You are on page 1of 5

Optimization Design of a Motor Embedded in a

Lightweight Robotic Joint


Haibin Yin1, 2, Yanming Yu1, Junfeng Li1,2
1
Wuhan University of Technology, Wuhan 430070, Hubei, China
2
Key Laboratory of Hubei Province for Digital Manufacture, China
chinaliuyin@whut.edu.cn

Abstract—As the lightweight robotic arm has been applied The driving system of joints is one of the most momentous
widely, it is significant to make lightweight robotic arms much safer parts could determine the whole robot performance. Especially
and lighter. In this study, the axial flux permanent magnet (AFPM) in multi-DOF driving module, output characteristics of each
brushless DC motor with high torque density is proposed and actuator could affect driving reliability and efficiency of overall
designed to apply for the robotic joint based on a prototype. A system. However, most previous researches about lightweight
major design objective is to achieve a lightweight AFPM motor
robotic joints don’t consider the actuators while the actuators
embedded in a robotic joint. This is accomplished by the complex
method, where the structural dimensions and electromagnetic design of robotic joints becomes a new trend. Compared with
parameters are parameterized as design variables. Constraints are conventional motors or actuators, the axial flux permanent
formulated on the basis of kinematic performance, dynamic magnet (AFPM) brushless DC motors have the characteristics of
performance and electromagnetic specifications. Finally, through slim topology, high torque at low speed, small size and light
the comparison between the original and optimal, the optimal weight [10] and has great significance for lightweight and
design has lighter weight and smaller size. miniaturization of robotic joints. In this study, an AFPM motor
embedded in the robotic joint based on a prototype is
Keywords—robotic joint; AFPM motor; optimization design
parameterized and designed. Meanwhile, the complex method is
used to accomplish the optimization design of the AFPM motor.
I. INTRODUCTION Finally, through the comparison between the original and the
With the development of robot technology, the lightweight optimal result, the optimal design has lighter weight and smaller
robotic arms are widely used in military, medical, domestic structure size.
services and the space exploration. At the same time, it faces
many problems varying from low payload-weight ratio, bulky
TABLE I. INITIAL PARAMETERS OF JOINT3
structure, high power consumption and low safety of human-
robot co-existence environment [1]. Therefore, the design of Description Unit Value
lightweight robotic arms needs new methods. The lightweight TR Required torque Nm 10.9
robotic arm consists of links and joints, and the joints are the NR Required speed rpm 30
basic parts of the robotic arm, its performances directly affect the
H1 Total height mm 86
overall system of the robotic arm, so the lightweight of joints
L1 Total length mm 145
becomes more important.
M1 Total mass kg 1.738
Many previous researches have launched on the design of
lightweight robotic joints mainly through the following ways.
The lightweight robotic joints are obtained through joints torque
optimization [2, 3] and trajectory optimization [4]. On the other
hand, some studies focus on the structure and drive trains of the
robotic joints, like Zhou [5] introduced an integrated dimensional
and drive-train design optimization of joints and Park [6]
proposed an approach of modular drive train with integrated
motors and gearboxes design. In addition, some lightweight
materials are utilized to design the lightweight joints, for example
aluminum alloy [7], magnesium alloy [8] and carbon-fiber
material [9]. Fig.1. Prototype of the lightweight robotic arm

978-1-5090-6161-7/17/$31.00 2017
c IEEE 1630
II. PROTOTYPE OF THE LIGHTWEIGHT ARM
In our previous study [11], a five degree-of-freedom (DOF)
lightweight robotic arm has been designed and the overall
structure optimization based on joints and links design has been
implemented to obtain an optimal lightweight robotic arm, as is
shown in Fig. 1. It is designed as a lightweight robotic arm, using
five revolute joints: two DOFs at the shoulder, two DOFs at the
elbow, and one at the wrist. L1 is the total length in axial direction
and H1 is the total height of the robotic joint.
It can be seen from Fig. 1, the joint has too long length in axial
direction, and the space of the joint is not efficiently used.
Therefore, in this study the AFPM motor embedded in the robotic
joint is proposed. Because joint 2, joint 3, joint 5 have the same Fig.2. Basic structure of the AFPM motor.
topological forms, the joint 3 is considered as a representative
joint to design and optimize. And the initial parameters of joint 3 1
Fc =Nc Bci ( Dmo − Dmi ) (4)
are listed in Table I. 2
And the rated torque on the coils is
III. DESIGN AND OPTIMIZATION 1
Tn = Fc ( Dmo − Dmi ) (5)
In this section, the optimization design of the AFPM motor 2
embedded in robotic joint is introduced. And the optimization The main dimensions of the AFPM motor can be determined
problem is generally formulated as an objective function to based on required torque and speed from the above formulas.
minimize the total mass of the AFPM motor in account of the According to the required torque and speed of the robotic joint,
constraint conditions. Meanwhile, the complex method is used the preliminary design of the AFPM motor embedded in the
to accomplish the optimization of the AFPM motor. robotic joint is accomplished, as shown in Fig.3. In this design,
the shell is made up of aluminum alloy, the stator of the AFPM
A. Design of the AFPM motor embedded in the robotic joint motor is glued to the shell through the yoke, the shaft and the
AFPM motors have many topological forms, the slotted one rotors are connected through interference fit. And the harmonic
with double rotors in this study is chosen. Because it is one of the reducer is fixed in the shell through bolt connection. In this
most promising machines in the same volume and can overcome design the harmonic reducer is HFUC-14 with HFUC-2UH
the unilateral magnetic force and reduce the rate of magnetic flux Series Units [13], ĭ=79mm, l=37mm and its ratio ȡ is 100 and its
leakage. The material of the stator core is soft magnetic efficiency Ș is 0.80. The shaft of the AFPM motor is connected
composites. The AFPM motor has 16poles and 12slots, moreover, to the harmonic reducer through tight set screws. The encoder in
the permanent magnet (PM) is sintered NdFeB and is bonded to this design is Omron rotary encoder E6B2-CW26C 360P/R and
the rotor disc along the radial direction, as shown in Fig.2. its main dimension is ĭ40mm×30mm, it is binding pasted to the
rear cover of the shell. L2 is the total length in axial direction and
According to electromagnetic theory, assuming that the
H2 is the total height of the robotic joint.
component of flux density perpendicular to the conductor Bc
does not change significantly, Dmo and Dmi are the outer and
inner diameter of stator, respectively. The back electromotive
force Ec produced by the coils of the motor during operation is
calculated using relative speed of conductor (nn)to the permanent
magnet and the total number of turns of coil Nc as follows
1
Ec = N c nn ( Dmo 2
− Dmi
2
) (1)
4
The dc current flow in conductor i can be calculated as
i = (U n − Ec ) / R (2)
Where Un is the dc rated voltage, R is the total resistance of
the coil, and R can be expressed as Fig.3. Model of the AFPM motor embedded in robotic joint

R = 4 N clρ /(πd 2 ) (3)


TABLE II. DESIGN VARIABLES OF AFPM MOTOR
Where l is the length of a turn coil, d is the diameter of the
conduct. So the total mechanical force on coils Fc can be DV Dmo(mm) Dmi(mm) Ls(mm) hm(mm) d(mm)
expressed as Range [55,75] [30,45] [4,12] [3,6] [0.1,0.5]

2017 12th IEEE Conference on Industrial Electronics and Applications (ICIEA) 1631
B. Parameterized Design Variables min f ( X ) = m1 ( X ) + m2 ( X ) + m3 ( X ) + m4 ( X ) + m5 ( X ) (8)
X
The robotic joint is composed of the AFPM motor, the robotic
joint set and harmonic reducer. In this study, we just optimize D. Implementation of optimization
the AFPM motor. And the mass of each part can be determined
by their main dimensions, and the main dimensions of each part The complex algorithm will be presented in this section. The
complex method is a mathematical programming technic, which
are shown in Fig.3. In this design the structural parameters are
prescribes a systematic procedure for obtaining an optial solution
considered, so the rated voltage Un here is 36V. The minimum to a nonlinear, constrained optimization problem. The detailed
diameter of the shaft dmin is 6mm here in order to fit with the theory of this method was presented by Box [12]. In this study,
harmonic reducer ,į is half the thickness of air gap, į=1mm in we specify several feasible design points (initial vertices of the
this design in account of the assembly and manufacturing complex) in MATLAB, and these points are calculated in ANSYS
requirements. hm is the thickness of the PM, Ls is the thickness Workbench in order to identify if they satisfy the related
of the stator, Ls=Ncd/24 double coil winding. So the design constraints. Then, the objective function value of each design
variables include the outer and inner diameter of stator Dmo, Dmi , points is calculated. The design point with minimum objective
the total number of turns of coil Nc, the thickness of the PM hm function value is defined as the best point Xb, on the contrary, the
and the diameter of the conductor d. For the sake of brevity, the worst is Xw. If a new design point is better than the worst, the
design variables of the motor are described as a vector X= [Dmo, worst design point Xw will be replaced by the new. The new Xnew
Dmi, Ls, hm, d]. According to the calculation, the range of design is α times as far from the centroid Xc of the remaining points as
variables (DV) of the AFPM motor is listed in Table II. the reflection the worse in the centroid.
C. Constrains and objective function m

To ensure the AFPM motor has enough ability to drive the


X c = 1 /(m − 1) ¦ i =1
X i, Xi ≠ Xw

robotic joint, the following design criteria of drive trains are (9)
evaluated and expressed as
X new
= X c +α(X c − X w)
ηρTn ( X ) ≥ S1TR , nn ( X ) ≥ S1ρN R , (6) (10)

Where Tn(X) is the rated torque and n n(X) is the rated speed Where m is the number of the feasible design points, Į is
from the motor, respectively. T R and N R are the required torque reflection coefficient. However, if the new design point is worse
than the worst, we need regulate the algorithm procedure.
and required speed of the joint, respectively, S1 is a safety
coefficient in this design, Ș and ȡ are the efficiency and ratio of X new = X c + α ( X c − X sw )
the harmonic reducer respectively.
(11)
The shaft and the rotors are the main moving parts, and some
Where Xsw: f(Xsw)=max{f(Xi), i=1,2,…,m, XiXw}. This
related constraints on structural strength have to be formulated
optimization design will be terminated until the difference
and the shear strength of the adhesive surface of the stator should
be considered too. between the mass of the best design point and the worst is less
than a defined tolerance İ. The flow chart of the optimization
S 2σ m1 ( X ) ≤ σ y1 , S2σ m 2 ( X ) ≤ σ y 2 , S2τ s ( X ) ≤ [τ ], (7) design approach is shown in Fig. 4.

Where ım1(X) and ım2(X) denote the maximum von-Mises


Equivalent Stress of the shaft and the rotor respectively, ıy1 and
ıy2 are the yield strength of the structural material of the shaft and
the rotor respectively. S2 is the safety coefficient to improve
safety margin. Here these constraints are evaluated through the
Static Structural Analysis in ANSYS Workbench. IJs(X) is the
shear stress of the adhesive surface of the stator and [IJ] is the
allowable shear strength of adhesive.
In this study, the objective of optimization design focus on
minimizing the total mass of the AFPM motor. And the total
mass of the AFPM motor is consist of mi (i=1, 2, 3, 4, 5), where
m1(X) is the mass of permanent magnets, m2(X) is the mass of the
rotors, m3(X) is the mass of stator, m4(X) is the mass of the shaft,
m5(X) is the mass of the coil. Here the optimization problem is
generally formulated as an objective function f(X), which can be
represented by
Fig.4. The flow chart of optimization design

1632 2017 12th IEEE Conference on Industrial Electronics and Applications (ICIEA)
IV. RESULTS AND DISCUSSION 75 45

70
In this design, for the sake of that the initial robotic joint can

Dmi (mm)
Dmo (mm)
40
be replaced completely by the designed joint, the rated output 65

torque and speed of the designed joint should be the same as the 35
60
initial joint. T R=10.9 Nm, N R=30 rpm, the safety coefficient S1
and S2 are set to 1.3 and H1=H2. The results of optimization 55
30
0 200 400 600 800
design are obtained by implementing the flow chart in Fig. 5. In 0 200 400
Iteration Number
600 800
Iteration Number
this study, the solutions to the optimal lightweight AFPM motor (a) (b)
are got through 800 iterations and the120 feasible design points 12 6
are generated to execute the complex algorithm. In the procedure 11
of the algorithm, the convergence error is set to 0.001. The 10
5

hm (mm)
Ls (mm)
convergences of weight of the best point and the worst for the 9
4
AFPM motor are depicted in Fig.5. And in selection procedure, 8
7
the structural dimensions are shown in Fig.6. 3
6

The main parameters of the optimal and initial are listed in 5


0 200 400 600 800
2
0 200 400 600 800
Table III, it can be seen that the optimal mass of the AFPM Iteration Number Iteration Number
motor is 0.46kg with a reduction of 32.4% corresponding to the (c) (d)
initial design. Moreover, the topological structure of AFPM 0.5 1
motor embedded in the robotic joint makes the joint more 0.45 0.8
compact, and the shell of the robotic joint is the shell of motor

Tn(X) (Nm)
d (mm)
too. In this way, the space is made full use and the connection is 0.4 0.6

simpler and more convenient. Compared with the initial joint, the 0.35 0.4

total length of the optimized joint is 108mm with a reduction of 0.3 0.2
25.5% corresponding to the initial joint while the total mass of
0.25 0
the joint is 1.663kg and it is similar to the initial. 0 200 400 600 800 0 200 400 600 800
Iteration Number Iteration Number
The shaft and the rotors are the main moving parts, and some (e) (f)
related constraints on structural strength should be guaranteed. Fig.6. Convergences of dimension

0.7 Here they are evaluated through the Static Structural Analysis
best design point in ANSYS Workbench. In this design the material of the shaft
Weight of Design Point (kg)

0.65
worst design point and rotors is 45 steel, its yield strength is 315-355MPa. The
maximum torque of the shaft is 0.8Nm during the process of
0.6
iteration. From Fig.7 we can see the maximum von-Mises
Equivalent Stress of the shaft is around 137MPa less than the
yield strength. And from Fig.8, the maximum von-Mises
0.55
Equivalent Stress of the rotor is around 1.8MPa far less than the
yield strength during parameters change. The shear strength of
0.5
the adhesive surface of the stator should be guaranteed too. And
it can be calculated by
0.45
0 200 400 600 800
Fs 2Tn ( X )
Iteration Number τs(X ) = =
Fig.5. Convergences of the best and worst design points S πDmo
2
t (12)

TABLE III. OPTIMAL RESULT

Description Initial Optimal


Outer diameter Dmo [mm] 75 55
Inner diameter Dmi [mm] 44 32
Thickness of stator Ls [mm] 12 6
Thickness of PM hm [mm] 6 3
Diameter of conductor d [mm] 0.5 0.28
Total mass [kg] 0.68 0.46 Fig.7. Max Equivalent Stress of the shaft

2017 12th IEEE Conference on Industrial Electronics and Applications (ICIEA) 1633
6

Max Equivalent Stress (Pa)


1.8
x 10 [3] Vinay, G., Himanshu, C and Subir, K., “Dynamics and actuating torque
optimization of planar robots,” Journal of Mechanical Science and
1.7
Technology, 29(7)(2015)2699-2704.
1.6
[4] Hsien-I Lin., “A Fast and Unified Method to Find a Minimum-Jerk Robot
1.5 Joint Trajectory Using Particle Swarm Optimization,” J Intell Robot Syst
1.4 (2014) 75:379-392.
1.3 [5] Zhou, L., Shaoping B., and Michael Rygaard Hansen., “Integrated
1.2
dimensional and drive-train design optimization of a light-weight
0 200 400 600 800 anthropomorphic arm,” Robotics and Autonomous Systems 60(1)
Iteration Number
(2012)113-122.
(a) (b) [6] Park, C., Kyung, J. H., Choi, T. Y., Do, H. M., Kim, B. I., and Lee, S. H.,
“Design of an modular actuation module for a dual arm robot manipulator,
Fig.8. Equivalent Stress of the rotor
9th International Conference on Ubiquitous Robots and Ambient
Intelligence,” November, 2012, pp. 254–259.
[7] Guo, D., Mei, T., Luo, M., Feng, B and Zhao, J., “Dynamics Analysis and
0.07
LightHeight Design of Manipulator of Elderly Serve Robot,” J China
Mechanical Engineering, 23(10)(2012) 1146-1150.
Shear Stree (MPa)

0.06
[8] Wang, Z., Jia, L and Du, W., “Analysis of The Application of Magnesium
0.05 Alloy Materials in The Robot Lightweight,” J Advanced Materials Industry,
0.04 12(7)(2016) 14-17.
0.03 [9] Schrock, P., Farelo, F., Alqasemi, R., and Dubey, R., “Design, simulation
and testing of a new modular wheelchair mounted robotic arm to perform
0.02
activities of daily living,” IEEE International Conference on Rehabilitation
0.01
0 200 400 600 800 Robotics, June 2009, pp. 518–523.
Iteration Number
[10] Jung, S and Se-Hyun, J., “Design of Axial Flux Permanent Magnet
(a) (b) Brushless DC Motor for Robot Joint Module,” The 2010 International
Power Electronic Conference, IEEE, pp. 1336-1340. 2010.
Fig.9. Shear Stress of the adhesive surface of the stator [11] Haibin Y., Shansheng H., Mingchang H and Junfeng L., “An Overall
Where Fs is the shear force on the adhesive surface, S is the Structure for Light-Weight Robotic Arm”, The 11th IEEE Conference on
Industrial Electronics and Applications, June 2016.
area of the adhesive surface and t is the width of the yoke, here t
[12] Box, M. J., “A New Method of Constrained Optimization and a
is 4mm. The shear stress of the adhesive surface is shown in Fig.9. Comparison With Other Methods,” The Computer Journal, vol. 8, pp. 42–
The adhesive is AB glue in this design, its allowable shear 52, 1965.
strength [IJ] is more than 25MPa and it is far larger than the [13] Harmonic Drive. Engineering Data SHD-2SH Units. Harmonic Drive AG.
maximum shear stress of the adhesive surface of the stator Limburg/Lahn, Germany. [Online] Available: http://harmonicdrive.de/
0.07MPa from Fig.9. So the strength meets the requirement. produkte/media/catalog/category/2014_12_ED_1019647_SHD_2SH.

V. CONCLUTION
In this study, an AFPM motor embedded in the robotic joint is
proposed and designed based on a prototype. Meanwhile, the
optimization of the AFPM motor is accomplished by means of
the complex method. Finally, through the comparisons, the
optimized joint has lighter weight and smaller size. Therefore,
the applications of AFPM motors embedded in robotic joints
have great significance for lightweight and miniaturization of
robotic joints, and it will be greater developed and applied in the
future. However, in order to demonstrate the advantages more
accurately, the experimental prototypes will be trial-produced and
tested in our future research.

REFERENCES
[1] Hirzinger, G., Albu-Schaffer, A., Hahnle, M., Schaefer, I., and Sporer,
N., “On a new generation of torque controlled light-weight robots,”IEEE
International Conference on Robotics and Automation, 2001, pp. 3356-
3363.
[2] Gao, X., Du, M., Wu, H., Sun, H., Jia, Q., Chen, G., and Wang, Y., “A
Joint Torque Optimization Method for Space Manipulators Considering the
In-Orbit Motion Reliability,” Jourmal of Astronautics, 37(7) (2016) 784–
794.

1634 2017 12th IEEE Conference on Industrial Electronics and Applications (ICIEA)

You might also like