You are on page 1of 5

Fusion Engineering and Design 124 (2017) 507–511

Contents lists available at ScienceDirect

Fusion Engineering and Design


journal homepage: www.elsevier.com/locate/fusengdes

Static stiffness modelling of EAST articulated maintenance arm using


matrix structural analysis method
Shanshuang Shi a,b,∗ , Huapeng Wu a , Yuntao Song b , Heikki Handroos a , Ming Li a ,
Yong Cheng b , Bingyan Mao b
a
Laboratory of Intelligent Machines, Lappeenranta University of Technology, 53810, Finland
b
Institute of Plasma Physics, Chinese Academy of Sciences, Hefei 230031, China

h i g h l i g h t s

• A general mathematical approach to estimate the static stiffness of robot system based on matrix structural analysis method has been introduced.
• Stiffness model of EAST Articulated Maintenance Arm (EAMA) has been built for purpose of error prediction.
• Deflections of EAMA robot have been respectively calculated by the mathematical and simulation models.
• Less than 5% deviations have been supposed for EAMA stiffness model.

a r t i c l e i n f o a b s t r a c t

Article history: The remote inspection and maintenance of the small pieces in vacuum vessel requires certain accuracy
Received 16 September 2016 to the serial long-reach robot applied in EAST tokamak. The elastic deformations due to gravity and
Received in revised form 4 May 2017 dynamical forces, however, will cause a big trajectory tracking error to the robot end. Therefore, the error
Accepted 15 May 2017
prediction and compensation strategy should always be studied in advance to achieve a high required
Available online 23 May 2017
accuracy. This paper presents a mathematical method to estimate the static stiffness of EAST articulated
maintenance arm (EAMA) robot. Firstly, the stiffness matrix of the modular parallelogram mechanism of
Keywords:
robot is obtained based on the Matrix Structural Analysis (MSA) method. Considering the real constraint
EAST
Articulated maintenance arm
conditions of robot arms, the matrix is further simplified to a final version with minimum dimension to
Stiffness modeling improve the computational efficiency. Secondly, the simulation examples have been given to evaluate
Error prediction the accuracy and reliability of the stiffness equations. And finally, the results show that the presented
Matrix structural analysis approach has an acceptable prediction error (less than 5%), which can be applied to other similar robotic
systems with long-reach complex links.
© 2017 Elsevier B.V. All rights reserved.

1. Introduction zontal plane) and elevation (±45◦ in the vertical plane). Benefiting
from the parallelogram mechanism, robot arms can provide a huge
To ensure remote inspection and simple maintenance oper- reduction ratio and withstand strong torques generated from the
ations of inner components inside vacuum vessel (VV), EAST long-reach arm and its gravity. Besides, the parallelogram structure
Articulated Maintenance Arm (EAMA), a highly redundant serial can also keep the rotation axis of clevis joint always in vertical posi-
robot system with 11◦ of freedoms (DOFs) in total, has been devel- tion. However, due to the long-reach mechanisms with a weight
oped since 2014. It mainly consists of five serial modular arm more than 100 kg, the effect of the gravity will cause huge flexible
segments and a storage cask with a shuttle base which can transfer deflections, which are unacceptable for the robot to run inside a
the robot into EAST VV [1]. The brief description of EAMA modular narrow and complex-shaped space as EAST VV.
segment design can be found in Fig. 1 [2]. Each modular arm can Since most serial long-reach arms normally run at a very low
mostly provide two kinds of motions: rotation (±90◦ in the hori- speed, study on the flexible characteristics in static condition will
be more convenient and efficient and also acceptable for a medium
precision. In this paper, a general model of combined matrix equa-
∗ Corresponding author at: Institute of Plasma Physics, Chinese Academy of Sci- tions have been built to estimate the static stiffness of EAMA robot
ences, Hefei, 230031, China. segments by using the Matrix Structural Analysis (MSA) method.
E-mail address: shiss@ipp.ac.cn (S. Shi). Based on the obtained stiffness equations, the key points’ displace-

http://dx.doi.org/10.1016/j.fusengdes.2017.05.068
0920-3796/© 2017 Elsevier B.V. All rights reserved.
508 S. Shi et al. / Fusion Engineering and Design 124 (2017) 507–511

Fig. 1. Modular design of EAMA robot arms: 5 segments with identical principle design, the 1st to 3rd segments have only rotation motions by using rigid rods instead of
elevation actuators; the 4th and 5th segments have both rotation and elevation motions.

⎡ ⎤
ments of the structure in different positions can be calculated. This [r] 03×3 03×3 03×3
paper presents the detailed process of this approach to the prob- ⎢ ⎥
⎢ 03×3 [r] 03×3 03×3 ⎥
[] = ⎢ ⎥
lems facing the deformation prediction and accurate path planning
⎢ ⎥ (3)
of flexible long-reach arm with complex mechanisms rather than ⎣ 03×3 03×3 [r] 03×3 ⎦
simple links. Besides, some simulations using finite element (FE)
method have been finished to verify the accuracy and reliability of 03×3 03×3 03×3 [r]
the stiffness equations.
2.3. Matrix optimization

2. General procedure for stiffness modelling In most of practical engineering, a complete mechanism will
always have loads and constraints at the same time. For purpose of
2.1. Element stiffness matrix (ESM) making the integrated SSM more efficient for computation, some
simplifications can be furthermore operated by partitioning Eq.
In MSA method, a structure can be divided into several basic (1) into restrained items (with subscript s) and those related to
beam elements. For an arbitrary element with 12◦ of freedoms remaining DOFs (with subscript f), as expressed in Eq. (4).
(DOFs) as shown in Fig. 2, complete stiffness can be denoted as  
a 12 × 12 matrix having zero coefficients for all of the uncoupled Ff Kff Kfs f
forces and displacements, which can be found in Eq. (1) [3–5]. = (4)
Fs Ksf Kss s
In the equations, i K stands for the 12 × 12 stiffness matrix
expressed in beam local reference frame, which is determined by  
As s = 0, the formula of deflections can be finally derived as
the inherent material and geometric properties; F represents the
follow:
loads (forces and moments, 12 elements in total) while  repre-    −1  
sents deflections (linear and angular ones, 12 elements in total); A: f = Kff Ff (5)
the cross-section area; L: length of the beam; E: Young’s modulus;
Iy, Iz, J are the quadratic and polar moments. 3. MSA method applied on EAMA modular robot
   
F1 1 For EAMA robot shown in Fig. 1, the modular structure can be
= iK (1) treated as a parallelogram mechanism, where four bars consist of
F2 2
horizontal rods, robot tube, and two clevises. Fig. 3 gives the kine-
matical model of the entire manipulator. The approximate beam
2.2. System stiffness matrix (SSM) model and its force conditions have been built as Fig. 4. The con-
strains are defined under the assumption that the first clevis is
The stiffness matrix of the entire structure can be derived by strong enough, which means the translational displacements of
assembling several basic ones based on the i K in the Eq. (1). When joint A and B are almost equal to zero. Besides, the total gravity
assembling, superposition principle of MSA should be used to get force and torques caused by the following segments are equivalent
the final complete stiffness matrix with a dimension of 6n × 6n (n to be applied uniformly on joint C and D.
is the number of joint points of whole structure). In this procedure, With this approximate beam model, the ESM of the five beams
the reference coordinate system (CS) should be firstly unified to the in global reference frame can be easily obtained with the method
global one by utilizing transformation formula expressed in Eq. (2), descripted in former section. Moreover, since the whole system is
where matrix [] is composed of four identical rotation matrices a kind of cantilever structure, which means that there are no forces
[r] from global to local frame of beam element (Eq.(3)). For better along z axis or moments along y axis in practice. Therefore, the five
understanding, some similar applications can be found in the work 12 × 12 ESM can be simplified into dimensions of 8 × 8. Then, the
described in [6–8]. reduced ESM can deconstructed into the form as expressed in Eq. (6)
0  −1 i
  for integrating. The leading superscript stands for coordinate frame
K = [] K [] (2) while the leading subscript represents the beam number, here it
S. Shi et al. / Fusion Engineering and Design 124 (2017) 507–511 509

Fig. 2. Definitions of basic beam element in MSA method.

Fig. 3. The kinematical model of the entire manipulator.

can be AB, BC, CD, AD, AC. The following subscript represents the Table 1
The forces and moments applied on robot.
node numbers. For example:0AB K AB represents: the effect on node A
caused from the forces on node B, for beam AB, in global CS. Loads Fyc Mxc Mzc Fyd Mxd Mzd
The final SSM can be obtained as Eq. (7) with a dimension of Value −500 N −500 Nm −750 Nm −500 N −500 Nm −750 Nm
16 × 16 by applying
 the superposition principle of MSA.
 
Furthermore, 0 K ff can be derived from 0 K by
(12×12) (16×16)
taking into account the constraints at joint A and B. Then, the deflec-
With the parameters listed in Tables 1 and 2, the results of robot
tions on the key points of one modular segment can be calculated deflection items with different elevation angles (from −45◦ to 45◦ )
according to Eq. (5). Among the equation, f (Eq. (8)) consists
have been respectively calculated by the MSA and FE method. In FE
  of the four joints. It can be calculated when the
of 12 deflections
simulations, the element of Solid 185 has been selected for static
elements of Ff (Eq. (9)) are known.
analysis in ANSYS Mechanical APDL 17.0. For a comparison, the
  results of proposed mathematical method and the FE simulation for
0  0 K
AB AA
0 K
AB AB the linear and the angular deflections at Joint C have been illustrated
AB K (8×8) =
0 K 0 K
(6)
AB BA AB BB
⎡0 ⎤
AB
K AA + 0AC K AA + 0AD K AA 0 K
AB AB
0 K
AC AC
0 K
AD AD
⎢ ⎥
0  ⎢ 0 K
AB BA
0 K
AB BB
+ 0BC K BB 0 K
BC BC
04×4 ⎥
K =⎢

⎥(7)

(16×16) 0 K 0 K 0 K
⎣ AC CA BC CB AC CC
+ 0BC K CC + 0CD K CC 0 K
CD CD ⎦
0 K 04×4 0 K 0 K + 0CD K DD
AD DA CD DC AD DD
   T
f = xa za xb zb uc vc xc zc ud vd xd zd (8)
   T
Ff = 0 0 0 0 0 Fyc Mxc Mzc 0 Fyd Mxd Mzd (9)
in Fig. 5 (The results at Joint D are similar with Joint C, not presented
here).
4. Numerical results and simulations
Some facts can be summarized from the results:

Based on the above stiffness model, the deflections of all nodes


of the parallelogram structure can be calculated with a set of exter- • For linear deflections: range around −2.5 to 2.5 mm in X (horizon-
nal loads as an input, which in practice can be easily obtained tal) direction, reaching maximum at ±45◦ while −2 to −3.5 mm
from the dynamical equations of the entire robot system. To eval- in Y (vertical) direction, reaching maximum at 0◦ ;
uate the accuracy of this mathematic model, several hypothetical • For angular deflections: range around −0.25 to −0.32◦ along X
loads (approximate level with the actual loads on first arm segment (arm) axis, reaching maximum at 0◦ while −0.28 to −0.36◦ along
during operation) have been adopted as an input which are given Z (elevation) axis, reaching maximum at ±45◦ ;
in Table 1. The material and geometric parameters are shown in • Less than 5% deviations have been found by comparing the results
Table 2. from proposed mathematical method and FE simulation.
510 S. Shi et al. / Fusion Engineering and Design 124 (2017) 507–511

Fig. 4. The approximate beam model and force conditions of EAMA modular segments.

Table 2
The material and geometric parameters of robot components.

Parts A(mm2 ) L(mm) Iz (mm4 ) Iy (mm4 ) J(mm4 ) E(MPa) v


AB 2300 75 758,000 1,010,000 182,000 108,000 0.35
BC 992 1035 3,096,700 3,096,700 6,193,400 108,000 0.35
CD 2300 75
 758,000 1,010,000 182,000 108,000 0.35
AC 254 394 6.936 − sin 5150 5150 10,300 108,000 0.35
AD 201 1035 3215 3215 6430 108,000 0.35

 is the elevation angle of the robot.

Fig. 5. The deflection results of Joint C calculated from mathematical stiffness model and FE simulations.

5. Discussions and conclusions parison with the FE simulation results under same input of external
loads, which shown a prediction error less than 5%.
A generalized stiffness modelling method has been introduced It should be noted that the work introduced in this paper is an
to build the mathematical equations of error prediction for long- application of MSA method for error prediction of only one arm. Due
reach manipulators with complicated link mechanisms by using to the identical parallelogram mechanism of all EAMA arms, the
the MSA technique. The method has been applied to EAMA robot. stiffness modeling can be performed on the other arms by modify-
The evaluation of stiffness model has been verified through a com- ing the related parameters if needed. After obtaining the deflections
of all five arms, the deflections of the entire manipulate can be
S. Shi et al. / Fusion Engineering and Design 124 (2017) 507–511 511

Fig. 6. Experiment on the entire manipulator system.

evaluated by modifying the D-H parameters of robot kinematical References


model.
In addition, the ideal geometric parameters are applied to the [1] S.S. Shi, Y.T. Song, Y. Cheng, et al., Design and implementation of storage cask
system for EAST articulated inspection arm (AIA) robot, J. Fusion Energy 34 (4)
mathematical model for the complex-section components instead (2015) 711–716.
of actual ones especially for the clevis joints, some uncertain factors [2] S. Shi, Y. Song, Y. Cheng, et al., Conceptual design main progress of EAST
may be ignored and cause a small difference between real system articulated maintenance arm (EAMA) system, Fusion Eng. Des. 104 (2016)
40–45.
and the mathematical model. The deviations between mathemat- [3] J.S. Przemieniechki, Theory of Matrix Structural Analysis, Courier Corporation,
ical results and real behavior was more significant in following 1985.
experimental measurements. This problem can be improved by [4] M. Li, Stiffness Based Trajectory Planning and Feedforward Based Vibration
Suppression Control of Parallel Robot Machines, Acta Universitatis
further studying on the stiffness nature of the uncertain complex- Lappeenrantaensis, 2014.
section components and calibrating the parameters through the [5] H. Yuan, Static and Dynamic Stiffness Analysis of Cable-driven Parallel Robots,
experimental measurements. The compensation strategy based on INSA de Rennes, 2015.
[6] G. Carbone, Grasping in Robotics, Springer, London, 2013.
the proposed method will be carried out focus on the entire manip-
[7] S. Ulrich, J.Z. Sasiadek, Trajectory tracking control of flexible-joint space
ulator system (Fig. 6) with a performance of ±20 mm error (more manipulators, Can. Aeronaut. Space J. 58 (1) (2012) 47–59.
than 100 mm without any compensation). More general deflection [8] M. Li, H. Wu, H. Handroos, Static stiffness modeling of a novel hybrid
control algorithms will be studied for the large remote handling redundant robot machine, Fusion Eng. Des. 86 (9) (2011) 1838–1842.

robots for ITER and the new fusion rectors like DEMO and CFETR.

Acknowledgments

This work was supported by the National “973” Program of China


(Chinese ITER Special Support Project, No. 2014GB101000), fund-
ing of China Scholarship Council and a collaboration framework
between ASIPP China and LUT Finland.

You might also like