3.1. Theoretical Modeling Error Analysis
In order to theoretically analyze the radial error motion measurement with the T-type CS, we consider a small radial displacement of the precision spindle, i.e., |
α/(
R −
g)| < 1. The power series expansion with first-order approximation (Equation (1)) was applied to express the equivalent spacing changing Δ
u. The exact derived expression of it is:
Then, the approximate error of Equation (1) can be assessed by:
Figure 7 shows the variation situation of error Δ
uapp.error with the changing of parameter terms
α/(
R −
g) and cos(
θ −
β). As mentioned above, the radial displacement
α of a precision spindle is generally very small. In order to achieve relatively high sensitivity, the cylindrical electrode radius
R is often designed to be tens of millimeters or even larger. Thus, the variation range of the term
α/(
R −
g) is set to be [0, 0.1]. It can be observed that, when the ratio
α/(
R −
g) is less than 0.01, the error Δ
uapp.error basically keeps lower than 2% with the variation in the term cos(
θ −
β) in the entire phase region (the value range of cos(
θ −
β) is [0, 1] correspondingly), except for a few phase angles. Therefore, a highly accurate value of the changing Δ
u calculated by Equation (1) can be obtained if the ratio
α/(
R −
g) is less than 0.01, which can be achieved by adjusting the structural parameters of the sensor.
Since a highly accurate value of Δ
u can be obtained by setting the ratio
α/(
R − g) ≤ 0.01, the expressions (3)–(6) established based on Equation (1) could be more accurate to reflect the capacitance changing caused by the radial displacement, and is also for the expressions of the differential output capacitance of the REG derived as:
In a manner similar to the simplification process on the equivalent spacing changing Δ
u, the first-order approximation of the Equations (31) and (32) could intuitively reflect the variation’s relation of the radial displacement and the differential capacitance (CrX and CrY). Thus, the expressions of the displacement
α and phase angle
β can be deduced in the following form:
Therefore, the expression used to assess the approximate error (taking the CrX as an example) is given as:
As shown in
Figure 8, at each phase angle
β, the approximate error CrX
app.error all increases with the increase in the ratio
α/
g. When the ratio
α/
g becomes larger than 0.2, the variation tendency is more significant.
Figure 9 shows the error of the calculated capacitance CrX
tl relative to the simulation value CrXf as the rotor has a displacement along the direction of y = x. The dependence of the relative error on the parameter
α is basically consistent with the variation tendency shown in
Figure 8. Note that the value CrXf was obtained based on the simulation model of the T-type CS previously established with the design parameters in [
16]; these parameters were also adopted to calculate the theoretical capacitance.
From the analysis in
Figure 8, it can be known that diminishing the ratio
α/
g could make the error CrX
app.error reduce further, but the measurement range and sensitivity of the sensor will be affected to some extent. By considering the inherent nonlinearity of the capacitor output with spacing change, the power series expansion with a third-order approximation was derived to improve the calculation accuracy [
18]:
Compared to the error distribution shown in
Figure 8, the error CrX
app.error is significantly decreased overall for the calculation expression of the CrX with third-order approximation, as shown in
Figure 10. As the phase angle
β varies from 0° to 90°, the error CrX
app.error remains less than 0.4% for the ratio
α/
g = 0.1 and no more than 1.2% for
α/
g = 0.2.
Figure 11 compares the simulated differential capacitance CrXf and theoretical counterparts CrX
tl3 calculated with Equation (36). The phase angle
β equals 45°, that is, the rotor has a displacement along the direction of y = x. It can be observed that the deviation between the value CrXf and the CrX
tl3 is closer to the actual value.
In order to theoretically analyze the tilt error motion during the measurement, we considered tiny tilt displacement of a precision spindle and reasonable structural parameters of the sensor, that is, |
kρ/
a| < 1 for the parameter term
kρ/
a in Equation (15). The power series expansion with first-order approximation (Equation (38)) was also applied to express the integrand term in Equation (15):
Correspondingly, the expression used to assess the approximate error is given as:
By considering the relative tiny tilt displacement of a precision spindle and reasonability of the sensor structural parameters, the variation range of the term
kρ/
a was set to be [0, 0.2]. As shown in
Figure 12, the approximate error IGe
app.error is relatively smaller over the range of the parameter terms cos(
φ +
γ) and
kρ/
a. According to the simulation model parameters of the sensor [
16], the ratio of the term
kρ/
a is about 0.04 with the tilt displacement of 200 arc-sec; thus, the error IGe
app.error remains no more than 0.2% within the whole range of the term cos(
φ +
γ), i.e., the entire phase region.
Figure 13a compares the theoretical and simulated capacitance of the end part electrode. The theoretical capacitance is integrated by the integrand term with a second-order approximation (C8
tl2). It can be observed that there is little difference between C8
tl2 and the counterpart using first-order approximation (C8
tl), which is also the same for the theoretical value of the CeY (the differential output capacitance of the EPEG) shown in
Figure 13b. Besides, similar to the theoretical value of the CrX calculated by the expression with third-order approximation (see in
Figure 11), the difference between the theoretical capacitance CeY
tl and the simulation one CeYf remains roughly unchanged relative to the simulation value, which indicates that the theoretical capacitance of the end part electrode solved by adopting IGe
tl as integrand can meet the design needs for relatively accurate results.
3.2. Influence of Fringe Effects
Due to the existence of diverging electric fields in the plate edges in
Figure 14, i.e., the “fringe effects” [
19], the measured capacitance of a sensor contains some additional capacitance introduced by the fringe effects. As for the sensor electrode with finite dimensions, the variation in its output capacitance is relatively small, especially for the measurement of tiny displacement. As such, the influence of the fringe effects cannot be neglected.
Figure 15 shows the distribution of the electric field between the cylindrical excitation electrode (
Ed) and curved sensing electrodes (
Es) for the fundamental configuration of the radial curved plate capacitor. It can be seen that the electric field concentration appears in the junction of the inner circle face of the curved plate and its side face. As shown in
Figure 15a, a different intensity distribution of the electric field between the
Ed and the side face of the
Es and its adjacent outer circular face formed along the radial direction. Once the equipotential guard ring (
Egr) is employed outside the
Es, the electric field distribution between the
Ed and the outer circular face of the
Es and between the
Ed and local region of the side face near the outer circular face can be eliminated, as shown in
Figure 15b. The
Egr is also classically named Kelvin guard-ring.
The variation in the electric field distribution is also reflected in the output capacitance of the
Es, as illustrated in
Figure 16. The use of the
Egr can effectively reduce the additional capacitance introduced by the fringe effects (see
Figure 16a,b). Moreover, the guard ring could decrease the variation in the additional capacitance with the reduction in radial spacing (see
Figure 16d), which is helpful to improve the measurement accuracy and linearity. For the T-type CS, the
Egr is distributed around the periphery of the sensing electrodes with a tiny gap (
λ = 0.1 mm), forming a coplanar configuration between the
Egr and the electrodes. Under the effect of the
Egr, the amount of variation in the additional capacitance significantly decreased (about 10 times), and the corresponding nonlinear error reduced from 4.8% to 1.7%, as indicated in
Figure 17.
3.3. Installation Errors of the Sensing Electrodes
The sensing electrodes of the T-type CS include the cylindrical and fan-shaped electrodes integrated into the REG and EPEG, respectively. These electrodes are fabricated by the FPCB process in one procedure, along with the equipotential guard ring, and the sensing electrodes have a fixed position relationship with the guard ring.
The fabricated REG with planar form (as shown in
Figure 18a, solid line box) is installed in the annular groove of the stator cylindrical bore by the surface mounting method. The coaxiality error, polar angle position error and tilt error (the axial boundary of the electrode relative to the axial reference of the stator) relative to the mounting reference were produced during the process. By utilizing the electrode measurement and sensor calibration method, the coaxiality and polar angle position error can be modified and removed, respectively. Thus, the analysis of the tilt error is the main focus for the cylindrical electrodes.
By spreading out the installed REG, as shown in
Figure 18a, it can be seen that a certain tilt angle
εe is produced relative to the ideal position of the REG. The point
Om is assumed to be the tilt point. The dash line box represents the position of the cylindrical electrodes, and the dash line segment represents the axial boundary of the REG. As mentioned above, the in-plane relationship between the cylindrical electrodes is fixed during the mounting process, and thus the tilt angle
εe is the tilt error of the electrodes.
As shown in
Figure 18b, by taking a micro element with arc length
dθ on the boundary
D′C′ of the electrode
Er4, the axial length
wJ′ of the electrode at this location can be considered as the function of the polar angle
θJ′ at point
J′. The solution region is vertically divided into three parts according to the constraint boundary of the length
wJ′, and the corresponding calculation function can be expressed as follows:
where
kI =
k1 + 1/
k1,
kIII = −
kI,
bI =
b3 −
b4,
bII =
b3 −
b1,
bIII =
b2 −
b1;
k1 is the slope of the boundary
D′C′;
b1~
b4 are the intercepts of the lines including
D′C′,
C′B′,
B′A′,
D′A′, respectively.
Referring to Equation (6), the approximate expression of the output capacitance of the electrode
Er4 under the tilt angle
εe can be obtained:
By considering the minor radial displacement of a precision spindle, at the condition |
α/
g| < 1, the power series expansion with neglection of the high-order term is applied to express the integrand term in Equation (43), and thus we have:
Similarly, the approximate expression of the output capacitance of the electrodes
Er3~
Er1 are derived as follows:
where,
k4I =
k3I =
k2I =
kI,
k4III =
k3III =
k2III =
kIII,
b4II =
b3II =
b2II =
bII,
biI =
b3 −
bi4,
biIII =
bi2 −
b1(
i = 4, 3, 2);
bi2,
bi4 are the intercepts of the lines, including
Ci′Bi′,
Di′Ai′ (
i = 4, 3, 2), respectively;
i represents the electrodes
Er1~
Er3 in turn.
Further, the influence of the tilt error
εe on the output capacitance of the cylindrical electrode (taking
Er4 as an example) can be reflected by:
By referring to the structural design of the T-type CS, the range of the tilt angle
εe was set to be [0.006, 0.06] deg (about 200 arc-sec) in the numerical simulation using Matlab. A tiny circumferential displacement of the cylindrical electrode was produced due to the tilt angle
εe, which produces a relative error of the output capacitance under the radial displacement along the same direction. The dependence of the relative error on the angle
εe is shown in
Figure 19. For the single electrode, the influence of the tilt angle
εe on the output capacitance is very small, with respect to the amplitude of the relative error.
Figure 20 presents the dependence of the differential capacitance on the tilt angle
εe. It can be known that the influence of the tilt angle
εe on the differential output capacitance of the REG is slightly larger. By considering the rotor of run-out in the given directions (
β = 5°, 45°, 85°), the differential output capacitance (CrXt-
εe and CrYt-
εe) can be calculated by the capacitance of single electrodes under the angle
εe. The difference between the differential output capacitance under the angle
εe and the counterpart under an ideal position (CrXt and CrYt) could be transformed to the difference of displacement parameters, utilizing Equations (33) and (34). The difference in the displacement parameters (magnitude
α and phase angle
β) are shown in
Figure 21. It can be known that the produced tilt error
εe of the cylindrical electrode only generates a certain deviation of solving the phase angle
β, which is consistent in different displacement directions (brings about the overall phase advance or lag of the displacement trajectory); the effect of tilt error
εe on solving the radial displacement
α can be neglected, especially for the measurement range of less than 0.1 mm.
The fabricated EPEG was also in planar form (as shown in
Figure 22a, solid line box) and installed in the annular groove located at the end part of the stator by the surface mounting method. The coaxiality error, polar angle position error and parallelism error (the electrode plane relative to the axial reference of the stator) relative to the mounting reference were produced during the process. For the parallelism error, it can be modified by the electrode measurement. Thus, the analysis of the fan-shaped electrodes is emphasized on the coaxiality error and polar angle position error.
Due to the position error between the shape boundary of EPEG and the sensing electrodes and the manipulation precision, there may be an eccentricity between the geometric center
Oe′ of the fan-shaped electrodes in the EPEG and the medial axis of the stator (the origin position of the coordinate system). As shown in
Figure 22a,
δe denotes the magnitude of the eccentricity, and
φe is the phase angle.
δe is the coaxiality error of the fan-shaped electrode.
By taking the electrode
Ee1 as an example, the region confined by the points
C′D′S′T′ (dash line) represents the position of
Ee1 under the coaxiality error
δe. As shown in
Figure 22b, the micro-plane element ΔB
P′ (oblique line filled) at any point
P′ is taken as the solving unit, and the area of this element is denoted as Δ
SP′. Then, the Δ
SP′ is approximately given by:
The projection point of
P′ on the
x′-axis is the point
P′x′. By referring to Equation (13), the expression for calculating the spacing
t′ at this point could be derived as:
Correspondingly, the capacitance of the plane parallel capacitor composed of the micro-planes ΔB
P′ and ΔA
Q′ is:
Through the integration of Equation (51) over the area of the fan-shaped electrode, the output capacitance of the electrode
Ee1 is approximately expressed as:
where the integral domain
D is a closed region confined by the boundary curves of the fan-shaped electrode, which can be divided into three sub-regions, i.e.,
DI,
DII and
DIII. These three sub-regions can be expressed as follows:
where
xB and
yB are the coordinates of the geometric center
Oe′;
D′x,
C′x,
E′x and
T′x represent the abscissas of the corresponding points, respectively.
Equation (52) can be rewritten as:
As for multivariate function, it can be expanded to power series under certain conditions by analogy with the univariate power series expansion [
20]. By considering the tiny tilt displacement of a precision spindle and reasonable structural parameters of the sensor, there is generally |(
k/
a)∙(
xcos(
γ) −
ysin(
γ))| << 1, the power series expansion with neglecting of a high-order term is applied to express the integrand term in Equation (53). Then, it can be expressed as:
By calculating Equation (54) within the integral sub-domains
DI,
DII and
DIII, we have:
Similarly, the approximate expressions of the output capacitance of the electrodes
Ee2~
Ee4 can be derived as follows:
where
D′ix,
C′ix,
E′ix and
T′ix(i=1,2,3) represent the abscissas of boundary points of the electrodes in the 2nd to 4th quadrant, respectively;
;
μ =
kcos(
γ)/
a,
ν =
ksin(
γ)/
a.
Further, the expression used to assess the influence of the coaxiality error on the output capacitance of the fan-shaped electrode (taking
Ee1 as an example) is given as:
In the numerical simulation using Matlab, the range of the coaxiality error
δe is set to be [−0.2, 0.2] mm, and the variation quantity of the phase angle
φe is 8 degrees referring to the sensor structural design.
Figure 23 shows the relative error variation in the output capacitance of fan-shaped electrodes due to the error
δe under different phase angles
φe. For each fan-shaped electrode, the relative error of output capacitance increases with the increase in the error
δe. It should be pointed out that the magnitude of the relative error is quite small within the whole value range of the error
δe. Moreover, the influence of the phase angle
φe variation in the relative error is also very limited.
From Equations (27) and (28), it can be known that the tilt displacement
ε and yaw angle
γ are obtained based on the calculated differential output capacitance of the EPEG (CeY, CeX). As shown in
Figure 24, under different yaw angle
γ, the produced coaxiality error
δe and phase angle
φe do not cause the changing of the capacitance CeY and CeX, which indicates that the measurement of the rotor tilt displacement is basically not affected by the coaxiality error of fan-shaped electrode.
The polar angle position error of the fan-shaped electrodes equals each other and does not change the increment of the electrode polar angle. Thus, the effect of the polar angle position error on the yaw angle γ measurement can be eliminated by the sensor calibration method.