Professional Documents
Culture Documents
Precision Engineering: Xiaozhong Guo, Zhaoyao Shi, Bo Yu, Baoya Zhao, Ke Li, Yanqiang Sun
Precision Engineering: Xiaozhong Guo, Zhaoyao Shi, Bo Yu, Baoya Zhao, Ke Li, Yanqiang Sun
Precision Engineering
journal homepage: www.elsevier.com/locate/precision
MSC: Gears are core components of transmission systems. The current trend of gear measurements is the rapid and
00-01 full digitization. To rapidly obtain information on the three-dimensional (3D) shape of the gear tooth flank,
99-00 a 3D point cloud measurement system based on a line structure light sensor and high precision air floating
Keywords: rotary table is proposed. The system can acquire more than 1.5 million 3D data points on one side of the gear
Gear measurement tooth flank within less than 5 s. In this study, the measured 3D point cloud data is used to calculate the profile
3D point cloud error, pitch error, then compared to those obtained from traditional contact measurements. In addition, tooth
Line structured light sensor
flank error was evaluated. The results show that the 3D point cloud measurement system can perform fast and
Involute error
accurately in 3D measurements of gears and it is a novel system for gear measurement and error calculation.
∗ Corresponding author.
E-mail address: shizhaoyao@bjut.edu.cn (Z. Shi).
https://doi.org/10.1016/j.precisioneng.2019.10.013
Received 31 July 2019; Received in revised form 11 October 2019; Accepted 17 October 2019
Available online 19 October 2019
0141-6359/© 2019 Elsevier Inc. All rights reserved.
X. Guo et al. Precision Engineering 61 (2020) 160–169
Table 1
Rotary table geometry and precision parameters.
Parameter Value
Diameter 500 mm
Variable Speed 0.1 ∼ 5 rpm
Torque 20 Nm continuous/93 Nm peak
Flatness of Platen Surface 0.005 mm
Runout of Platen Surface 0.005 mm
Parallelism of Tabletop to Base 0.010 mm
Angular System Resolution 0.2 arc second 0.0001◦
Angular Repeatability ±0.2 arc second
Angular indexing error ±3 arc seconds
Radial — Max Deviation ±0.0001 mm 0.0002 mm TIR
Axial — Max Deviation ±0.0001 mm 0.0002 mm TIR
Coning of Axis ±0.5 arc second
Table 2
Line structured light sensor parameters.
Parameter Value
Mounting conditions Specular reflection
Fig. 1. 3D point cloud measurement device. Reference distance 60 mm
Measurement range (Z-axis (height)) ±8 mm
Measurement range (X-axis (width))-near side 8 mm
Measurement range (X-axis (width))-Reference distance 15 mm
four structured light sensors to measure the tooth flank and a line Measurement range (X-axis (width))-far side 15 mm
structured light sensor can be placed above the gear to measure the Light source Type Blue semiconductor laser
Light source Wavelength 405 nm (visible beam)
inner circle, however, 3D measurement of gears using this instrument is Repeatability Z-axis (height) 0.4 μm
still in the experimental stages. Further to this, Heartig et al. proposed Repeatability X-axis (width) 5 μm
a 3D point cloud assessment method and principle for gears, but did Profile data interval 20 μm
not present a suitable measuring instrument [7]. To rapidly obtain the Sampling cycle (trigger interval) Top speed: 16 μs
3D shape data of a gear tooth flank and to evaluate gear machining
precision, a line structured light sensor and high-precision air-floating
rotary table were used to build a 3D measurement device, which
can be used to quickly and accurately obtain the 3D point data and
dimensional error of gear teeth.
2. Measurement system
161
X. Guo et al. Precision Engineering 61 (2020) 160–169
[ ]
Where D = 𝑥1 , 𝑥2 , 𝑥3 , … , 𝑥𝑁 is the distance vector from each point
to the center of the ellipse.
The short axis of the ellipse is the cylindrical radius 𝑅, then ac-
cording
√ to the ellipse formula, the short axis of the ellipse is B =
−2𝑓
√
2 ( )2 , which gives get
𝑎+𝑐+ 𝑏2 + 𝑎−𝑐
𝑓
8
⎡ 1 0 1 0 0 ⎤
𝐵2
⎢ 0 −1 0 0 0 0 ⎥⎥
⎢ 8
⎢ 1 0 1 0 0 ⎥
𝒂𝑇 𝐶𝒂 = 𝒂𝑇 ⎢ 𝐵2 ⎥ 𝒂 = 1 (3)
⎢ 0 0 0 0 0 0 ⎥
⎢ 0 0 0 0 0 0 ⎥
⎢ 8 8 64 ⎥
⎣ 82
0 𝐵2
0 0 𝐵4 ⎦
According to the above mathematical relations, the following equa-
tion set can be obtained,
{
𝑆𝑎 = 𝜆𝐶𝑎
, 𝑆 = 𝐷𝑇 𝐷 (4)
𝑎𝑇 𝐶𝑎 = 1
Where 𝜆 is the eigenvalue of matrix 𝑆 relative to matrix 𝐶. In the
matrix calculation results, the eigenvector corresponding to the non-
negative eigenvalue is the real solution(in real number field) of the
equation.
To solve this generalized eigenvalue problem, the parameters of the
general ellipse expression 𝒂𝟎 and the central coordinates of the elliptic
Fig. 3. Positional relationship between the line structured light sensor and cylindrical geometric parameters (𝑋𝑐 , 𝑌𝑐 ), the long axis 𝐴, the short axis 𝐵, and
standard component.
the inclination angle 𝜃 are obtained, and the geometrical parameter of
[ ]
ellipse is 𝒈𝟎 = 𝑋𝑐0 𝑌𝑐0 𝑎0 𝑏0 𝜃0 .
Taking 𝒈𝟎 as the initial value of geometric distance least square
of the output laser beam with the sensor light plane and the cylinder. method, accurate elliptic parameters are obtained after several itera-
The output of the line structured light sensor is the distance from the tions of jacobian matrix, the jacobian matrix is
sensor zero plane to the ellipse surface. The ellipse has the following
⎡ 𝐽𝑋 ′ ,𝑋 𝐽𝑋 ′ ,𝑌𝑐 𝐽𝑋 ′ ,𝑎 𝐽𝑋 ′ ,𝜃 ⎤ ⎛𝛥𝑋 ⎞ ⎛𝑋 ′′ ⎞
𝐽𝑋 ′ ,𝑏
characteristics: ⎢ 1 𝑐 1 1 1 ⎥ 1 𝑐 1
⎢ 𝐽𝑌1′ ,𝑋𝑐 𝐽 𝐽
𝑌1′ ,𝑌𝑐 𝐽
𝑌1′ ,𝑎 𝐽𝑌 ′ ,𝜃 ⎥ ⎜ 𝛥𝑌𝑐 ⎟ ⎜ 𝑌1′′ ⎟
𝑌1′ ,𝑏
⎢ ⋮
1 ⎜ ⎟ ⎜ ⎟
1. The plane in which the ellipse is located is the plane of the line
⎢ ⋮ ⋮ ⋮ ⋮ ⎥⎥ ⎜ 𝛥𝑎 ⎟ = ⎜ ⋮ ⎟ (5)
structured light sensor laser plane. ⎢𝐽𝑋𝑚′ ,𝑋𝑐 𝐽𝑋𝑚′ ,𝑌𝑐 𝐽𝑋𝑚′ ,𝑎 𝐽𝑋𝑚′ ,𝑏 𝐽𝑋𝑚′ ,𝜃 ⎥ ⎜ 𝛥𝑏 ⎟ ⎜𝑋𝑚 ⎟
⎢𝐽 ′ ⎜ ⎟ ⎜ ′′ ⎟
2. The center of the ellipse is on the axis line of the cylinder
⎣ 𝑌𝑚 ,𝑋𝑐 𝐽𝑌𝑚′ ,𝑌𝑐 𝐽𝑌𝑚′ ,𝑎 𝐽𝑌𝑚′ ,𝑏 𝐽𝑌𝑚′ ,𝜃 ⎥⎦ ⎝ 𝛥𝜃 ⎠ ⎝ 𝑌𝑚 ⎠
component. [ ]
3. The elliptical short axis is equal to the radius of the cylinder then get the final ellipse parameters 𝒈𝐟 = 𝑋𝑐 𝑌𝑐 𝑎 𝑏 𝜃 , 𝒈𝐟 is
component. the most accurate fitting result.
The ellipse-fitting algorithm used in the coordinate registration of
First, the relationship between the ellipse short-axis and ellipse pa- the measuring device differs to ordinary ellipse fitting, in fact, geo-
rameters (𝑎, 𝑏, 𝑐) are used as dual-constraints for the algebraic distance metric fitting with a small sample range of measurements. However,
least squares method is used to solve for the ellipse parameters [8]. the spatial distribution of the measured points are concentrated within
The result is then used as the initial value and the geometric distance a small region, which can easily distort the fitted result. Thus, ellipse
least squares method is used to iteratively solve for the exact elliptic fitting based on incomplete arc segments presents a difficult problem
geometry parameters [9]. in the field of precision measurement, engineering mapping, and model
According to the positional relationship between the ellipse and line reconstruction.
structured light sensor, the measurement data can be converted into The fitting effect has an important relationship with the measure-
coordinates in the elliptical plane 𝑂′ 𝑋 ′ 𝑌 ′ . The inclination ratio of the ment data error, spatial distribution, and constraints. Results of the
𝑂′ 𝑋 ′ 𝑌 ′ plane with respect to the rotary table end face 𝑂𝑋𝑌 is the ratio general least squares method are often unsatisfactory or may even
of the major and minor axes of the ellipse. A point in the elliptical seriously deviate from theoretical values. In this paper, the fitting data
coordinate system 𝑂′ 𝑋 ′ 𝑌 ′ can be converted into cylindrical coordinates is derived from the high-precision line structured light sensor but the
using the inclination angle; therefore, it is possible to convert the measurement range is small. Therefore, the constrained least squares
line structured light sensor measurement data into three-dimensional method can improve the accuracy of the ellipse fitting to meet the
coordinate data along the basic rotary axis. measurement accuracy requirements.
According to the definition of an ellipse, an ellipse expression is
established, 3. Tooth flank error parameters
162
X. Guo et al. Precision Engineering 61 (2020) 160–169
And
{
𝑥 = 𝑎𝑧 + 𝑏
(10)
𝑦 = 𝑐𝑧 + 𝑑
where
Fig. 4. Model of the involute helicoid. 𝐴 𝐴 𝐵 𝐵
a= , 𝑏 = 𝑥0 − 𝑧0 , 𝑐 = , 𝑑 = 𝑦0 − 𝑧0 (11)
𝐶 𝐶 𝐶 𝐶
The tooth reference line parameters of each tooth flank are obtained
𝜃 angle can determine the shape of the tooth on the end section of an using the least squares fitting algorithm, and then the actual tooth
axial position, and 𝑧 can determine the axial position corresponding to reference line is obtained. As shown in Fig. 5, (for spur gear), the red
the tooth profile of the end section. straight line is the tooth reference line of the corresponding tooth [11].
In Fig. 4, M is a point on the involute flank, the position of this A point in the middle of the tooth width 𝑧𝑚𝑖𝑑 = 12 𝑏 is taken as the
specified point is: pitch deviation calculation point on the actual fitted tooth reference
{ }
⃖⃖⃖⃖⃖⃗ + 𝐸𝐹
𝑟⃗ = 𝑂𝐸 ⃖⃖⃖⃖⃖⃗ + 𝐹
⃖⃖⃖⃖⃖⃖⃖
𝑀⃗ line, P = 𝑝1 , 𝑝2 , … , 𝑝𝑁 , where 𝑁 is the number of teeth. Individual
(6) single pitch deviations can be calculated by the length of the arc
⃗
= 𝑥⃗𝑖 + 𝑦𝑗⃗ + 𝑧𝑘 difference between the theoretical pitch of two adjacent points. The
where: maximum value among deviations is 𝑓𝑝𝑡 .
𝑥 = 𝑂𝐸 cos(𝜏 + 𝜇 + 𝜃) + 𝐹 𝑀 sin(𝜏 + 𝜇 + 𝜃) Selecting one deviation as the starting point and successively cal-
culating the difference between each point and the starting point, the
= 𝑟𝑏 cos(𝜏 + 𝜇 + 𝜃) + 𝜃𝑟𝑏 sin(𝜏 + 𝜇 + 𝜃)
arc length and theoretical gear pitch can then be calculated to obtain
𝑦 = 𝑂𝐸 sin(𝜏 + 𝜇 + 𝜃) − 𝐹 𝑀 cos(𝜏 + 𝜇 + 𝜃) (7) the cumulative deviation of the gear pitch. The difference between its
= 𝑟𝑏 sin(𝜏 + 𝜇 + 𝜃) − 𝜃𝑟𝑏 cos(𝜏 + 𝜇 + 𝜃) maximum and minimum values is the 𝐹𝑝 .
𝑧 = 𝐸𝐹 = 𝑧
3.2. Profile deviation
The 𝑟𝑏 is the radius of base cylinder(mm), 𝜏 is the position angle
about the cylindrical of this specified involute flank(rad). All of the
The deviation of tooth profile could be caused by manufacturing
gear errors calculation and result analysis are based this geometric
error, installation error of cutter or short cycle error of machine tool
modeling.
transmission chain. The profile deviation is the amount by which a
measured profile deviates from the design profile [10]. This deviation is
3.1. Pitch deviation
commonly determined from one measured profile line along the tooth,
as shown in Fig. 6. However, in the presence of 3D points along a
Pitch deviation is one of the most important indicators of gear
broader surface area, we can expand evaluation of profile deviation
performance. Smaller pitch deviation is beneficial as it can reduce
to consider multiple profile lines along the tooth axial direction. The
vibrations and noise and improve transmission accuracy. The evalua-
additional dimension enables us to express profile deviations in terms
tion index mainly consists of the single pitch deviation (𝑓𝑝𝑡 ) and total
of a box as shown in Fig. 7. The box contains all profile deviation lines
cumulative pitch deviation 𝐹𝑝 [10].
and points among profile evaluation scope, therefore, the total profile
The pitch error of the cylinder gear can be calculated using the gear
tolerance would be calculated by the maximum and minimum value of
tooth reference line (the line that intersects the tooth flank and face
all profile deviation values.
reference cylinder). First, points near the tooth reference line on each
tooth flank are extracted by letting 𝑟 represent the theoretical reference
circle radius and 𝑟′ denote the distance from the measurement point to 3.3. Tooth flank form error
the gear center axis, and taking 𝑟′ − 𝑟 ≤ 𝜖 points as the tooth reference
line points, as is shown in Fig. 5. Selecting a tooth flank of the measured gear as the evaluation
The theoretical tooth reference line can be considered as a straight object, the theoretical involute tooth flank equation corresponding to
line perpendicular to the plane of the gear end face, and the actual the selected tooth can be obtained as shown in Fig. 8. From the involute
163
X. Guo et al. Precision Engineering 61 (2020) 160–169
164
X. Guo et al. Precision Engineering 61 (2020) 160–169
Fig. 11. User interface of gear 3D point cloud measurement and evaluation software.
Table 3
Gear parameters.
Parameter Value
Number of teeth 28
Modulus 3
Pressure angle 27.5◦
Tip height 2.527 mm
Root height 3.75 mm
Accuracy level 6
relationship between the current line structured light sensor and the
rotary table. Then the registration system automatically generated a
coordinate transformation matrix.
The experimental results show that the slant angle of laser plane
of the line structured light sensor to the end plane of measured gear
is within the range of 10◦ ∼ 85◦ , and the above ellipse fitting and
registration methods are highly accurate, and can be used to accurately
acquire 3D data on the flank of the object being tested.
Fig. 12. Original 3D raw point measurement data for gear.
165
X. Guo et al. Precision Engineering 61 (2020) 160–169
When the sensor laser beam leaned to the right and the rotary
table rotated clockwise, the spatial relationship between the laser beam
of the sensor and the measured gear tooth flank is divided into 5
stages [13]:
The spatial relationship and sensor data are shown in Fig. 14. The
top of the figure shows the relationship between the laser beam and
tooth flank, while the bottom side depicts the line structured light
sensor receiving area distance value and the sensor laser beam point
sensitivity.
From Fig. 14, it can be observed that the line structured light sensor
receives the light reflected off the left tooth flank of the gear more
fully and evenly compared to the right side, and all laser line range
could be effectively measured; Whereas, the curvature of the right tooth
flank relative to the laser beam causes the reflection angle of the laser
beam to be larger, and the sensor is unable to accurately measure the
reflected light. However, when the laser beam leaned to the left, then
the right flank measurements were effectively and the left flank could
not be accurately measured.
166
X. Guo et al. Precision Engineering 61 (2020) 160–169
𝜎 𝜎
[𝜇 − 𝑢 𝛼 √ ,𝜇 + 𝑢𝛼 √ ](𝛼 = 0.05) for the gross error in measure process.
2 𝑛 2 𝑛
Fig. 15. P26 gear measurement instrument. In Fig. 19, the measured flank surface points are shown as a scatter
plot. The two planes envelop all valid surface points and are located
at 𝑧𝑙𝑜𝑤𝑒𝑟 = −7.86 μm and 𝑧𝑢𝑝𝑝𝑒𝑟 = 8.01 μm. the 𝐹𝛼 (Profile deviation,
total) is 15.87 μm, i.e 15.8 μm. The 𝐹𝛼 result of Klingelnberg P26
is 14.9 μm,multiple comparisons verified that the gear measurement
system and error calculation method can accurately measure profile
error [15].
Fig. 16. Tooth flank evaluation area (green) after point cloud preprocessing.
After calculating the involute tooth flank error point-by-point, de-
viations across the entire tooth flank evaluation area were obtained.
cumulative pitch deviation is 19.1 μm. the differences between the two As shown in Fig. 20, error of the display gear near the tooth reference
instruments in the measurement of pitch deviation are from 2 to 3 μm. line is small and the processing quality is good at both the root and
tip of the tooth. Deviation in the top direction gradually increases, and
5.3. Calculation of profile error the error distribution reflects poor processing quality at the top of the
tooth root of the gear error form [16].
Following the 3D point cloud profile deviation box model discussed Fig. 21 shows the statistical distribution of points on the tooth
in Section 3.2, 50 profile deviation lines were extracted along the tooth flank evaluation area and deviations from the theoretical tooth flank.
direction, comparing more than 30,000 surface points. The distribution Based on the analysis of the distribution map, the 3D error distribution
of profile deviations is shown in Fig. 18, with 95% of the deviations of the tooth flank is consistent with the distribution calculated using
contained within the shaded region. 95% of the measured deviations the method proposed in Section 4.2. The distribution fully reflects
are contained within the lower and upper boundaries as denoted by the shape of the pitch error and is consistent with the pitch error
the solid red lines in Fig. 18. These boundaries are given by the values distribution obtained in Section 5.2.
Fig. 17. Pitch error result comparison (a) Pitch error result of 3D measurement system; (b) pitch error result of Klingelnberg P26 measurement.
167
X. Guo et al. Precision Engineering 61 (2020) 160–169
Acknowledgments
References
[1] Wang X, Shi Z. Evaluation system of gear accuracy based on full information.
Sci Sin Technol 2017;47(1):46–59. http://dx.doi.org/10.1360/N092016-00072.
[2] Goch G, Ni K, Peng Y, Guenther A. Future gear metrology based
on areal measurements and improved holistic evaluations. CIRP Annals
2017;66(1):469–74. http://dx.doi.org/10.1016/j.cirp.2017.04.046, URL http://
Fig. 20. Point-by-point 3D representation of tooth flank error.
www.sciencedirect.com/science/article/pii/S000785061730046X.
[3] Peters J, Goch G, Günther A. Helical gear measurement using structured light.
In: Proceedings of the XVI IMEKO world congress. Wien; 2000. p. 227–30.
[4] Leopold J, Günther H. Fast 3D measurement of gear wheels. In: Seventh
international symposium on laser metrology applied to science, industry, and
everyday life, vol. 4900. International Society for Optics and Photonics; 2002,
p. 185–94.
[5] Chen Y, Chen J. Optical inspection system for gear tooth surfaces using a
projection Moiré method. Sensors 2019;19(6):1450. http://dx.doi.org/10.3390/
s19061450.
[6] Kumar A, Jain P, Pathak P. Curve reconstruction of digitized surface us-
ing k-means algorithm. Procedia Eng 2014;69:544–9. http://dx.doi.org/10.
1016/j.proeng.2014.03.024, URL http://www.sciencedirect.com/science/article/
pii/S1877705814002707.
[7] Härtig F, Stein M. 3D involute gear evaluation – Part I: Workpiece
coordinates. Measurement 2019;134:569–73. http://dx.doi.org/10.1016/j.
measurement.2018.10.088, URL http://www.sciencedirect.com/science/article/
pii/S0263224118310315.
[8] Fitzgibbon A, Pilu M, Fisher RB. Direct least square fitting of ellipses. IEEE
Trans Pattern Anal Mach Intell 1999;21(5):476–80. http://dx.doi.org/10.1109/
34.765658.
[9] Liu Y, Dong X. Constrained least squares fitting of ellipse. In: 2010 international
conference on computational intelligence and software engineering; 2010. p. 1–3.
http://dx.doi.org/10.1109/CISE.2010.5676736.
[10] ISO. Cylindrical gears—ISO 1328:2013 system of flank tolerance classification—
Fig. 21. 3D error distribution on gear tooth flank. Part 1: Definitions and allowable of value of deviations relevant to flanks of gear
teeth. Central Secretariat Geneva.
168
X. Guo et al. Precision Engineering 61 (2020) 160–169
[11] Wang Q, Peng Y, Wiemann A-K, Balzer F, Stein M, Steffens N, et al. [14] VDI/VDE. VDI/VDE 2612 BLATT 1:2018 Measurement and testing of gears -
Improved gear metrology based on the calibration and compensation of ro- Evaluation of profile and helix measurements on cylindrical gears with involute
tary table error motions. CIRP Annals 2019;68(1):511–4. http://dx.doi.org/10. profile. I have add this information into my manuscript and as a reference. Verlag
1016/j.cirp.2019.04.078, URL http://www.sciencedirect.com/science/article/pii/ des Vereins Deutscher Ingenieure.
S0007850619301076. [15] Chernoff H, Savage IR. Asymptotic normality and efficiency of certain nonpara-
[12] Ni K. Areal gear metrology with modified flanks [Ph.D. thesis], The University metric test statistics. Ann Math Stat 1958;29(4):972–94, URL http://www.jstor.
of North Carolina at Charlotte; 2017. org/stable/2236941.
[13] Auerswald MM, von Freyberg A, Fischer A. Laser line triangulation for fast 3D [16] Brandão JA, Seabra JH, Castro MJ. Surface fitting of an involute spur gear tooth
measurements on large gears. Int J Adv Manuf Technol 2019;100(9):2423–33. flank roughness measurement to its nominal shape. Measurement 2016;91:479–
http://dx.doi.org/10.1007/s00170-018-2636-z. 87. http://dx.doi.org/10.1016/j.measurement.2016.05.076, URL http://www.
sciencedirect.com/science/article/pii/S0263224116302433.
169