h, which is double of 2200 s plus extra time for robot transformation. However, if a close-up inspection is demanded, the duration will extend.
Fig. 7. Inclined observation.
into finite number (N) of unit areas.
N
Sr =
ın · r rn
n=1 N
ı n=1 n
N
× 100% =
ı n=1 n
· cos ˇn · cos n
N
ı n=1 n
× 100% (4)
where ın is defined as importance weights of every unit area according to the maintenance requirements of the relative in-vessel components, ranging from 0 to 1, a bigger value of which means the area is more important for inspection. In the early study on the ITER remote handling, K. Shibanuma and other scholars have classified all ITER components into Class 1–4 [6], from highly dependent on scheduled remote maintenance or replacement to never. Similar methods are expectable to quantify the importance weight of every unit area taken by cameras. Computational approach is capable of detailed derivation of the success rate. A simple result for VH plan with 250 mm closeness shows that the success rate is about 91%, where ın is set to be constant 1.0 for simplicity meaning that every in-vessel surface is same important.
3.2. Success rate 3.3. Fast localization As mentioned above, precise images by normal observation are expected in order to get sufficient information for whole vessel diagnosis. However, as shown in Fig. 4, the normal pose of vision system is not fully achievable during the vertical section scanning, due to the irregular in-vessel contour also the dimensional limitation of the vision system. In addition, as shown in Fig. 5, it also has to inspect inclined when the big arm slides near to its base. It is necessary to find a mathematical approach to quantify the level of normal observation, and then evaluate the level of successful inspection. Note that the resolution of unit area drops if an image is taken inclined. By defining two inclined angles, ˇ in vertical section and in equatorial plane in Fig. 7, it becomes possible to measure a ratio of the actual resolution over ideal resolution of unit area in Eq. (3). r r=
resolution ideal · cos ˇ · cos × 100% resolution ideal
(3)
Define success rate Sr in time in Eq. (4) to represent how well a whole in-vessel inspection is achieved by dividing all of the images
All of the in-vessel images together with the relative coordinate information by encoders are merged into remote virtual reality system in real time. If any suspect image at any position is found by comparing with standard ones, either a marker is recorded or a command is sent directly to robot side for instant close-up inspection. The robot returns back to routine inspection later. This fast localization technique enhances the inspection efficiency with less cost. 4. Implementation and experiments 4.1. Implementation of the binocular vision system The design of binocular vision system is shown in Fig. 8 in detail. We utilize two digital cameras with Ethernet interface vertically up and down, one is 2448 × 2050 resolution CCD with manual focus lens in charge of close-up inspection, and the other is 1600 × 1200 resolution CCD for routine scanning. A LED array circuit board is
Please cite this article in press as: W. Zhang, et al., Analysis and optimization on in-vessel inspection robotic system for EAST, Fusion Eng. Des. (2015), http://dx.doi.org/10.1016/j.fusengdes.2015.07.024
G Model
ARTICLE IN PRESS
FUSION-8239; No. of Pages 5
W. Zhang et al. / Fusion Engineering and Design xxx (2015) xxx–xxx
5
Fig. 10. The thermal test of vision system.
Fig. 8. The structure of binocular vision system.
4.3. Tests of the vision system Basic tests on the vision system were carried out in high temp box furnace in Fig. 10. It was carefully set up from room temperature to 120 ◦ C by steps. Each step kept 1–2 h long. Images (in the right side of Fig. 10) recorded by the remote system showed a good performance of the vision system. 5. Conclusion
Fig. 9. The small arm.
attached in front as light source. All the electrical components are protected by copper tubing wound aluminum alloy outer case. Cooling liquid flowing through the copper tubing takes away the heat from the environment. With vacuum sealing components (such as Viton O-rings) and inert-gas pumped into the outer case, the vision system is capable of working in EAST. The dimension of the developed vision system is 308 mm × 80 mm × 166 mm (L × W × H), weight is about 6.5 kg, including a part of the fourth revolute axis of small arm. While, the joint to camera front distance, namely the parameter c in Section 2.2, is 253 mm. 4.2. Configuration of the small arm
Acknowledgements
As shown in Fig. 9, all of the mechanical and electrical components for the four axes, including DC motors, gear boxes, encoders, controllers, are designed with a nearby distribution to their relative axes and protected by thermal shields. The joint to joint lengths (a and b in Section 2.2) are designed 350 mm identically. Table 2 shows the spec in detail. All the controllers are connected serially with each other through Ethercat interface, and controlled by a real time control system.
Table 2 Detailed spec of small arm.
Motor Gearbox Encoder Brake Stroke Max speed
In this paper, a detailed analysis and optimization of a 10 DOFs robotic system is discussed, which is particularly developed for the use in EAST with many challenging constraints and expectations, such as high remanent working temperature (100 ◦ C) and vacuum (10−3 pa) environment, close-up and precise inspection demand, also irregular D-shape in-vessel structure. An upgraded robotic system with a binocular vision system, an underactuated mechanism in big arm, and a virtual reality system is described. The reason of binocular vision system and close-up inspection is explained. Path planning, inspection efficiency, success rate in time, and fast localization are defined and discussed. The results shown in the paper give a good estimation and pre-evaluation of the developed robotic system. A detailed design, implementation, and experiments of the binocular vision system and part of the robotic system are stated. Now, the whole system including the big arm and base is under development. In future, reports will be given on the implementation and performance tests of the entire robot system, actual experiments and in-vessel task in EAST are expected.
Axis 1
Axis 2
Axis 3
Axis 4
250 W DC Harmonic Absolute 16-bit Yes ±170◦ 60◦ /s
200 W DC Harmonic Absolute 16-bit Yes −15◦ to 30◦ 80◦ /s
200 W DC Harmonic Absolute 16-bit Yes ±170◦ 80◦ /s
60 W DC Harmonic Absolute 16-bit No ±120◦ 360◦ /s
This work is supported by the China Domestic Research Project for the International Thermonuclear Experimental Reactor (ITER) under Grant 2012GB102001, also is partially supported by National Natural Science Foundation of China (NSFC), No. 51275286, and No. 51175324. References [1] X. Ji, Y. Song, et al., Optimization and update of EAST in-vessel components in 2011, Plasma Sci. Technol. 15 (3) (2013) 277–281. [2] D. Yao, L. Bao, et al., Design, analysis and R&D of the EAST in-vessel components, Plasma Sci. Technol. 10 (3) (2008) 367–372. [3] L. Gargiulo, P. Bayetti, et al., Operation of an ITER relevant inspection robot on Tore Supra tokamak, Fusion Eng. Des. 84 (2009) 220–223. [4] X. Peng, J. Yuan, et al., Kinematic and dynamic analysis of a serial-link robot for inspection process in EAST vacuum vessel, Fusion Eng. Des. 87 (2012) 905–909. [5] L. Gargiulo, J.J. Cordier, et al., Towards operations on Tore Supra of an ITER relevant inspection robot and associated processes, Fusion Eng. Des. 82 (2007) 1996–2000. [6] K. Shibanuma, T. Burgess, et al., Overview of ITER remote handling, in: Fusion Engineering, 16th IEEE/NPSS Symposium, vol. 1, 1995, pp. 280–283.
Please cite this article in press as: W. Zhang, et al., Analysis and optimization on in-vessel inspection robotic system for EAST, Fusion Eng. Des. (2015), http://dx.doi.org/10.1016/j.fusengdes.2015.07.024