Beruflich Dokumente
Kultur Dokumente
Capsules
I.
INTRODUCTION
Fig. 1. Three types of the electrodes, namely Type I, II, and III, which were
used to study the affects of the shape and size of the electrodes on the
sensitivity of the capacitive sensor, and stability of the response to the small
displacement of the capsule. The first column and second column show the
front and side view of the three electrode types, respectively. When moving
through the sensor, the axis of the capsule is aligned with the x axis of the
corresponding set of electrodes.
i
me = i =0 K i rm
{1... N }
TABLE I
THE NORMALIZED SENSITIVITY SN OF THE SENSOR FOR THE THREE TYPES OF
ELECTRODES WITH CORRESPONDING RELATIVE CHANGE IN THE SENSITIVITY
AS A FUNCTION OF 0.2 MM CAPSULE DISPLACEMENT IN THE DIRECTIONS OF
X, Y, AND Z AXIS (FIG. 1). THE MEASUREMENTS PERFORMED ON A REAL SET
OF ELECTRODES ARE GIVEN IN BOLD. THE DIMENSIONS OF THE ELECTRODES
FOR TYPE I WERE W=25 MM, H=15 MM, D=10 MM, AND D=1 MM, FOR TYPE
II W=25 MM, D=1 MM, D1=2 MM, AND R=5 MM, AND FOR TYPE III W=25 MM,
D=1 MM, D1=4 MM, AND R=5 MM.
SN
Direction x (%)
Direction y (%)
Direction z (%)
Type I
0.74 (0.7)
0.02 (0.0)
0.15 (0.2)
0.02 (0.0)
Type II
1.00 (1.0)
0.01 (0.0)
0.92 (1.0)
0.31 (0.3)
Type III
0.80
0.01
0.17
5.37
Fig. 3. The effect of the distance between the flat electrodes (Type I) D on
the sensor response at 0.2 mm displacement of the capsule in the directions
of x, y and z axis.
RMSD =
1
M
(mi j =0 K 0, j rm,i )
=1
(7)
TABLE II
THE TWO SETS OF CAPSULES USED FOR THE CHARACTERIZATION OF THE
MEASUREMENT SYSTEM. THE TRUE WEIGHT OF EACH CAPSULE WAS
DETERMINED WITH ACCURACY OF 0.1 MG.
Capsule Diameter/length Weight range Number of Calibration Validation
set
(mm)
(mg)
samples
set
set
Set I
5.3 / 14.3
45 - 285
5 x 10
3 x 10
2 x 10
Set II
21.7 / 7.6
90 - 600
5 x 21
3 x 21
2 x 21
Fig. 4. Two typical capsules from the Set I and Set II.
rm 1 rm rm 2
(11)
IV. RESULTS
This section summarizes the results obtained by measuring
the response of the measurement system using the flat
electrodes (Type I) to the two test sets of capsules (Table II).
The distance between the electrodes D was set to 10 mm in
order to maintain sufficient space between the capsule and the
electrodes and to minimize the effect of the capsule
positioning error on the weight measurements (Fig 3.). In
order to maximize the sensitivity of the sensor the width W
and height H were set to 25 mm and 15 mm, respectively.
The results showed that fitting a second order polynomial
was sufficient to compensate the small nonlinearity of the
measurement system. The accuracy A (RMSE), precision P,
correlation coefficient k, stability RMSD, and nonlinearity NL
of the measurement system as a function of the ambient
temperature, humidity and capsule size are shown in Table
III.
The response of the sensor rm to the typical empty and
full capsules from Set I and Set II as a function of capsule
displacement in the direction x corresponding to the normal
movement of the capsule through the sensor is shown in Fig.
5.
( mi m e )
A = i =1
M
(8)
M
Additionally the root-mean-square error of the estimated
weight me was calculated for each of the two validation sets.
RMSE =
M
i =1
(mi me )
(9)
M 1
The precision P of the measurement system was estimated
by repeating the measurements of each capsule in the
validation set for several times and then calculating the mean
standard deviation (STD) of the estimated weight me over all
the capsules in the validation set.
STD(me,i )
P = i=1
M
(10)
M
The nonlinearity NL of the measurement system was
characterized by the maximum of the absolute difference
between the line of best fit K1 and the best-fit polynomial KB
yielding the lowest RMSE and at the same time not overfitting
the data. The value of the sensor response rm was limited by
its minimum rm1 and maximum rm2 observed within the
calibration set corresponding to the empty and full capsule,
respectively.
Fig. 5. The normalized response of the sensor to the typical empty and full
capsules from the Set I and Set II as a function of the capsule position x
while passing between the electrodes.
Fig. 6. (a,c) The normalized sensor response to each of the capsules in the corresponding test set and the transfer functions of the measurement system estimated
by a second order polynomial for the test Sets I and II, respectively. (b,d) The correlation coefficient between the estimated and the true capsule weight for the test
Sets I and II, respectively. All the four plots are based on measurement at nominal ambient temperature of 19 OC and RH of 55%.
TABLE III
THE ACCURACY, PRECISION, CORRELATION COEFFICIENT, STABILITY, AND
NONLINEARITY OF THE MEASUREMENT SYSTEM AS A FUNCTION OF THE
AMBIENT TEMPERATURE, HUMIDITY, AND CAPSULE SIZE.
The test sets and the ambient temperature and humidity
55% RH
23 OC
19 OC,
25OC
45% RH
60% RH
Set I Set II Set I Set II Set I Set II Set I Set II
A (mg)
0.1 -0.2 0.0
0.1
0.1
0.1 -0.1 0.2
RMSE (mg) 1.3
2.8
1.2
3.0
1.4
3.1
1.3
3.0
P (mg)
1.5
3.1
1.6
3.1
1.5
3.2
1.6
3.2
k
0.999 0.999 0.999 0.999 0.999 0.999 0.999 0.999
RMSD (mg) 1.1
1.3
0.8
0.6
1.2
1.4
NL (mg)
5.2 10.1 5.1 10.3 4.9 10.5 5.1
9.9
V. DISCUSSION
During the design of the measurement system several
factors affecting the sensitivity and precision of the
measurements were taken into the account among these the
shape and size of the electrodes, ambient temperature, relative
humidity, and complexity of the calibration procedure. A
[2]
[3]
[4]
[5]
[6]
[7]
[8]
August 2008)
August 2008)
Wiley, 2002.