Академический Документы
Профессиональный Документы
Культура Документы
This content has been downloaded from IOPscience. Please scroll down to see the full text.
(http://iopscience.iop.org/0957-0233/24/11/115101)
View the table of contents for this issue, or go to the journal homepage for more
Download details:
IP Address: 81.161.248.93
This content was downloaded on 23/07/2015 at 15:56
Abstract
A novel structure for a six-axis force/moment sensor is proposed in this paper. Parallel plate
structures are combined in the one-block structure of the sensor. A method for estimating the
sensitivity and stiffness is developed by comparing the results of numeric estimations with
those of finite element method (FEM) analyses. A detailed FEM model was used to analyse
the surface strain in the area where the strain gauge was attached. Tests carried out on the
characteristics revealed a sensitivity design error of less than 8.5%, a relative measurement
error of less than 2.13%, and an interference error of less than 9.51%. A performance
enhancement method is also proposed.
Keywords: six-axis force/moment sensor, parallel plate structure, sensitivity design, FEM
(Some figures may appear in colour only in the online journal)
0957-0233/13/115101+10$33.00 1 2013 IOP Publishing Ltd Printed in the UK & the USA
Meas. Sci. Technol. 24 (2013) 115101 W Wang
300
10
Y4 Z2
2w
=2l
0
by
2d Mz X4
z
ty
Z1
Y2
41
tz
bz
Lz X3
Figure 1. Wind tunnel vehicle model test. X1
Z3
x Y1
Strain gauge
2d
Ly
P P y
+
M
x
Y3
X2 Z4
2d M y
L
bx
Weight = 2kg
tx
Lx
+
t
2
Meas. Sci. Technol. 24 (2013) 115101 W Wang
GFz1:T Z1 Z2
X1 Y2 Y4
z GFy2:C GFz2:C
X4 GFy4:C
GFx1:T x z
GFx4:C GFz3:T
y x
x
GFx2:C y GFz4:C y
GFx3:T
X2
X3 GFy3:T
Y3 Y1 GFy1:T
Z4 Z3
(a) Fx (b) Fy (c) Fz
G :T GMy1:T
Z 1 GMx3:T GMx1:T Z Z1 My3
Z2
GMz3:T GMz4:C Y4
2
z z Y2
x GMx2:C x GMy2:C
GMx4:C
y GMy4:C y x
y
Z3 Y1
Z4 GMz2:C
Z3 Z4 GMz1:T
Y3
(d) Mx (e) My (f) Mz
T C GMz4: - b - t
1 4 Y4
GMz3: + b+ t
Output
Y2
2 3
C T
Input x
y
Y1
Figure 5. Gauge bridge.
Y3 GMz1: + b+ t
GMz2: - b - t
Furthermore, on the basis of equation (2), the stiffness of the
proposed sensor in the corresponding direction is given by Figure 6. Deformation caused by Mz.
4Ebt 3
kf = . (5) the moment sensor because only the moment arm d needs to
L3
be adjusted.
However, as discussed later, it was confirmed by FEM that
3.2.2. Rotational deformation mode. As shown in the design error of Mz was extremely large compared to those
figures 4(d)( f ), the deformations of the plates under Mx, My of the components about the other axes. A more precise Mz
or Mz are obviously different from those of the PPS shown estimation model is therefore proposed for the sensor design.
in figure 2. As shown in figure 4(d), Mx twists plates Z1Z4.
As shown in figures 4(e) and ( f ), the tangential directions of
the two ends of each of plates Z1Z4 or plates Y1Y4 are not 3.2.3. Deformation due to Mz. For the measurement of
parallel under My or Mz. However, to estimate the sensitivity Mz using plates Y1Y4, as shown in figure 6, the sensitivity
and stiffness in the initial design phase, a simple method that estimation requires the calculation of both b (the strain due
ignores the difference between the actual deformation and that to the y-direction deformation, which is approximately the
of the PPS shown in figure 2 can be used. PPS deformation) and t (the strain due to the x-direction
Mx is equal to the moment generated by the reaction forces deformation). The reason is that both strains contribute to the
of the two equivalent PPSs formed by plates Z1Z4, where the output of the bridge circuit, as shown in figure 6, wherein
moment arm dMx is the distance between the PPS and the x- the tension strain is indicated by + and the compression is
axis. The moment arms dMy and dMz are also defined as shown indicated by .
in figure 3. The gain of the moment sensor can therefore be A calculation model is proposed to determine the
estimated using quantitative relationship between the sensitivity and the design
3L parameters. In the model, plates Y1Y4 are replaced by
Gm = . (6) equivalent springs, as shown in figure 7. The x-direction
Ebt 2 d
The stiffness of the moment sensor can be estimated using stiffness can be calculated using the size of the plate and the
Youngs modulus of the material:
4Ebt 3 d 2
km = (7) Ebt
L3 . kx = (8)
where d is the moment arm relative to the specific axis. Here, L
the advantage of this method is that the sensitivity and stiffness The y-direction stiffness is associated with the bending
of the force sensors are not affected by the design process of deformation, which is regarded in the model as the PPS
3
Meas. Sci. Technol. 24 (2013) 115101 W Wang
The reaction force of the equivalent springs and the load Mz are
therefore
fx = kx x (12)
fy = ky y (13)
Mz = 4( fx w + fy l). (14)
The stiffness is given by
Mz Ebt 2 Ebt 3 2
kMz = =4 w + 3 l . (15)
L L
To design the sensitivity, the strain related to the x-direction
Figure 7. Model of Mz sensor. deformation can be obtained using
fx
t = . (16)
Ebt
Also, the strain related to the y-direction deformation is given
by
3L fy
b = . (17)
Ebt 2
Considering that the strain output of the bridge circuit is four
times the sum of t and b, the gain can be obtained as
4(t + b ) L(3lt + wL)
GMz = = . (18)
Mz Ebt(L2 w 2 + t 2 l 2 )
Figure 8. Geometric constraint.
3.2.4. Key parameters. Following the trade-off between
sensitivity and stiffness, the key parameters (shown in figure 3)
deformation. From equation (2), the y-direction stiffness can of a specific scaled vehicle model for the wind tunnel test were
be obtained as set according to the specifications listed in table 1. A material
Ebt 3 with Youngs modulus E = 72.6 GPa was used. The stiffness
ky = . (9) of the sensor was used to estimate the resonance frequency of
L3
the modelsensor system.
Considering the geometric constraint of the structure shown in
The vehicle model inertial load is vital to the proposed
figure 8, the deflection of the equivalent springs is given by
sensor when applied to a wind tunnel model vibration test.
x = sin = w (10) The maximum inertial load P was calculated to check the
capacity of the sensor under the most severe experimental
y = cos = l. (11) conditions expected, which is when the vibration frequency
is 10 Hz and the amplitude is 5 mm or 3 . The load
4
Meas. Sci. Technol. 24 (2013) 115101 W Wang
capacity N was calculated using the gain estimation equations sensitivity for the FEM validation. Second, barring the thin
and the maximum strain in the linear elastic range of the plates, the deformation of the rest of the sensor structure
material (about 6000 m m1) observed during the capacity was not considered for the estimation equations, whereas
test using the early prototype of the sensor. All the values the deformation of the entire sensor structure was taken into
thus obtained were greater than the expected maximum load, account for the FEM validation. For example, figure 11 shows
with the exception of the significantly erroneous value of Mz the strain distribution in the direction of L on plate X2 for
calculated using the gain estimation equation (6). Fx = 1000 N.
The FEM gain of My is also approximately 70% of
3.3. FEM calculation the estimated gain. However, as shown in figure 12, an
asymmetrical strain distribution in the plates was obtained
An FEM model was built to validate the above sensitivity in the case of FEM.
and stiffness design method. As shown in figure 9, the detailed The asymmetric strain distribution can be explained by
deformation shape and strain distribution can be obtained using the specific structure of the sensor. As shown in figure 3, the
the model, which has a fine mesh of deformable plates. The stiffness of part C is apparently higher than that of part B, i.e.,
attachment point of the strain gauge is shown in red. The details the values of the support stiffness at the two ends of each of
of the strain gauge attachment point are shown in figure 10. plates Z1Z4 are different. The strain at the end connected to
The average strain of the elements in the attachment area was part C is therefore greater than that at the end connected to
used to calculate the gain of the sensor for the FEM validation. part B.
The FEM results and the estimation results obtained by the The FEM gain of Mx was approximately 80% of the
simplified equations are compared in table 2. estimated gain. As shown in figure 13, the two-dimensional
The FEM gains are approximately 70% of the estimated surface strain distribution due to the twist deformation and the
values for Fx, Fy and Fz. Considering that the deformations attachment of the strain gauge can be studied using FEM.
of the plates are almost the same as the PPS deformation Owing to space limitation, the strain gauges that make up
in these cases, the estimation error might be primarily due the bridge circuit that measures Mx can be attached along
to the following two factors. First, the estimation equations the outer line to the surfaces of plates Z1Z4. The strain
give the maximum strain at the end of the plates, whereas distribution in the direction of L is shown in figure 14. The
the average strain of all the elements in the area where gain will be 261 (m m1) N1 m1 if the gauges GMx1
the strain gauge was attached was used to calculate the GMx4 are attached along the centre line of the plates, i.e., at the
5
Meas. Sci. Technol. 24 (2013) 115101 W Wang
attachment points of gauges GMz1GMz4. Similar to the gains estimated by equation (18). The strain distribution is shown in
for the other axes, this gain is 73% of the estimated gain, which figure 15.
was 359 (m m1) N1 m1. Table 3 compares the maximum surface strain of the
The FEM gain of Mz was only 38% of that estimated by elements in the gauge area with the analytically obtained strain
equation (6), whereas the accuracy of the estimation model in for assumed applied force and moment of 1000 N and 100 N m,
section 3.2.3 was acceptable. The FEM again was 93% of that respectively. The difference can be attributed to the effect of
6
Meas. Sci. Technol. 24 (2013) 115101 W Wang
Figure 14. Strain distribution on Z4 (Mx = 100 N m). This effect of the series connection and the compliance of
the entire structure of the sensor barring the plates contribute
to the lower stiffness of the sensor in the FEM calculation.
Compared to equation (7), equation (15) produces a more
reasonable estimate of Mz-axis, i.e., the stiffness estimated by
the equation is higher than that obtained by FEM.
4.1. Fabrication
The proposed six-axis force/moment sensor shown in figure 16
was produced from a single aluminium alloy (A7075-T6)
block by wire cutting. The Youngs modulus E was 72.6 GPa.
The single-block structure reduces the hysteresis of the sensor
output, which is caused mainly by friction between adjacent
parts.
Figure 15. Strain distribution on Y3 (Mz = 100 N m).
4.2. Test of characteristics
the elastic support of the thin plates, which was considered for
FEM but not in the analytical method. A special apparatus was developed for the calibration test of
As regards the eigenfrequency of the model-sensor the proposed six-axis force/moment sensor. The method for
system, the FEM results were lower than the estimated generating the pure force and pure moment is similar to that
values. Only the deflection of the plates in the direction of a six-component force/moment sensor testing machine [9].
of L was considered by the estimation equations (5) and As shown in figure 17, each axis can be loaded by a
(7). The deflections of the plates in other directions affect rearrangement of the experimental apparatus. The calibration
the compliance of the sensor along the corresponding axes. results of Fx-axis and Mx-axis are shown in figures 18 and 19,
7
Meas. Sci. Technol. 24 (2013) 115101 W Wang
(f)
(d) (e)
Figure 18. Calibration results for Fx input. Figure 19. Calibration results for Mx input.
8
Meas. Sci. Technol. 24 (2013) 115101 W Wang
9
Meas. Sci. Technol. 24 (2013) 115101 W Wang
5. Conclusion References
We have described our development of a six-axis [1] Park J J and Kim G S 2005 Development of the 6-axis
force/moment sensor for the simultaneous measurement of force/moment sensor for an intelligent robots gripper
Sensors Actuators A 118 12734
the forces Fx, Fy and Fz and moments Mx, My and Mz when [2] Kim G S 2004 Development of a small 6-axis force/moment
testing a scaled model in a wind tunnel. The PPS was adopted sensor for robots fingers Meas. Sci. Technol. 15 22338
for the proposed sensor. [3] Kim G S et al 2008 Development of 6-axis force/moment
The PPS theory equations and FEM analyses were used to sensor for a humanoid robots intelligent foot Sensors
design the sensor. Equations for estimating the sensitivity and Actuators A 141 27681
[4] Kim G S 2007 Design of a six-axis wrist force/moment sensor
stiffness were proposed. FEM analysis was used to validate using FEM and its fabrication for an intelligent robot
the estimation method, which is especially useful in the initial Sensors Actuators A 133 2734
design phase. [5] Chao L P and Chen K T 1997 Shape optimal design and force
The results of the characteristic tests showed that the errors sensitivity evaluation of six-axis force sensors Sensors
between the FEM and calibration results were less than 8.5%. Actuators A 63 10512
[6] Joo J W et al 2002 Design and evaluation of a six-component
The relative measurement errors of the proposed sensor were load cell Measurement 32 12533
less than 2.13%, and the interference errors were below 4.38%, [7] Aschwanden P, Muller J and Knornschild U 2006
with the exception of the MxMy interference, which was Experimental study on the influence of model motion on the
9.51%. aerodynamic performance of a race car SAE Technical
Another gauge attachment position plan was proposed to Paper 2006010803
[8] Hatamura Y et al 1988 A trial on a 6-axis force sensor for
reduce the interference error of the sensor. In a future work, robots Japan Soc. Mech. Eng. Trans. C 54 2416
an attempt will be made to enhance the performance of the [9] Kim G S 2000 The development of a six-component
sensor, which will also be used to measure the aerodynamic force/moment sensor testing machine and evaluation of its
forces acting on a scaled model in a wind tunnel test. uncertainty Meas. Sci. Technol. 11 137782
10