Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

Precision Engineering 74 (2022) 232–241

Contents lists available at ScienceDirect

Precision Engineering
journal homepage: www.elsevier.com/locate/precision

Development of an on-machine measurement system for ultra-precision


machine tools using a chromatic confocal sensor
ZaoZao Chen a, ZhenDong Wang a, MingJun Ren a, XinQuan Zhang a, *, LiMin Zhu a,
XiangQian Jiang b
a
State Key Laboratory of Mechanical System and Vibration, School of Mechanical Engineering, Shanghai Jiao Tong University, Shanghai, 200 240, China
b
Engineering and Physical Sciences Research Council (EPSRC) Future Metrology Hub, School of Computing and Engineering, University of Huddersfield, Huddersfield,
HD1 3DH, UK

A R T I C L E I N F O A B S T R A C T

Keywords: In ultra-precision machining, the accuracy of a workpiece can be generally improved by surface measurement
Ultra-precision machining and compensation machining. Since on-machine surface measurement can avoid the errors caused by reposi­
On-machine measurement tioning workpieces and use the machine axes to extend the measuring range, there has been a trend from offline
Confocal sensor
measurement to on-machine measurement in recent years. However, the workpiece surfaces inspected by on-
Nonlinear least squares method
machine measurement system in previous studies generally have small slopes and depths. In this paper, an
on-machine measurement device based on a chromatic confocal sensor is designed, which can inspect workpiece
surfaces with larger slopes and depths. The on-machine measurement system and measurement procedure are
introduced first. A nonlinear least squares method is then used to further reduce the adverse influence of the
alignment error between the axes of the sensor and the spindle of the machine tool. Finally, machining and
measuring experiments of the spherical and free-form surfaces are performed using a Nanotech 650FG ultra-
precision machine lathe to validate the proposed on-machine measurement method. The profile pattern and
the form error obtained from the on-machine measurement device are compared with those obtained from a
Taylor-Hobson 3D stylus profilometer. There is good agreement between the two results, which verifies the
effectiveness of the proposed on-machine measuring method.

1. Introduction machine structural errors, vibration and thermal deformation, may


cause a machined surface to deviate from the design surface [10]. Thus,
With the development of modern industry, surfaces with sub­ surface inspection and compensation machining are indispensable to
micrometric form accuracy and nanometric surface roughness are improve machining accuracy. The offline measurement method, which
widely used in optics [1,2], machinery manufacturing, biomedicine [3], involves moving the parts from the machine tools to the measurement
and national defence and the military [4]. Corresponding ultra-precision instruments, was proposed earlier and has been intensively investigated
machining technologies have also attracted attention [5]. Currently, a [11]. Although the offline measurement method usually has the ad­
variety of techniques have been developed to fabricate these surfaces, vantages of high precision, reliable performance and a large measuring
mainly including laser processing technology [6], micro-electrical range, moving and repositioning the workpiece between the measure­
discharge machining technology [7], micro-grinding technology [8] ment and manufacturing processes may significantly affect the mea­
and particle or electron beam processing technology [9]. However, surement accuracy and efficiency [12]. Therefore, there has been a trend
mechanical machining, such as ultra-precision single-point diamond from offline measurement to on-machine measurement(OMM) to in­
cutting technology, still plays a key role in high-precision surface crease measurement efficiency and accuracy [13]. On-machine surface
fabrication, providing the advantages of low surface roughness, high inspection can avoid the errors caused by remounting workpieces and
shape accuracy and manufacturing efficiency. can use the machine axes to extend measuring ranges. The measuring
Although ultra-precision machine tools can achieve high motion path can also be followed conveniently by using the CNC system of the
accuracy, some external factors, including environmental factors, machine tool. OMM methods can mainly be classified into contacting

* Corresponding author.
E-mail address: zhangxinquan@sjtu.edu.cn (X. Zhang).

https://doi.org/10.1016/j.precisioneng.2021.11.015
Received 27 July 2021; Received in revised form 22 October 2021; Accepted 29 November 2021
Available online 8 December 2021
0141-6359/© 2021 Elsevier Inc. All rights reserved.
Z. Chen et al. Precision Engineering 74 (2022) 232–241

and optical non-contacting methods according to the sensor working until the distance between the probe and the sphere along the Z-axis was
principles [14]. The contacting method typically utilizes a stylus or minimized. The sensor axis and the spindle axis were considered to be
probe to scratch the surface of the component to obtain the section coincident at that position [27].
profile. Suzuki et al. developed a new contact-type OMM system to Although the principle and application of chromatic confocal sensors
measure aspherical optical parts. The contact angle between the probe have been investigated by many researchers, there is little research on
axis and the contact surface was kept constant to reduce variation in the their application to OMMs for ultra-precision machining. Meanwhile,
probe friction force [15]. Chen et al. proposed a compensation approach the chromatic confocal sensor was mostly employed to measure curved
for grinding tungsten carbide aspheric moulds based on on-machine surfaces with small depths in previous work, only tens of microns to
contact measurements [16]. Since the radius of the conventional con­ hundreds of microns. In this research, we design an on-machine mea­
tact probe usually lies in the millimetre range, which limits the mea­ surement device based on the chromatic confocal sensor, which enables
surement resolution, atomic force microscopy(AFM) and scanning collecting the sensor data and the real-time position data of motion axes
tunnelling microscopy(STM) have also been integrated into OMM sys­ synchronously. The sensor moves through the Z-axis during the mea­
tem [10]. Gao et al. designed and constructed an AFM-head for the surement process to ensure that the distance between the sensor and the
surface profile measurement of sinusoidal XY-grid workpieces on a workpiece surface remains unchanged. Therefore, the measurement
diamond turning machine [17]. Zhu et al. developed a new OMM system range is theoretically the same as the Z-axis motion range of the machine
based on the scanning tunnelling microscope principle for precision tool. Then the proposed OMM system is applied to the inspection of a
form control of machined microstructures [18]. Nonetheless, contact spherical surface and a free-form surface with the depths of several
between the probe and the inspected surface is unavoidable with these millimeters respectively, which exceed the original range of the sensor.
methods. As a result, the measured surface may be damaged to some Considering the influence of the alignment error between the axes of the
degree by the probes [19]. Meanwhile, the measurement efficiency is sensor and the spindle, a flat surface with an annular groove, rather than
also limited. Therefore, non-contacting OMM methods have been a spherical surface, is used to calibrate the sensor position first, and the
developed to avoid these disadvantages. nonlinear least squares method is proposed to further calculate the
Noncontact OMM methods share the advantages of fast data acqui­ alignment error accurately from the measurement data, which effec­
sition and a non-destructive nature [10]. Interferometry is a mainstream tively improves the accuracy of the measurement result.
optical OMM system thanks to its nanometric precision. Li et al. inte­ The remainder of this paper is organized as follows. The proposed
grated an interferometric probing system on an ultra-precision turning OMM system configuration and the measurement procedure are
machine. To improve the performance of the proposed OMM system, described in Section 2. The compensation of the alignment error based
they calibrated and compensated for several systematic errors, including on the nonlinear least squares method is presented in Section 3. The
on-machine vibration, machine tool kinematic errors and linearity er­ experimental setup and procedures are described in Section 4. The
rors [13,20]. Finally, the measurement system was applied to the in­ experimental results and discussion are presented in Section 5. Finally,
spection and characterization of microlens array features [21]. the conclusion is summarized Section 6.
However, the interferometry is inconvenient to inspect reflecting sur­
faces with large slopes according to its measuring principle. The 2. Measurement methodology
maximum sensor tilt angles of the current commercial
interferometric-based sensors are only a few degrees, such as the inter­ 2.1. Measurement system
ferometer IMS5x00 series products developed by MICRO-EPSILON
(German). The chromatic confocal technology is an alternative way In this research, a chromatic confocal sensor was integrated into an
for non-contacting surface measurement with high accuracy and ultra-precision machine tool to measure the machined components.
increasingly utilized in the ultra-precision machining applications. This Fig. 1 shows the structural configuration of the proposed OMM system.
unique measuring system makes use of the controlled chromatic aber­ The machine tool is equipped with three linear hydrostatic axes (X-, Y-
ration to focus each light wavelength at a specific distance and then and Z-axis) and two rotational axes (B- and C-axis). The chromatic
receive the light that is reflected from the target surface and transfer it to confocal sensor is installed with a fixture on the platform of the B-axis.
the controller. The reflected light is focused back on to a detector This allows it to be driven to move along the Z-axis and rotate around the
through a pinhole. If the focal point of the probe is in focus, the reflected B-axis. Meanwhile, the workpiece is held on a vacuum chuck, which can
light passes through the pinhole and is detected as an intensity peak. be driven to move along the X- and Y-axes and rotate around the C-axis.
Then the spectrometer can detect the intensity and the wavelength at Thus, arbitrary three-dimensional movements of the sensor relative to
which the maximum of intensity occurs. Therefore, the wavelength at the workpiece can be achieved and the whole surface can be inspected. A
which the maximum of intensity occurs can be associated with the diamond cutting tool is fixed on a tool holder, which is installed on the
surface height through calibration [22]. Based on the measurement other side of the B-axis platform. When machining, the B-axis platform
principle, its maximum detectable slope of the workpiece surface rea­
ches tens of degrees and is much bigger than that of
interferometric-based sensors. Nouira et al. developed a calibration
gauge for the characterization of a chromatic confocal probe, and the
proposed system could identify major error sources [23]. Rishikesan
et al. introduced a chromatic confocal displacement sensor for the
estimation of surface profile parameters [24]. Quinsat et al. presented an
approach using a confocal chromatic sensor to carry out on-machine
measurements of the surface roughness on a five-axis milling center
[25]. Keferstein et al. proposed a new optical system based on a chro­
matic confocal sensor to take measurements on a grinding machine for
quality assurance and optimization of the grinding process [26]. Zou
et al. integrated a chromatic confocal measurement sensor with an
ultra-precision diamond turning machine, and the measurement results
of the proposed OMM method were compared with those of the stylus
method. Before measurement, the sensor was moved along the X- and
Y-axes to scan the surface of a standard sphere affixed in a vacuum chuck Fig. 1. Structural configuration of the on-machine measurement system.

233
Z. Chen et al. Precision Engineering 74 (2022) 232–241

rotates 180◦ to make the cutting tool face the workpiece. Therefore, the platform is rotated manually to make the sensor axis parallel to the
sensor does not need to be installed and removed repeatedly, which spindle axis. Here, a micrometer is utilized to judge whether the two
improves the measurement efficiency. Additionally, the sensor and the axes are parallel by placing the probe of the micrometer on the
fixture will not be damaged during the machining process. The chro­ sensor. If the two axes are parallel, the displayed value of the
matic confocal probe is connected to its controller through an optical micrometer should be constant when the sensor moves along the Z
fibre, which is then connected to an external Power PMAC controller direction. Since the B-axis platform does not rotate during the
with a network cable for sensor data acquisition and storage, as pre­ measurement process, the sensor axis does not need to intersect with
sented in Fig. 2. The corresponding encoder signal output modules of the the B-axis.
machine tool axes are also connected to the Power PMAC controller to 2 The calibration cylinder is installed on the vacuum chuck, and the
collect the real-time position data of the machine tool axes at the same eccentric error between the axes of the calibration cylinder and the
time. Therefore, the sensor data can be matched with the position data spindle is measured using the micrometer. The position of the cali­
and the whole surface can be inspected. bration cylinder is adjusted with a rubber hammer until the eccentric
error is less than 0.1 μm.
3 The end face of the cylinder with an annular groove is measured by
2.2. Sensor position calibration scanning the sensor along the X-axis as shown in Fig. 3(a), and the X-
axis coordinates are recorded when the sensor output changes
Before measuring, the relative position of the sensor and the machine sharply, denoted as X1 and X2. Then, the surface along the Y-axis is
tool needs to be calibrated. Generally, the sensor axis is adjusted to align measured, as shown in Fig. 3(b), and the Y-axis coordinates are
with the spindle axis of the machine tool at first. In previous research, recorded when the sensor output changes sharply again, denoted as
spherical surfaces were employed by some scholars for sensor position Y1 and Y2.
calibration. However, considering the small curvature near the extreme 4 The X- and Y-axis coordinates of the machine tool are adjusted to
point of the spherical surface, the variation of the measurement value is X1 +X2
and Y1 +Y
2 , respectively. In this position, the sensor axis in­
2
2
only one hundredth of the displacements along the X- and Y-axes. tersects with the spindle axis. Since the sensor axis has been adjusted
Additionally, external noises affects the measurement results. Therefore, to be parallel to the spindle axis in step 1, the two axes are considered
it is difficult to ensure that the sensor axis coincides with the spindle axis to coincide at this position. Finally, the current position in the CNC
accurately using this method. In this research, a two-step calibration system of the machine tool is saved for subsequent use. The align­
method is proposed. A flat surface with an annular groove is used to ment error will be further reduced with the method in Section 3.2.
calibrate the sensor position first, which can reduce the alignment error
to a few microns and will be illustrated in this section. Then the
nonlinear least square method is proposed to further reduce the align­ 2.3. Surface measurement and reconstruction
ment error to sub-micron level by using the measurement data of a
freeform surface, which will be introduced in Section 3.2. After completing the calibration of the sensor position, the B-axis
In the first step, a flat surface with an annular groove is scanned by platform needs to be rotated around the B-axis to keep the probe away
the sensor along the X-axis and Y-axis as shown in Fig. 3 and the posi­ from the machining zone. Then, the machine tool can be used for
tions of the two linear axes are recorded when the sensor data changes workpiece machining. After machining, the workpiece remains on the
sharply. Then, the position of the sensor axis relative to the spindle axis spindle, and the sensor is moved back to the recorded position deter­
can be calculated. The detailed calibration steps are as follows: mined in the calibration process. Before surface measurement, the initial
relative distance between the sensor and the surface should be within
1 The confocal sensor is first installed on the B-axis platform with an the measuring range when moving the probe along the Z-axis. In this
elaborately designed and machined fixture. By ensuring the flatness way, the sensor, workpiece and cutting tool do not need to be repeatedly
of the mating planes of the fixture via ultra-precision grinding, the installed and disassembled. Meanwhile, the sensor only needs to be
sensor axis is parallel to the B-axis platform plane. Then, the B-axis calibrated once to complete the machining, measurement and
compensation processes.
In this study, a spiral trajectory is adopted to inspect the spherical
and free-form surfaces. Through the synchronous movements of the X-
and C-axes, the confocal sensor scans the workpiece surface in a spiral
scanning pattern from inside to outside of the measurement zone.
Considering that the depth of the workpiece surface along the Z-axis
may exceed the sensor measuring range, the sensor needs to be moved
synchronously along the Z-axis according to the theoretical surface
model and the coordinates of the X- and C-axes. During the measure­
ment process, the coordinates of the machine tool axes and the distance
data of the sensor are recorded synchronously in real time using the
Power PMAC controller. The maximum acquisition frequency can reach
2 kHz.
Before measuring, the probe axis is adjusted to align with the spindle
axis and the distance between the sensor and the measured surface is
adjusted to be within the measurement range, as mentioned above. The
initial sensor position is denoted as S0, and the corresponding mea­
surement point on the surface is denoted as p0, as shown in Fig. 4.
During the measurement, the sensor moves to position Si along the Z-
axis, and the corresponding measurement point on the surface is deno­
ted as pi at time i. When reconstructing the surface based on the mea­
surement data, using point p0 as the origin of the workpiece coordinate
system, the coordinates of point pi can be expressed in the polar coor­
Fig. 2. Data acquisition scheme of the on-machine measurement system. dinate system as follows:

234
Z. Chen et al. Precision Engineering 74 (2022) 232–241

Fig. 3. Schematic diagram of the calibration process using a calibration cylinder with an annular groove on the end face.

Fig. 4. Schematic diagram of the measurement process using a spi­


ral trajectory.

r p i = Xi − X0 (1) Fig. 5. The influence of the axis alignment error on measurement points.

θ pi = Ci − C0 (2)
spindle axis after calibration, the origin of the coordinate system for the
spiral scanning measurement would move from O to O’. The deviations
Zpi = (Di + Zi ) − (D0 + Z0 ) (3)
in the X direction and the Y direction are denoted as Δx and Δy,
where Xi, Ci, and Zi are the coordinates of the X-, C- and Z-axes at time i, respectively. After the preliminary calibration with an annular groove,
X0, C0, and Z0 are the coordinates of the X-, C- and Z-axes at the initial the values of these two deviations are on the order of a few microns. the
actual measurement points (p1 , p2 , p3 , p4 , …) will also be different
′ ′ ′ ′
time, and Di and D0 are the sensor output data collected at time i and the
initial time, respectively. from the theoretical ones (p1 , p2 , p3 , p4 , …), which may cause the
The measurement point pi can also be expressed in the Cartesian distortion of the measured surface profile. The adverse influence will be
coordinate system as follows: enhanced when the slope of the measured surface increases. To discuss
the influence of the alignment error on the measurement results when O’
Xpi = rpi ∗ cos(θpi ) = (Xi − X0 ) ∗ cos(Ci − C0 ) (4) is in different quadrants, the deviation is expressed in terms of the polar
coordinates (Δr, Δθ) as follows:
Ypi = rpi ∗ sin(θpi ) = (Xi − X0 ) ∗ sin(Ci − C0 ) (5) √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
Δr = Δx2 + Δy2 (7)
Zpi = (Di + Zi ) − (D0 + Z0 ) (6)
Δy
Δθ = arctan( ) (8)
3. Alignment error compensation method Δx
Therefore, the value of Δr is on the order of a few microns, and the
3.1. Analysis of the alignment error range of Δθ is from 0 to 360◦ .
Fig. 6 and Fig. 7 present the simulation results of the profile distor­
Considering the machining error of the calibration gauge and mea­ tions caused by the axis alignment error when measuring a spherical
surement noises, the alignment error can only be reduced to a few mi­ surface with a radius of 30 mm and a chord height of 1.7 mm. The error
crons after the calibration step with an annular groove presented in component Δr is increased from 0 to 40 μm, and Δθ is set to 45◦ in the
Section 2.2. Therefore, the nonlinear least squares method is employed simulations summarized in Fig. 6. The error component Δr is set to 20
to reduce the alignment error further in this section. The influence of the μm, and Δθ is increased from 0◦ to 180◦ in the simulations summarized
axis alignment error on the measurement results is analyzed first. As in Fig. 7. The maximum error of the simulated profile is nearly 10 μm
shown in Fig. 5, if the sensor axis does not coincide exactly with the

235
Z. Chen et al. Precision Engineering 74 (2022) 232–241

||p − q|| = min ||p − x|| (9)


x∈S

Then, the distance between point p and surface S is defined as:


d(p) = ||p − q|| (10)
Clearly, when point q is located on the S, point p is located on the
normal line of surface S at point q, as shown in Fig. 8. That is
p − q = ±d(p)nq (11)

where nq is the unit normal vector of the surface S at point q. The sign
depends on whether nq points to the same side of point p or the opposite
side. Point q is designated as the projection of point p on surface S.
Furthermore, the directed distance from point p to surface S can be
defined as [28]:
ds (p) = (p − q)nq (12)
When the axis alignment error is zero, since the theoretical mea­
surement point pi is on the spherical surface S, point pi coincides with
projection point qi, and the directed distance from point pi to surface S
Fig. 6. Simulation results of profile distortions caused by the alignment error is:
when measuring a ball with a radius of 30 mm (the error component Δθ = 45◦ ).
ds (pi ) = (pi − qi )nqi = 0 (13)

and caused by an axis alignment error of 40 μm, which is unacceptable When the axis alignment error is considered, the actual measurement
in ultra-precision measurement. point is pi . Although point pi is also on surface S, the X and Y co­
′ ′

ordinates of pi are determined using the position of the machine axes


3.2. Calibration principle and will be affected by the axis alignment error when reconstructing the
surface, as shown in Fig. 9. Thus, point pi ′′ used in the actual calculation
To further reduce the alignment error, a freeform surface, which is deviates from point pi with an increment Δp. As shown in Fig. 5, the

machined with an ultra-precision machine lathe, is measured again on axis alignment error is characterized by Δr and Δθ. Denoting the coor­
the basis of the previous calibration, and the nonlinear least squares dinate of the C-axis at time i as Ci, Δp can be expressed in the Cartesian
method is utilized to calculate the alignment error accurately using the coordinate system defined in Section2.3 as:
measurement data. The part is attached to the vacuum chuck while the Δp = (Δr ∗ cos(Ci − Δθ), Δr ∗ sin(Ci − Δθ), 0) (14)
surface axis is adjusted to coincide with the spindle axis.
s
Then, we define the distance function from the point to the surface as The directed distance from point pi to surface S is denoted as d (pi ′′ ).
′′

follows: Given a point p ∈ R3 and a smooth regular surface S ∈ R3 in the To calculate Δr and Δθ, the sum of the squared distance of all the
space, there is at least one point q ∈ S, which satisfies measuring points should be minimized. Therefore, the optimization

Fig. 7. Simulation results of profile distortions caused by the alignment error when measuring a ball with a radius of 30 mm (the error component Δr = 20um).

236
Z. Chen et al. Precision Engineering 74 (2022) 232–241

by the alignment error is added to point p, and the Z coordinate of the


new point p′ is calculated using the free-form surface equation. 3) The
coordinates of point p′′ can be obtained by combining the X and Y co­
ordinates of point p and the Z coordinate of the corresponding point p′ ,
which are the final measurement data used in the simulation.
As shown in Fig. 10(a), the maximum simulation error is nearly 3.3
μm due to the axis alignment error. Using the proposed method intro­
duced in Section 2.2, the alignment error components Δr and Δθ can be
calculated, and the corresponding deviation can be compensated. The Δr
and Δθ calculated using the simulation measurement date are 20 μm and
45◦ , respectively, which are exactly equal to the theoretical values. After
compensation, the distortion caused by the axis alignment error is
almost eliminated, as shown in Fig. 10(b).
In addition, to consider the environmental factors, random noise
with a mean of 0 and a standard deviation of 0.5 μm are added to the
simulation measurement data. The simulation measurement points
containing random noise are shown in Fig. 11(a). Fig. 11(b) shows that
Fig. 8. Schematic diagram of the distance from the point to the surface. the maximum error of the calculation result is 4.17 μm if the compen­
sation method is not applied in the surface reconstruction. By using the
proposed method, the Δr and Δθ calculated with the simulation mea­
surement date are 19.896 μm and 44.657◦ , respectively, which are close
to the theoretical values and satisfy the accuracy requirements. The
maximum simulation error decreases to 1.5 μm, as shown in Fig. 11(c).
The error can be further reduced by using a Gaussian filter. Fig. 11(d)
shows that the final maximum resulting error is less than 0.2 μm. The
influence of the axis alignment error is almost eliminated, and the re­
sidual error is mostly caused by random noise. To verify the repeat­
ability of simulation, the above simulation process is repeated five times.
The calibration errors in the X and Y directions are all at the sub-micron
level and less than 0.3 μm, which shows good simulation repeatability.

4. Experimental setup

Machining and measuring experiments were carried out on a


Nanotech 650FG ultra-precision machine tool to validate the proposed
on-machine measurement system and method. Fig. 12 illustrates the
integrated measurement system on the Nanotech 650FG machine tool.
Fig. 9. Schematic diagram of the position error of measuring point pi.
The relevant parameters of the machine tool are listed in Table 1. The
confocal sensor model employed in this research is IFS2405–0.3(MICRO
objective function can be expressed as follows: EPSILON, Germany). The measuring range and the maximum tilt angle
∑ of the confocal sensor are 0.3 mm and 34◦ , respectively. The measure­
min E(Δr, Δθ) = [ds (pi ′′ + Δpi )]2 (15)
i ment resolution along the axis is 10 nm. The chromatic confocal sensor
was installed on the B-axis platform with an elaborately designed
In this research, the Levenberg-Marquardt algorithm(LMA) is adopted to fixture. The relative position of the sensor with respect to the machine
obtain the optimal solution to the above optimization problem. The LMA tool was calibrated first using the method described in Section 2.3 with a
interpolates between the Gauss–Newton algorithm (GNA) and the calibration cylinder attached via the vacuum chuck. Then, the nonlinear
gradient descent method. Thus, the LMA is more robust than the GNA, least squares method was utilized to further calculate the alignment
which means that in many cases, it finds a solution even if it starts very error, as illustrated in Section 3, using the freeform surface. After cali­
far from the final minimum. The readers can refer to Ref. [29] for a brating the probe position, the B-axis platform was rotated to keep the
detailed description of this method.As deduced in Ref. [28], the gradient probe away from the machining zone, and the workpiece was installed
vector of the distance function ∇ds(p), which can be used in the LMA, is on the spindle for machining. After machining, the machined workpiece
presented as follows: was retained on the vacuum chuck and the probe was returned to the
recorded position to start measuring with a programmed spiral
∇ds (p) = nq (16)
trajectory.
The surfaces machined and inspected in the experiments included
3.3. Simulation result both spherical and free-form surfaces, as shown in Fig. 13. The spherical
radius of the former workpiece was 30 mm and the chord height was
A simulation study is presented to verify the compensation method. approximately 1.7 mm. The free-form surface can be expressed by XY
The nominal surface is a freeform surface with a length of 10 mm and a polynomials. The length and width of the free-form surface were 32 mm
width of 10 mm. The free-form surface can be expressed by XY poly­ and 25 mm respectively. The depth in the Z direction was approximately
nomials as z = 0.008*x2 + 0.006*y2. The alignment error components 3 mm. During the measurement process, the feed speed of the X-slide
Δr and Δθ are set to 20 μm and 45◦ , respectively, in the simulation. The was 5 mm/min, and the rotation speed of the spindle was 100 rpm. The
measurement data are simulated as follows: 1) The simulated mea­ feed speed of the Z-slide varied according to the theoretical measure­
surement point p is selected every two degrees along the spiral trajec­ ment surface profile. The sampling frequency for axis positions and
tory, as shown in Fig. 4, and the pitch of the spiral is set to 40 μm. The sensor data acquisition was set to 2 kHz. The entire measurement times
total number of simulation points is 63 000. 2) The deviation Δp caused of the spherical and free-form surfaces were approximately 2 min and 4

237
Z. Chen et al. Precision Engineering 74 (2022) 232–241

Fig. 10. Simulation results of profile distortions caused by the alignment error when measuring the freeform surface(Δr = 20 μm, Δθ = 45◦ ). (a) Calculated error
distribution without compensation. (b) Calculated error distribution with compensation.

Fig. 11. Simulation results of profile distortions caused by alignment error (Δr = 20 μm, Δθ = 45◦ ) with the consideration of random noise (μ = 0, σ = 0.5). (a)
Simulation measurement points containing random noise. (b) Error distribution after fitting without compensation. (c) Error distribution after fitting with
compensation. (d) Error distribution after fitting with compensation and Gaussian filtering.

Table 1
The relevant parameters of the machine tool.
Axis Programming Feedback Maximum Straightness Positioning
resolution Resolution Speed Accuracy

X 0.01 nm 0.008 nm 2000 mm/ 0.3 μm over /


min full travel
Y 0.01 nm 0.008 nm 2000 mm/ 0.3 μm over /
min full travel
Z 0.01 nm 0.008 nm 2000 mm/ 0.3 μm over /
min full travel
B 0.000 000 1◦ 0.005 arc 50 rpm / ±1.0 arc
seconds seconds
C 0.000 000 1◦
0.005 arc 3000 rpm / ±1.0 arc
seconds seconds

min, respectively. After the machined workpieces were measured


through the proposed OMM system, they were detached from the vac­
uum chuck and measured using a Taylor-Hobson 3D stylus profilometer.

Fig. 12. The integrated measurement system on the Nanotech 650FG ma­ 5. Results and discussions
chine tool.
After completing the measurements, the results provided by the two

238
Z. Chen et al. Precision Engineering 74 (2022) 232–241

Fig. 13. The spherical and free-form surfaces inspected in the experiments.

measurement instruments were compared and analyzed. The measured As for the measurement uncertainty, since the machine tool is placed
spherical surfaces and form errors obtained from the two measurement in a room with constant temperature and humidity, the effect of tem­
methods are shown in Fig. 14. The measured free-form surfaces and form perature and humidity can be neglected; the sensor is fixed on the ma­
errors obtained from the two measurement methods are shown in chine tool after calibration, thus avoiding the influence of reinstallation
Fig. 15. The profile patterns obtained from the OMM system agree well errors; with the calibration method, the axis alignment error is at the
with those obtained from the 3D stylus profilometer, demonstrating the sub-micron level, and its influence on the measurement uncertainty is
consistency of the measurement results. also negligible; meanwhile, the ultra-precision machine tool Nanotech
To compare the measurement results more accurately, the form er­ 650FG used in this research has been accurately calibrated, and its
rors of the two surfaces characterized in terms of the root mean square motion accuracy and verticality meet the measurement accuracy re­
(RMS) and peak-to-valley (PV) values calculated from the measurement quirements. Therefore, the main factors that affect the measurement
data are summarized in Table 2. For both the spherical and free-form uncertainty are the resolution of the sensor and the vibration when the
surfaces, the deviations of the RMS values obtained from the two mea­ sensor tracks the Z-axis movement. The linear resolution of the chro­
surement methods are within 0.2 μm and the deviations of the PV values matic confocal sensor is 10 nm. The vibration error can be obtained by
do not exceed 0.4 μm, which shows that the proposed OMM method comparing the Z-axis coordinate of the machine tool with the sensor
meets the accuracy requirements for ultra-precision machining appli­ distance value when measuring a fixed point with a moving Z-axis. The
cations. To verify the repeatability of experimental results, the mea­ measured vibration error is about 100 nm, which can be further reduced
surement of the spherical surface and the freeform surface is also by designing more stable fixture.
repeated three times. The measurement data are summarized in Table 3,
which shows good measurement repeatability.

Fig. 14. Spherical surface measurement results by on-machine and offline measurements. (a)Measured spherical surface and form error obtained by on-machine
measurement. (b)Measured spherical surface and form error obtained by offline measurement.

239
Z. Chen et al. Precision Engineering 74 (2022) 232–241

Fig. 15. Free-form surface measurement results by on-machine and offline measurements. (a)Measured free-form surfaces and form error obtained by on-machine
measurement. (b)Measured free-form surfaces and form error obtained by offline measurement.

the work reported in this paper.


Table 2
Form error parameters of the on-machine and offline measurements.
Acknowledgement
Sample Parameter On-machine Offline Deviation
measurement measurement
This work was partially supported by the National Natural Science
Spherical RMS/um 1.493 0 1.320 0.173 0 Foundation of China under Grants 52 075 332, U2013211 and
surface PV/um 5.248 0 4.886 0.362 0
Freeform RMS/um 0.209 6 0.062 1 0.147 5
51 975 375, the Strategic Priority Research Program of Chinese Acad­
surface PV/um 0.883 9 0.531 9 0.352 0 emy of Sciences under Grants XDA25040000, XDA25040200 and
XDA25040202and Shanghai Pujiang Program under Grants
19PJ1404500.
Table 3
Form error parameters of the repeated on-machine measurements. References
Sample Parameter No.1 No.2 No.3 [1] Kim Jiyoon, Kim Se-Um, Lee Bo-Yeon, Suh Jeng-Hun, Lee Sin-Doo. Lenticular lens
Spherical surface RMS/um 1.493 0 1.415 1.433 array based on liquid crystal with a polarization-dependent focusing effect for
PV/um 5.248 0 4.875 5.126 2d–3d image applications. J. Inf. Disp. 2015;16(1):11–5.
[2] Xiang Li, Tian Lan, Wang Yun, Wang Long-Hui. Design and study of fresnel lens for
Freeform surface RMS/um 0.209 6 0.186 0.214
an antenna in indoor visible light communication system. Acta Phys Sin 2015;64:
PV/um 0.883 9 0.718 0.873
024201.
[3] Miyoshi Hiromi, Ju Jungmyoung, Lee Sang Min, Cho Dong Jin, Ko Jong Soo,
Yamagata Yutaka, Adachi Taiji. Control of highly migratory cells by
6. Conclusion microstructured surface based on transient change in cell behavior. Biomaterials
2010;31(33):8539–45.
[4] Yang Zhi, Dai Yi-fan, Zhang Pei. The study of hybrid refractive-diffractive
This paper develops an OMM device based on a chromatic confocal telephoto lens. Laser Technol 2007;31(2):206–8.
sensor. According to the measuring principle of the sensor, the work­ [5] Lee Wing Bun, Kong Ling bao, Cheung Chi Fai, To Suet, Chen Xin, Liu Qiang. An
piece surfaces with large slopes and depths can be inspected using this overview of ultra-precision diamond machining of microstructured freeform
surfaces. Jixie Gongcheng Xuebao/Journal of Mechanical Engineering 2013;49
measurement system. To consider the adverse influence of the alignment
(19):144–55.
error between the axes of the sensor and the spindle, a two-step cali­ [6] Vorobyev AY, Guo Chunlei. Multifunctional surfaces produced by femtosecond
bration method is proposed, and a nonlinear least squares method is laser pulses. J Appl Phys 2015;117(3):033103.
developed for fine calibration. The proposed OMM system and method [7] Pham Duc Truong, Dimov Stefan Simeonov, Bigot Samuel, Ivanov Atanas, Popov K.
Micro-edm—recent developments and research issues. J Mater Process Technol
are validated by machining and measuring experiments on both spher­ 2004;149(1–3):50–7.
ical and free-form surfaces. The measurement results show that the [8] Aurich JC, Engmann J, Schueler GM, Haberland R. Micro grinding tool for
profile patterns and form errors obtained from the proposed OMM de­ manufacture of complex structures in brittle materials. CIRP annals 2009;58(1):
311–4.
vice are consistent with those obtained from an offline measurement [9] Xu ZW, Fang FZ, Fu YQ, Zhang SJ, Han T, Li JM. Fabrication of micro/nano-
device, which verifies the effectiveness of the proposed OMM system structures using focused ion beam implantation and xef2 gas-assisted etching.
and method. There are sub-micron deviations between the results of the J Micromech Microeng 2009;19(5):054003.
[10] Li Duo, Wang Bo, Tong Zhen, Blunt Liam, Jiang Xiangqian. On-machine surface
on-machine and offline measurements, which may be caused by the measurement and applications for ultra-precision machining: a state-of-the-art
vibration and noise during the measurement and still need to be review. Int J Adv Manuf Technol 2019;104(1):831–47.
explored in the future. [11] Kim Ho-Sang, Lee Kwang-Il, Lee Kyoung-Min, Bang Young-Bong. Fabrication of
free-form surfaces using a long-stroke fast tool servo and corrective figuring with
on-machine measurement. Int J Mach Tool Manufact 2009;49(12–13):991–7.
[12] Zhang Xiaodong, Zeng Zhen, Liu Xianlei, Fang Fengzhou. Compensation strategy
Declaration of competing interest for machining optical freeform surfaces by the combined on-and off-machine
measurement. Opt Express 2015;23(19):24800–10.
[13] Li Duo, Tong Zhen, Jiang Xiangqian, Blunt Liam, Gao Feng. Calibration of an
The authors declare that they have no known competing financial interferometric on-machine probing system on an ultra-precision turning machine.
interests or personal relationships that could have appeared to influence Measurement 2018;118:96–104.

240
Z. Chen et al. Precision Engineering 74 (2022) 232–241

[14] Li Duo, Jiang Xiangqian, Tong Zhen, Blunt Liam. Development and application of [21] Li Duo, Wang Bo, Zheng Qiao, Jiang Xiangqian. Ultraprecision machining of
interferometric on-machine surface measurement for ultraprecision turning microlens arrays with integrated on-machine surface metrology. Opt Express 2019;
process. J Manuf Sci Eng 2019;141(1). 27(1):212–24.
[15] Suzuki H, Onishi T, Moriwaki T, Fukuta M, Sugawara J. Development of a 45 [22] Leach Richard. Optical measurement of surface topography, vol. 8. Springer; 2011.
degrees tilted on-machine measuring system for small optical parts. Annals of the [23] Nouira H, El-Hayek N, Yuan X, Anwer N. Characterization of the main error
CIRP 2008;57(1):414. sources of chromatic confocal probes for dimensional measurement. Meas Sci
[16] Chen FJ, Yin SH, Huang H, Ohmori H, Wang Y, Fan YF, Zhu YJ. Profile error Technol 2014;25(4):044011.
compensation in ultra-precision grinding of aspheric surfaces with on-machine [24] Rishikesan V, Samuel GL. Evaluation of surface profile parameters of a machined
measurement. Int J Mach Tool Manufact 2010;50(5):480–6. surface using confocal displacement sensor. Procedia Materials Science 2014;5:
[17] Gao Wei, Aoki Jun, Ju Bing-Feng, Kiyono Satoshi. Surface profile measurement of a 1385–91.
sinusoidal grid using an atomic force microscope on a diamond turning machine. [25] Quinsat Yann, Tournier Christophe. In situ non-contact measurements of surface
Precis Eng 2007;31(3):304–9. roughness. Precis Eng 2012;36(1):97–103.
[18] Zhu Wu-Le, Yang Shunyao, Ju Bing-Feng, Jiang Jiacheng, Sun Anyu. Scanning [26] Keferstein Claus P, Honegger Daniel, Thurnherr Hugo, Gschwend Bernard. Process
tunneling microscopy-based on-machine measurement for diamond fly cutting of monitoring in non-circular grinding with optical sensor. CIRP annals 2008;57(1):
micro-structured surfaces. Precis Eng 2016;43:308–14. 533–6.
[19] Zhu Yongjian, Jingxin Na, Pan Weiqing, Zhi Yanan. Discussions on on-machine [27] Zou Xicong, Zhao Xuesen, Guo Li, Li Zengqiang, Sun Tao. Non-contact on-machine
measurement of aspheric lens-mold surface. Optik 2013;124(20):4406–11. measurement using a chromatic confocal probe for an ultra-precision turning
[20] Li Duo, Jiang Xiangqian, Tong Zhen, Blunt Liam. Kinematics error compensation machine. Int J Adv Manuf Technol 2017;90(5):2163–72.
for a surface measurement probe on an ultra-precision turning machine. [28] Han Ding, Zhu Limin. Geometric theories and methods for digital manufacturing of
Micromachines 2018;9(7):334. complex surfaces. Science Press; 2011.
[29] Ranganathan Ananth. The levenberg-marquardt algorithm. Tutoral on LM
algorithm 2004;11(1):101–10.

241

You might also like