Al-khwarizmi
Engineering 
  Journal 

Al-Khwarizmi Engineering Journal, Vol.4  , No.1 , pp 17- 26 (2008 )

Minimizing error in robot arm based on design optimization for high 
stiffness to weight ratio.

Ahmed Abdul Hussain Ali
University of Baghdad

College of Engineering, Mech. department 

(Received 8 May 2007; accepted 3 December 2007)

Abstract:
In this work the effect of choosing tri-circular tube section had been addressed to minimize the 

end effector’s error, a comparison had been made between the tri-tube section and the traditional 
square cross section for a robot arm, the study shows that for the same weight of square section and tri-
tube section the error may be reduced by about 33%.
A program had been built up by the use of MathCAD software to calculate the minimum weight of a 
square section robot arm that could with stand a given pay load and gives a minimum deflection. The 
second part of the program makes an optimization process for the dimension of the cross section and 
gives the dimensions of the tri-circular tube cross section that have the same weight of the 
corresponding square section but with less deflection.

Key word: robot arm stiffness, flexible manipulator, robot structure analysis, flexible link robot. 

Introduction:
The links of serial manipulators are 

usually over designed in order to be able to 
support the subsequent links on 
the chain and the pay load to be manipulated. 
However, increasing the size of the links 
unnecessarily requires the use of larger 
actuators resulting in higher power 
requirements. Optimum robot design has been 
addressed by many researchers as found in the 
open literature; Shiakolas and koladiye [1] 
discuss the application and comparison of the 
evolutionary techniques for optimum design of 
serial link robot manipulators based on task 
specifications. The objective function 
minimizes the required torque for a defined 
motion subjected to various constraints which 
considering kinematics, dynamic and structural 
conditions. The design variables examined are 
the link parameters and the link cross sectional 
characteristics, the developed environment was 
employed in optimizing the design variables for 
a SCARA and an articulated 3-DOF PUMA 
type manipulators. In the work developed by 

Marcus Pettersson et al. [2] an optimization 
problem are formulated to minimize the weight 
of the gearboxes, by choosing different discrete 
gear boxes, and changing the lengths of the 
arms continuously, subjected to a few 
requirements on acceleration capability reach 
and pay load capacity. Analysis of stiffness of 
manipulator link can be found in Abdel malek, 
K. and Paul, B.[3] where aspects of the 
structural design of the manipulator arm are 
presented. Prismatic joints of manipulator arm 
are based upon a cross sectional design of the 
links that provides a high stiffness to weight 
ratio compared with a hollow round cross-
section. 

The case that we study in this work is 
the robot that consists of three arms as shown 
in fig. (1). Where the first arm is vertical and 
the second and third arm are horizontal this 
gives the maximum reach (completely stretched 
out) for the robot arm and will yield the 
maximum deflection for the robot.



Ahmed Abdul Hussain Ali                Al-khwarizmi Engineering Journal, Vol.4 , No. 1 PP 17-26 (2008)

18

Prismatic joints:
Most manipulator link cross- section are 

either hollow round or hollow rectangular. 
Hollow links provide convenient conduits for 
electric power and communication cables, 
hoses, power transmission members, etc. 
Rivin[4] has studied the influence of cross-
sections on the deflections both in bending and 
torsion. He had compared hollow square with 
hollow circular cross sections. Rivin states that 
a square cross section can provide a 69 to 84 
percent increase in bending stiffness over a 
circular hollow cross section with only a 27 
percent increase in weight. 

In this paper a different cross-section is 
introduced, consisting of three tubes centered 
on the vertices of an equilateral triangle. This 
cross section is referred to as a tri-tube 
configuration the hollow square link will be 
referred to as a uni-tube configuration, as 
shown in fig. (2).

Deflection due to pure bending:
Links with an open end manipulator are 

normally modeled as cantilevers. Consider a 
simple cantilever with solid or hollow cross –
section as shown in fig.(3).To study the 
proposed cross –section, we use the following 
equations for moments of inertia (2nd moment 
of area) about any diametrical axis through the 
centroid of area. 

Uni –tube:
For the uni- tube depicted fig.(2,a) the moment 
of inertia about the neutral axis is 

tBb
bB

I tubeuni 2,
12

44






Where B, b is the outer and inner sides, 
respectively for the uni-tube construction, t is 
the thickness of the uni-tube and the area is 

22 bBA tubeuni 

Tri-tube:
For the tri-tube depicted in fig.(2,b) the 
moment of inertia about the neutral axis is 

   

 
2

22

2244

60sin
3

1

4
2

60sin
3

2

464

3





















hdD

hdDdDI tubetri





  

tDd 2

Where D and d are the out side and inside 
diameters, respectively, for each tube on the 
equilateral triangle, t is the thickness of each 
tube in the tri – tube construction. 
The area of each tube is 

  tDddDA tubetri 2,
4

22 


To demonstrate the deflection due to loading, 
consider the third arm beam depicted in 
fig.(3,c).













83

1
4

33
3

33

3
3

LqLW

EI


gmWA
L

ALg

L

gM
q 


 ,



Where M is the mass of each arm, m is the 
mass of the gear box and the mass of the load 
to be manipulated at the end of the arm, q is the 
weight per unit length of the beam, W is the 
load in Newton, g is the gravitational 
acceleration and L is the length of the arm, A is 
the cross sectional area of the beam,  is the 
specific density.
To get the reactions (force and moment) at the 
fixed end of the third arm we equate the 
summation of forces and moment to zero i.e.

  33330 LqWFFy

  33
2

33
3

2
0 LW

Lq
MOM

The same thing may be said for the second arm 
(fig.3-b) taking in to account the effect of 
moment in calculating the deflection i.e.

 















283

1
3

23
4

22
3

223

2
2

LMOLqLWF

EI


The reactions at the fixed end will be 

32222 FWLqF 

32322

2
22

2
2

MOLFLW
Lq

MO 

For the first arm (fig.3-a) we assume that the 
deflection at the free end is due to bending 
moment only and the effect of compressive 
loads on the whole deflection is neglected 
therefore 

1

2
11

1
2EI

LMO




Ahmed Abdul Hussain Ali                Al-khwarizmi Engineering Journal, Vol.4 , No. 1 PP 17-26 (2008)

19

The total deflection at the end effecter of the 
robot manipulator arm will be 

  21
2

23  total
The sequence of analysis in this work is to 
calculate the weight of the lightest structure 
that has a square hollow section and with stand 
the given loading condition this may be 
achieved by letting the stress in each arm 
reaches the maximum allowable stress to avoid 
failure of the structure, the equation for 
calculating the stress in the third arm may be 
written as

12

2/*
4

3
4

3

33
3

bB

BMO

I

YM




By letting the stress equal the allowable stress 
and assuming the thickness of the tube walls to 
be 2mm we may found the dimension of the 
third arm, this had been done by the aid of a 
program built up using MathCAD software. 
The stress in the second arm may also be 
calculated in the same way i.e.

  

12

2/*
4

2
4

2

22
2

bB

BMO


   

The dimension of the first arm fig.(3.a) is 
calculated by equating the maximum stress 
induced in it with the maximum allowable, this 
maximum stress is found by the Rankine-
Gorden formula [5] which is a combination of 
the Euler and crushing loads for a strut

ceR FFF

111


For very short strut eF is very large, 
eF

1
can 

therefore be neglected and cR FF  , for very 

long struts  eF is very small and 
eF

1
is very 

large so that 
cF

1
can be neglected. 

Thus eR FF  . The Rankine formula is 
therefore valid for extreme values of 
slenderness ratios. It is also found to be fairly 
accurate for the intermediate values. Thus, re –
writing the formula in terms of stresses

)/(1

111

111

eY

Y

Ye

Ye
R

Ye

Ye

YeR

YeR AAA



























For a strut with one end free and the other fixed 

2

2

4 L

IE
Fe


 and   

AL

IE
e 2

2

4


 

The crushing load on the first arm is 

121 WFFFc 

A

Fc
Y 

The final stress  1 on the first arm is thus the 
sum of the direct stress calculated by Rankine 
formula and that due to bending generated by 
the exerted moment  2MO as was explained in 
figure (3-a and b)

  1
12

1

2/*

/! I

BMO

eY

Y
bendingR 







From this equation we may find the dimension 
of the first arm. After knowing the dimension 
of each arm the weight of each arm may be 
found and also the total weight of the 
manipulator structure will be determined. The 
next step in the analysis is to input those 
information to the program to began the 
process of changing the dimension of the cross 
section to minimize the total error (deflection) 
at the end of the robot arm this process gives 
many generations of the dimensions of the arm 
cross section which satisfies the conditions 
specified for the maximum and minimum error 
allowed at the end effecter and also the 
permissible increase in the weight of the robot 
structure specified from us, from between all 
those generation the program select the best 
generation or probability that gives the lightest 
weight and the less deformation. The next step 
in the analysis is to input the new weight of the 
robot arms to the optimization process for the 
tri –tube cross section shown in figure(2-b) and 
trying to find the best dimensions that gives the 
highest moment of inertia for the cross section 
so as to minimize the deflection in each arm 
and also the total deflection of the robot at the 
end effecter, i.e. the mass of uni –tube section 



Ahmed Abdul Hussain Ali                Al-khwarizmi Engineering Journal, Vol.4 , No. 1 PP 17-26 (2008)

20

found by the program should be equal to the 
mass of the tri –tube section which is equal to   
























)
4

3

(2)(
4

3

2

222
3

3 





D

RSdDL
VM

2/
3

sin
3

2
DhR 



Where  is the density of robot arm metal, R
is the radius of the two flanges (stiffeners) 
welded at each end, S is the thickness of the 
flanges and h is the distance between the 
vertices of the equilateral triangle. 
The optimization problem is defined as follow 

tDDh

Iimize

MM

tubetri

tubeunitubetri

2,

)(max









In the optimization problem the thickness of the 
tubes )(t and the thickness of the flanges )(S
are assumed to be 2mm.
The results of the optimization problem showed 
that the tri –tube section that have the same 
weight (mass) of a uni –tube may improve the 
stiffness of the robot and minimize the total 
deflection in about 33%, this results means that 
we may construct a robot having tri –tube 
section which is less in weight from that of uni 
–tube section and both of them having the same 
end effecter deformation. 
Results:
In order to verify the analysis of the previous 
section a run had been done which has the 
following characteristics for the robot arm 

29 /10200 mNE 
26 /10120 mNall  , 

2sec/81.9 mg  , 
mLmLmL 4.,45.,5. 321  , 

3/7850 mkg , m0022.0max  , 
m0005.0min  , mT 002.0 ( tube thickness 

), mS 002.0 (stiffener thickness), 
kgm 5.91  (mass of the first gear box), 
kgm 4.42  (mass of the second gear box), 

kgm 503  (manipulated mass). The available 
gear boxes for the application are given in the 
list of table 1
The results of the program shows that for the 
given configuration the minimum weight for 
the structure of the uni –tube robot is 

(Wmin=19.986 N), the robot with such structural 
weight could manipulate the load with out 
failure because the stress in each arm is less 
than or equal to the allowable stress, but the 
deflection of the end point effecter is very 
large. The iteration process for increasing the 
dimension of the section to minimize the 
deflection and letting it be within the range 
(0.0005< <0.0022) shows that there are 22 
generation all of which has a deflection 
(0.0005< <0.0022) and also a weight 
(W<Factor * Wmin ) the permissible weight 
factor (Fac.) for increasing the weight was 
chosen to be (Fac.=1.35). The dimensions of 
the inner side of the uni –tube section for the 22 
generation are shown in fig.(4). The relation 
between the total deflection at the end -effecter 
and number of generation is shown in fig.(5). 
The relation between the new weight of the 
robot structure and its generation is shown in 
fig.(6). The program chooses the best 
generation which has the less variable 
(variable= weight*deflection), the relation 
between the variable and the generation is 
shown in fig(7) it is obvious that the generation 
no. 16 has the minimum value, the dimensions 
of the section for that generation are B1
=0.07464m, B2 =0.06397m, B3=0.05657m and 
has a deflection total =1.87125*10

-3  m and a 
total structural weight Wtotal =26.917N. Those 
results are the input for the next step in the 
program for calculating the dimensions of the 
tri –tube section in which an optimization 
problem where solved to maximize the moment 
of inertia for the section in terms of the 
dimensions h and D, the results of the program 
are shown in table (2)
The total deflection for the tri –tube 
configuration 

  21
2

23)(  tubeTriTotal =1.31462*10
-3 m
The deflection for the tri –tube configuration is 
less than that for the uni –tube which was found 
to be ( 310*8725.1  tubeuni ) this results 
shows that the tri –tube section reduces the 
deflection in a bout 29.7% from that of uni –
tube section, the result may be improved to 
reach a value of 33.38% if we change the 
weight factor (Fac.) to make it equal to (1.3) on 
the other hand if (Fac.) is increased to(1.65) the 



Ahmed Abdul Hussain Ali                Al-khwarizmi Engineering Journal, Vol.4 , No. 1 PP 17-26 (2008)

21

improvement in deflection would be less and 
equal to (6.02%) those results for iteration and 
optimization are shown in table(3)
Another interesting feature in the design of the 
uni –tube section is that if the weight factor 
(Fac.) was chosen to be 1.3 and the range of 
deflection is limited to  <0.0022 we would 
find only five generation which satisfies the 
previous mentioned configuration and if (Fac.) 
was changed to 1.65 and  < 0.00104 we 
would find only 4 generation these results 
which are shown in table (4) shows the band of 
limits of the design of robot in other word we 
can not find a robot with a weight factor less 
than (Fac.=1.3) and has a deflection less than 
0.0022m or we cannot find a robot with a 
weight factor less than (Fac.=1.65) and had a 
deflection less than 0.00104m those results of 
iteration are shown in table(4).
A flow chart of the program used is shown in 
fig.(8).

Conclusions:
This paper presents a method for optimization 
of robot design in the conceptual design stage. 
The robot is modeled in the MathCAD package 
and the optimization problem is formulated as 
to determine the dimension of robot arm in 
order to minimize the weight and maximize 
stiffness this formulation can be interpreted as 
to design the cheapest possible robot that will 
still meets the design demands. The 
optimization method showed good capability in 
finding an optimum set of dimension of the arm 
of robot manipulator with three degree of 
freedom.
The optimization method shows that the tri –
tube is superior to uni –tube section in 
minimizing deflection in about 33%.
The presented work provides a good support 
for conceptual robot design. 

Fig. (1) robot configuration
(b) (a)

Fig.(2) (a) uni-tube configuration 
           (b) tri-tube configuration.

   

2q

2

3L2L

1

   

3q

1L

m3
m1 m2

3



Ahmed Abdul Hussain Ali                Al-khwarizmi Engineering Journal, Vol.4 , No. 1 PP 17-26 (2008)

22

Fig.(3) modulation of manipulator links

Table 1 torque–mass relation for available gear 
boxes [2]

Out put torques (N.m) Mass
(kg)

101
231
572

1088
1499
2176
4361
6135

2.5
4.4
9.5
12.7
18
28
47
69

Table (2) dimension of the tri –tube section
arm
No.

h (m) D (m) 
*10-3

Itri –tube(m
4)

*10-7
Deflection 
( )
m *10-4

3
2
1

0.10028
0.10440
0.10440

7.52652
9.22085
11.92

1.75056
2.48207
3.42024

8.09087
7.35704
3.0044

Table (3) results for iteration and optimization 
problem

Fac.
tubeTri

(m)*10-3
tubeUni 

(m)*10-3

Improvement
In deflection

1.3
1.35
1.4
1.45
1.5
1.55
1.6
1.65

1.39973
1.31462
1.2189
1.16048
1.09686
1.04631
0.985197
.948732

2.10114
1.87125
1.6237
1.48623
1.3437
1.22646
1.08422
1.00955

33.38%
29.7 %
24.9 %
21.9 %
18.37%
14.68%
9.13 %
6.02 %

Table (4) limits of deflection-weight factor for 
robot design

Fac.  max (m) generations
1.3
1.35
1.4
1.45
1.5
1.55
1.6
1.65

0.0022
0.0019
0.0017
0.0015
0.0014
0.0013
0.0011
0.00104

5
2
3
2
3
0
2
4



Ahmed Abdul Hussain Ali                Al-khwarizmi Engineering Journal, Vol.4 , No. 1 PP 17-26 (2008)

23

In
n
er

-s
id

e 
d
im

en
si

o
n 

 (
B

) 
fo

r 
 u

n
i-

tu
be

 s
ec

ti
o
n 

in
 

(m
m

)

1 3 5 7 9 11 13 15 17 19 21
0.04

0.05

0.06

0.07

0.08

0.09
B1
B2
B3

B1
B2
B3

generatonsgenerations
fig.(4) correlation between inner side dimension and no. of generations.

T
o
ta

l 
de

fl
ec

ti
o
n 

total

(m
m

)

0 2 4 6 8 10 12 14 16 18 20 22
0.0017

0.00176

0.00182

0.00188

0.00194

0.002

0.00206

0.00212

0.00218

0.00224

0.0023

generations
generations

Fig.(5) correlation between the deflection and no. of generations.



Ahmed Abdul Hussain Ali                Al-khwarizmi Engineering Journal, Vol.4 , No. 1 PP 17-26 (2008)

24

T
o
ta

l 
w

ei
g
h
t 

 (
N

) 
  
  
  
  
 

0 2 4 6 8 10 12 14 16 18 20 22
25.5

26

26.5

27

generations

generations
Fig.(6) correlation between robot weight and no. of generations.

  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
 V

ar
ia

b
le

 (
V

A
R

.)
  
  
 

0 2 4 6 8 10 12 14 16 18 20 22
0.05

0.055

0.06

generations
Fig.(7) correlation between (Variable) and no. of generations.



Ahmed Abdul Hussain Ali                Al-khwarizmi Engineering Journal, Vol.4 , No. 1 PP 17-26 (2008)

25

References:
[1]. P.S. shiakolas, D. koladiya and J.kebrle 
“optimum robot design based on task 
specifications using evolutionary techniques 
and kinematic, dynamic, and structural 
constraints”, international journal of inverse 
problems in engineering, volume 10, number 4, 
pp 359-375, 2002.
[2]. Marcus P., Peter K., Xiaolong F.A., Johan 
A. and doniel W. “industrial robot design 
optimization in the conceptual design phase”, 
IEEE –mec. Of robot, 2004.

[3]. Abdel –Malek K. and Paul B., “criteria for 
the design of manipulator arms for high 
stiffness to weight ratio”, SME journal of 
manufacturing systems, Vol.17, No.3, pp.209-
220.
[4]. E.I. Rivin, mechanical design of robots, 
1988, Mc Grow –Hill, Inc, New York.
[5]. E.j. Hearn “Mechanics of materials”, 
Pergamon press, 1977.

Fig.(8) flow chart of the program built-up by the use of MathCAD soft ware

Defining: E, all , m1,m2,m3,  ,T,S, 

Calculation of  B3, B2, B1 and 
Wmin=q3*L3+q2*L2+q1*L1

For B1 =B1+0.001
For B2=B2+0.001
For B3=B3+0.001

Calculation of: 

321 ,,  ,  
2

3
2

23  total ,Wrobt

If

maxmin   total

If
Wrobot>Fac.*Wmin

Saving 
generations

Finding minimum 

(VAR=Wrobot total )

1

1

Optimization for tri-tube 
section

h>D , Wtii-tube=Wuni-tube, D  2T
Maximize (I), find h,D

Finding: 123 ,, 

  21
2

23,  tubetritotal

No

No

Yes

Yes



Ahmed Abdul Hussain Ali                Al-khwarizmi Engineering Journal, Vol.4 , No. 1 PP 17-26 (2008)

26

على أساس التصمیم األمثل ولنسبة جساءة الى وزن ) الروبوت(تقلیل الخطاء في الذراع األلي 
  .عالیة
  

  أحمد عبد الحسین علي. د
  كلیة الھندسة/  جامعة بغداد

  قسم المیكانیك 
  

  :الخالصة
لطرفي في الذراع األلي ، تم في ھذا البحث تم دراسة تاثیر استخدام المقطع الثالثي األنابیب الدائریھ ألجل تقلیل الخطاء ا  

اجراء مقارنھ بین المقطع الثالثي األنابیب و المقطع المربع التقلیدي للذراع األلي ، الدراسھ بینت بانھ لكال الذراعین ذات المقطع 
  . ٣٣%الثالثي و المربع والذان لھما نفس الوزن ممكن تقلیل الخطاء بحدود 

ساب أقل وزن للذراع اآللي ذي المقطع المربع الذي یمكن أن یتحمل األوزان المسلطھ لح  MathCADتم كتابة برنامج باستخدام 
  .ویعطي أقل تشوه

ي الجزء الثاني من البرنامج یقوم بعملیة األمثلیھ ألجل ایجاد ابعاد المقطع ذي األنابیب الثالثیھ الدائریھ والذي لھ نفس وزن الذراع ذ
  .المربعالمقطع المربع ولھ تشوه أقل من نضیره 














all
s











r











min
max
,
d
d











2
q











t
B
b
b
B
I
tube
uni
2
,
12
4
4
-
=
-
=
-











2
2
b
B
A
tube
uni
-
=
-











(
)
(
)
(
)
2
2
2
2
2
4
4
60
sin
3
1
4
2
60
sin
3
2
4
64
3
÷
ø
ö
ç
è
æ
-
+
÷
ø
ö
ç
è
æ
-
+
-
=
-
h
d
D
h
d
D
d
D
I
tube
tri
p
p
p











t
D
d
2
-
=











(
)
t
D
d
d
D
A
tube
tri
2
,
4
2
2
-
=
-
=
-
p











÷
÷
ø
ö
ç
ç
è
æ
+
=
8
3
1
4
3
3
3
3
3
3
3
L
q
L
W
EI
d











g
m
W
A
L
A
L
g
L
g
M
q
=
=
=
´
=
,
g
r











g











å
+
=
=
3
3
3
3
0
L
q
W
F
F
y











å
+
=
=
3
3
2
3
3
3
2
0
L
W
L
q
MO
M











(
)
÷
÷
ø
ö
ç
ç
è
æ
+
+
+
=
2
8
3
1
3
2
3
4
2
2
3
2
2
3
2
2
L
MO
L
q
L
W
F
EI
d











3
2
2
2
2
F
W
L
q
F
+
+
=











3
2
3
2
2
2
2
2
2
2
MO
L
F
L
W
L
q
MO
+
+
+
=











1
2
1
1
1
2
EI
L
MO
=
d











(
)
2
1
2
2
3
d
d
d
d
+
+
=
total











12
2
/
*
4
3
4
3
3
3
3
b
B
B
MO
I
Y
M
-
=
=
s











12
2
/
*
4
2
4
2
2
2
2
b
B
B
MO
-
=
s











c
e
R
F
F
F
1
1
1
+
=











e
F











e
F
1











c
R
F
F
=











e
F











e
F
1











c
F
1











e
R
F
F
=











)
/
(
1
1
1
1
1
1
1
e
Y
Y
Y
e
Y
e
R
Y
e
Y
e
Y
e
R
Y
e
R
A
A
A
s
s
s
s
s
s
s
s
s
s
s
s
s
s
s
s
s
s
+
=
+
=
+
=
+
=
+
=











2
2
4
L
I
E
F
e
p
=











A
L
I
E
e
2
2
4
p
s
=











1
2
1
W
F
F
F
c
+
=
=











A
F
c
Y
=
s











(
)
1
s











(
)
2
MO











(
)
1
1
2
1
2
/
*
/
!
I
B
MO
e
Y
Y
bending
R
+
+
=
+
=
\
s
s
s
s
s
s











ú
ú
ú
ú
û
ù
ê
ê
ê
ê
ë
é
-
+
-
=
=
)
4
3
(
2
)
(
4
3
2
2
2
2
3
3
p
p
p
r
r
D
R
S
d
D
L
V
M











2
/
3
sin
3
2
D
h
R
+
=
p











r











R











S











h











t
D
D
h
I
imize
M
M
tube
tri
tube
uni
tube
tri
2
,
)
(
max
³
³
=
-
-
-











)
(
t











)
(
S











2
9
/
10
200
m
N
E
´
=











2
6
/
10
120
m
N
all
´
=
s











2
sec
/
81
.
9
m
g
=











m
L
m
L
m
L
4
.
,
45
.
,
5
.
3
2
1
=
=
=











3
/
7850
m
kg
=
r











m
0022
.
0
max
=
d











m
0005
.
0
min
=
d











m
T
002
.
0
=











m
S
002
.
0
=











kg
m
5
.
9
1
=











kg
m
4
.
4
2
=











kg
m
50
3
=











d











d











total
d











(
)
2
1
2
2
3
)
(
d
d
d
d
+
+
=
-
tube
Tri
Total











3
10
*
8725
.
1
-
-
=
tube
uni
d











d











d











3
L











d











tube
Tri
-
d











tube
Uni
-
d











d











1
3
5
7
9
11
13
15
17
19
21
0.04
0.05
0.06
0.07
0.08
0.09
B1
B2
B3
B1
B2
B3
fig.(4) corrolation betwen inner side dimension and no. of generations
generatons
inner-side dimension (B) for uni-tube section in (mm)











total
d











0
2
4
6
8
10
12
14
16
18
20
22
0.0017
0.00176
0.00182
0.00188
0.00194
0.002
0.00206
0.00212
0.00218
0.00224
0.0023
fig.(5) corrolation betwenthe deflection and no. of generations
generations
total defflection (mm)











0
2
4
6
8
10
12
14
16
18
20
22
25.5
26
26.5
27
Fig.(6) corrolation betwen robot wight and no.of generations
generations
total weight (N)











0
2
4
6
8
10
12
14
16
18
20
22
0.05
0.055
0.06
generations
vareant(VAR)











2
L











3
q











1
L











1
2
3
,
,
d
d
d











(
)
2
1
2
2
3
,
d
d
d
d
+
+
=
-
tube
tri
total











³











total
d
´











max
min
d
d
d
á
á
total











3
2
1
,
,
d
d
d











(
)
2
3
2
2
3
d
d
d
d
+
+
=
total


















































Al-khwarizmi 

Ahmed abdul hussain ali
               Al-khwarizmi Engineering Journal, Vol.4 , No. 1 PP 17-26 (2008)

	Al-khwarizmi 
Engineering   Journal 
	Al-Khwarizmi Engineering Journal, Vol.4  , No.1 , pp 17- 26 (2008 )
	



Minimizing error in robot arm based on design optimization for high stiffness to weight ratio.
Ahmed abdul hussain ali

University of Baghdad
 College of engineering, Mech. department 

(Received 8 May 2007; accepted 3 December 2007)


Abstract:


In this work the effect of choosing tri-circular tube section had been addressed to minimize the end effector’s error, a comparison had been made between the tri-tube section and the traditional square cross section for a robot arm, the study shows that for the same weight of square section and tri-tube section the error may be reduced by about 33%.

A program had been built up by the use of MathCAD software to calculate the minimum weight of a square section robot arm that could with stand a given pay load and gives a minimum deflection. The second part of the program makes an optimization process for the dimension of the cross section and gives the dimensions of the tri-circular tube cross section that have the same weight of the corresponding square section but with less deflection.

Key word: robot arm stiffness, flexible manipulator, robot structure analysis, flexible link robot. 

Introduction:

The links of serial manipulators are usually over designed in order to be able to support the subsequent links on 

the chain and the pay load to be manipulated. However, increasing the size of the links unnecessarily requires the use of larger actuators resulting in higher power requirements. Optimum robot design has been addressed by many researchers as found in the open literature; Shiakolas and koladiye [1] discuss the application and comparison of the evolutionary techniques for optimum design of serial link robot manipulators based on task specifications. The objective function minimizes the required torque for a defined motion subjected to various constraints which considering kinematics, dynamic and structural conditions. The design variables examined are the link parameters and the link cross sectional characteristics, the developed environment was employed in optimizing the design variables for a SCARA and an articulated 3-DOF PUMA type manipulators. In the work developed by marcus Pettersson et al. [2] an optimization problem are formulated to minimize the weight of the gearboxes, by choosing different discrete gear boxes, and changing the lengths of the arms continuously, subjected to a few requirements on acceleration capability reach and pay load capacity. Analysis of stiffness of manipulator link can be found in Abdel malek, K. and Paul, B.[3] where aspects of the structural design of the manipulator arm are presented. Prismatic joints of manipulator arm are based upon a cross sectional design of the links that provides a high stiffness to weight ratio compared with a hollow round cross-section. 


The case that we study in this work is the robot that consists of three arms as shown in fig. (1). Where the first arm is vertical and the second and third arm are horizontal this gives the maximum reach (completely stretched out) for the robot arm and will yield the maximum deflection for the robot.

Prismatic joints:


Most manipulator link cross- section are either hollow round or hollow rectangular. Hollow links provide convenient conduits for electric power and communication cables, hoses, power transmission members, etc. Rivin[4] has studied the influence of cross-sections on the deflections both in bending and torsion. He had compared hollow square with hollow circular cross sections. Rivin states that a square cross section can provide a 69 to 84 percent increase in bending stiffness over a circular hollow cross section with only a 27 percent increase in weight. 


In this paper a different cross-section is introduced, consisting of three tubes centered on the vertices of an equilateral triangle. This cross section is referred to as a tri-tube configuration the hollow square link will be referred to as a uni-tube configuration, as shown in fig. (2).

Deflection due to pure bending:


Links with an open end manipulator are normally modeled as cantilevers. Consider a simple cantilever with solid or hollow cross – section as shown in fig.(3).To study the proposed cross –section, we use the following equations for moments of inertia (2nd moment of area) about any diametrical axis through the centroid of area. 
Uni –tube:

For the uni- tube depicted fig.(2,a) the moment of inertia about the neutral axis is 


 

Where B, b is the outer and inner sides, respectively for the uni-tube construction, t is the thickness of the uni-tube and the area is 



Tri-tube:

For the tri-tube depicted in fig.(2,b) the moment of inertia about the neutral axis is 


  



Where D and d are the out side and inside diameters, respectively, for each tube on the equilateral triangle, t is the thickness of each tube in the tri – tube construction. 

The area of each tube is 



To demonstrate the deflection due to loading, consider the third arm beam depicted in fig.(3,c).





Where M is the mass of each arm, m is the mass of the gear box and the mass of the load to be manipulated at the end of the arm, q is the weight per unit length of the beam, W is the load in Newton, g is the gravitational acceleration and L is the length of the arm, A is the cross sectional area of the beam, 
 is the specific density.

To get the reactions (force and moment) at the fixed end of the third arm we equate the summation of forces and moment to zero i.e.





The same thing may be said for the second arm (fig.3-b) taking in to account the effect of moment in calculating the deflection i.e.



The reactions at the fixed end will be 





For the first arm (fig.3-a) we assume that the deflection at the free end is due to bending moment only and the effect of compressive loads on the whole deflection is neglected therefore 



The total deflection at the end effecter of the robot manipulator arm will be 



The sequence of analysis in this work is to calculate the weight of the lightest structure that has a square hollow section and with stand the given loading condition this may be achieved by letting the stress in each arm reaches the maximum allowable stress to avoid failure of the structure, the equation for calculating the stress in the third arm may be written as



By letting the stress equal the allowable stress and assuming the thickness of the tube walls to be 2mm we may found the dimension of the third arm, this had been done by the aid of a program built up using MathCAD software. The stress in the second arm may also be calculated in the same way i.e.

  
  

The dimension of the first arm fig.(3.a) is calculated by equating the maximum stress induced in it with the maximum allowable, this maximum stress is found by the Rankine-Gorden formula [5] which is a combination of the Euler and crushing loads for a strut



For very short strut 
 is very large, 
 can therefore be neglected and
, for very long struts  
 is very small and 
 is very large so that 
can be neglected. Thus
. The Rankine formula is therefore valid for extreme values of slenderness ratios. It is also found to be fairly accurate for the intermediate values. Thus, re –writing the formula in terms of stresses



For a strut with one end free and the other fixed 


 and   

The crushing load on the first arm is 





The final stress 
 on the first arm is thus the sum of the direct stress calculated by Rankine formula and that due to bending generated by the exerted moment 
 as was explained in figure (3-a and b)



From this equation we may find the dimension of the first arm. After knowing the dimension of each arm the weight of each arm may be found and also the total weight of the manipulator structure will be determined. The next step in the analysis is to input those information to the program to began the process of changing the dimension of the cross section to minimize the total error (deflection) at the end of the robot arm this process gives many generations of the dimensions of the arm cross section which satisfies the conditions specified for the maximum and minimum error allowed at the end effecter and also the permissible increase in the weight of the robot structure specified from us, from between all those generation the program select the best generation or probability that gives the lightest weight and the less deformation. The next step in the analysis is to input the new weight of the robot arms to the optimization process for the tri –tube cross section shown in figure(2-b) and trying to find the best dimensions that gives the highest moment of inertia for the cross section so as to minimize the deflection in each arm and also the total deflection of the robot at the end effecter, i.e. the mass of uni –tube section found by the program should be equal to the mass of the tri –tube section which is equal to   





Where 
 is the density of robot arm metal, 
 is the radius of the two flanges (stiffeners) welded at each end, 
 is the thickness of the flanges and 
 is the distance between the vertices of the equilateral triangle. 

The optimization problem is defined as follow 



In the optimization problem the thickness of the tubes 
and the thickness of the flanges
 are assumed to be 2mm.

The results of the optimization problem showed that the tri –tube section that have the same weight (mass) of a uni –tube may improve the stiffness of the robot and minimize the total deflection in about 33%, this results means that we may construct a robot having tri –tube section which is less in weight from that of uni –tube section and both of them having the same end effecter deformation. 

Results:

In order to verify the analysis of the previous section a run had been done which has the following characteristics for the robot arm 
 


, 
, 
, 
 , 
, 
, 
 ( tube thickness ), 
(stiffener thickness), 
(mass of the first gear box), 
(mass of the second gear box), 
 (manipulated mass). The available gear boxes for the application are given in the list of table 1

The results of the program shows that for the given configuration the minimum weight for the structure of the uni –tube robot is (Wmin=19.986 N), the robot with such structural weight could manipulate the load with out failure because the stress in each arm is less than or equal to the allowable stress, but the deflection of the end point effecter is very large. The iteration process for increasing the dimension of the section to minimize the deflection and letting it be within the range (0.0005<
<0.0022) shows that there are 22 generation all of which has a deflection (0.0005<
<0.0022) and also a weight (W<Factor * Wmin ) the permissible weight factor (Fac.) for increasing the weight was chosen to be (Fac.=1.35). the dimensions of the inner side of the uni –tube section for the 22 generation are shown in fig.(4). The relation between the total deflection at the end -effecter and number of generation is shown in fig.(5). The relation between the new weight of the robot structure and its generation is shown in fig.(6). The program chooses the best generation which has the less variable (variable= weight*deflection), the relation between the variable and the generation is shown in fig(7) it is obvious that the generation no. 16 has the minimum value, the dimensions of the section for that generation are B1 =0.07464m, B2 =0.06397m, B3=0.05657m and has a deflection 
=1.87125*10-3  m and a total structural weight Wtotal =26.917N. Those results are the input for the next step in the program for calculating the dimensions of the tri –tube section in which an optimization problem where solved to maximize the moment of inertia for the section in terms of the dimensions h and D, the results of the program are shown in table (2)

The total deflection for the tri –tube configuration 


=1.31462*10-3 m

The deflection for the tri –tube configuration is less than that for the uni –tube which was found to be (
) this results shows that the tri –tube section reduces the deflection in a bout 29.7% from that of uni –tube section, the result may be improved to reach a value of 33.38% if we change the weight factor (Fac.) to make it equal to (1.3) on the other hand if (Fac.) is increased to(1.65) the improvement in deflection would be less and equal to (6.02%) those results for iteration and optimization are shown in table(3)

Another interesting feature in the design of the uni –tube section is that if the weight factor (Fac.) was chosen to be 1.3 and the range of deflection is limited to 
<0.0022 we would find only five generation which satisfies the previous mentioned configuration and if (Fac.) was changed to 1.65 and 
< 0.00104 we would find only 4 generation these results which are shown in table (4) shows the band of limits of the design of robot in other word we can not find a robot with a weight factor less than (Fac.=1.3) and has a deflection less than 0.0022m or we cannot find a robot with a weight factor less than (Fac.=1.65) and had a deflection less than 0.00104m those results of iteration are shown in table(4).

A flow chart of the program used is shown in fig.(8).

Conclusions:

This paper presents a method for optimization of robot design in the conceptual design stage. The robot is modeled in the MathCAD package and the optimization problem is formulated as to determine the dimension of robot arm in order to minimize the weight and maximize stiffness this formulation can be interpreted as to design the cheapest possible robot that will still meets the design demands. The optimization method showed good capability in finding an optimum set of dimension of the arm of robot manipulator with three degree of freedom.

The optimization method shows that the tri –tube is superior to uni –tube section in minimizing deflection in about 33%.

The presented work provides a good support for conceptual robot design. 




Fig. (1) robot configuration
(b)


 (a)

Fig.(2) (a) uni-tube configuration 
           (b) tri-tube configuration.


Fig.(3) modulation of manipulator links

Table 1 torque–mass relation for available gear boxes [2]

	Out put torques (N.m)
	Mass

(kg)

	101

231

572

1088

1499

2176

4361

6135
	2.5

4.4

9.5

12.7

18

28

47

69


Table (2) dimension of the tri –tube section

	arm No.
	h (m)
	D (m) 

*10-3
	Itri –tube(m4)

 *10-7
	Deflection (
)

m *10-4

	3

2

1
	0.10028

0.10440

0.10440
	7.52652

9.22085

11.92
	1.75056

2.48207

3.42024
	8.09087

7.35704

3.0044


Table (3) results for iteration and optimization problem
	Fac.
	
 

(m)*10-3
	

(m)*10-3
	Improvement

In deflection

	1.3

1.35

1.4

1.45

1.5

1.55

1.6

1.65
	1.39973

1.31462

1.2189

1.16048

1.09686

1.04631

0.985197

.948732
	2.10114

1.87125

1.6237

1.48623

1.3437

1.22646

1.08422

1.00955
	33.38%

29.7 %

24.9 %

21.9 %

18.37%

14.68%

9.13 %

6.02 %


Table (4) limits of deflection-weight factor for robot design

	Fac.
	
max (m)
	generations

	1.3

1.35

1.4

1.45

1.5

1.55

1.6

1.65
	0.0022

0.0019

0.0017

0.0015

0.0014

0.0013

0.0011

0.00104
	5

2

3

2

3

0

2

4


	Inner-side dimension  (B) for  uni-tube section in (mm)
	

	
	generations

	
	fig.(4) correlation between inner side dimension and no. of generations.


	Total deflection 
 (mm)
	

	
	generations

	Fig.(5) correlation between the deflection and no. of generations.


	Total weight  (N)          
	

	
	generations

	
	Fig.(6) correlation between robot weight and no. of generations.


	                                 Variable (VAR.)     


	

	
	generations

	Fig.(7) correlation between (Variable) and no. of generations.



 

References:

[1]. P.S. shiakolas, D. koladiya and J.kebrle “optimum robot design based on task specifications using evolutionary techniques and kinematic, dynamic, and structural constraints”, international journal of inverse problems in engineering, volume 10, number 4, pp 359-375, 2002.

[2]. Marcus P., Peter K., Xiaolong F.A., Johan A. and doniel W. “industrial robot design optimization in the conceptual design phase”, IEEE –mec. Of robot, 2004.

[3]. Abdel –Malek K. and Paul B., “criteria for the design of manipulator arms for high stiffness to weight ratio”, SME journal of manufacturing systems, Vol.17, No.3, pp.209-220.

[4]. E.I. Rivin, mechanical design of robots, 1988, Mc Grow –Hill, Inc, New York.

[5]. E.j. Hearn “Mechanics of materials”, Pergamon press, 1977.

تقليل الخطاء في الذراع الألي (الروبوت) على أساس التصميم الأمثل ولنسبة جساءة الى وزن عالية.

د. أحمد عبد الحسين علي

جامعة بغداد / كلية الهندسة
 قسم الميكانيك

الخلاصة:

في هذا البحث تم دراسة تاثير استخدام المقطع الثلاثي الأنابيب الدائريه لأجل تقليل الخطاء الطرفي في الذراع الألي ، تم اجراء مقارنه بين المقطع الثلاثي الأنابيب و المقطع المربع التقليدي للذراع الألي ، الدراسه بينت بانه لكلا الذراعين ذات المقطع الثلاثي و المربع والذان لهما نفس الوزن ممكن تقليل الخطاء بحدود %33 .
تم كتابة برنامج باستخدام MathCAD  لحساب أقل وزن للذراع الآلي ذي المقطع المربع الذي يمكن أن يتحمل الأوزان المسلطه ويعطي أقل تشوه.

الجزء الثاني من البرنامج يقوم بعملية الأمثليه لأجل ايجاد ابعاد المقطع ذي الأنابيب الثلاثيه الدائريه والذي له نفس وزن الذراع ذي المقطع المربع وله تشوه أقل من نضيره المربع.
Defining: E,� EMBED Equation.3  ���, m1,m2,m3, � EMBED Equation.3  ���,T,S, � EMBED Equation.3  ���



� EMBED PBrush  ���



   

� EMBED Equation.3  ���



2



� EMBED Equation.3  ���



� EMBED Equation.3  ���















1



   � EQ �� EMBED Equation.3  ���





� EMBED Equation.3  ���



m3



m1



m2



3

















































Fig.(8) flow chart of the program built-up by the use of MathCAD soft ware



Yes



Yes



No



No



Finding:� EMBED Equation.3  ���� EMBED Equation.3  ���



Optimization for tri-tube section

h>D , Wtii-tube=Wuni-tube, D� EMBED Equation.3  ���2T

Maximize (I), find h,D



1



1



Finding minimum (VAR=Wrobot� EMBED Equation.3  ���)



Saving generations



If

Wrobot>Fac.*Wmin



If

� EMBED Equation.3  ���



Calculation of: � EMBED Equation.3  ���,� EMBED Equation.3  ���,Wrobt



For B1 =B1+0.001

For B2=B2+0.001

For B3=B3+0.001



Calculation of  B3, B2, B1 and Wmin=q3*L3+q2*L2+q1*L1





PAGE  
25

_1236645461.unknown

_1238937592.unknown

_1238977271.unknown

_1239145252.unknown

_1239587407.unknown

_1257023452.unknown

_1257176484.unknown

_1257176901.unknown

_1257176255.unknown

_1239760023.unknown

_1239585709.unknown

_1239587255.unknown

_1239584024.unknown

_1239585314.unknown

_1239583099.unknown

_1239583266.unknown

_1239051691.unknown

_1239145002.unknown

_1239145193.unknown

_1239052851.unknown

_1238980484.unknown

_1239051232.unknown

_1238978590.unknown

_1238937862.unknown

_1238938031.unknown

_1238939508.unknown

_1238937938.unknown

_1238937717.unknown

_1238937778.unknown

_1238937650.unknown

_1236726162.unknown

_1238937241.unknown

_1238937500.unknown

_1238937518.unknown

_1238937313.unknown

_1236726579.unknown

_1238937087.unknown

_1236726483.unknown

_1236725767.unknown

_1236725952.unknown

_1236725998.unknown

_1236725854.unknown

_1236645729.unknown

_1236645868.unknown

_1236645559.unknown

_1236562568.unknown

_1236644217.unknown

_1236645198.unknown

_1236645359.unknown

_1236645100.unknown

_1236641265.unknown

_1236641354.unknown

_1236641126.unknown

_1236641014.unknown

_1236641072.unknown

_1236640860.unknown

_1235773038.unknown

_1235773823.unknown

_1235774657.unknown

_1236561856.unknown

_1235774271.unknown

_1235773440.unknown

_1235773735.unknown

_1235773154.unknown

_1235691626.unknown

_1235771993.unknown

_1235772700.unknown

_1235771805.unknown

_1226829966.unknown

_1235690633.unknown

_1235691262.unknown

_1235690109.unknown

_1226830001.unknown

_1226829657.unknown

_1226829939.unknown


_1226829543.unknown