Forsch Ingenieurwes (2019) 83:683–690

https://doi.org/10.1007/s10010-019-00346-5

ORIGINALARBEITEN/ORIGINALS

Temperature influence on involute gear measurements

Ann-Kathrin Wiemann1 · Martin Stein1 · Karin Kniel1

Received: 29 March 2019 / Accepted: 18 June 2019

© Springer-Verlag GmbH Deutschland, ein Teil von Springer Nature 2019

Abstract
Involute gears are high-stress components used in a variety of drivetrain applications. Small manufacturing tolerances
require low measurement uncertainties in production metrology. A well-established traceability chain from industrial
facilities to national gear measurement standards is essential for reliable quality control.
Especially for large parts, the ambient and workpiece temperatures contribute to the measurement uncertainty considerably.
However, measurements on the shop floor can rarely be performed at exactly 20 °C, which is the international reference
temperature for dimensional metrology.
In this paper, the influence of thermal expansion on gear deviations is discussed. Theoretical considerations are compared
to results obtained via finite element analysis (FEA) as well as to experimental data. This also includes measurements
of PTB’s large gear measurement standards under different temperature conditions. The outcome of this study can be
used either to estimate measurement uncertainty contributors in industrial applications or to compensate thermally induced
errors in gear measurements.

Temperatureinfluss bei Messungen von zylindrischen Evolventenverzahnungen

Zusammenfassung
Verzahnungen mit evolventischem Profil sind stark beanspruchte Bauteile in einer Vielzahl von Getriebeanwendungen.
Um diesen Beanspruchungen standzuhalten, sind sehr geringe Fertigungstoleranzen einzuhalten, die wiederum bei der
Qualitätsüberprüfung noch viel kleinere Messunsicherheiten erfordern. Für eine zuverlässige Qualitätskontrolle ist eine
geschlossene Rückführungskette von nationalen Normalen bis zum Produkt in der Industrie notwendig. Vor allem bei großen
Bauteilen liefert die Umgebungs- und Materialtemperatur des Bauteils einen entscheidenden Beitrag zur Messunsicherheit.
Allerdings können Messungen unter fertigungsnahen Bedingungen meist nicht bei der definierten Bezugstemperatur von
20 °C durchgeführt werden, bei der die Kalibrierung des Normals stattfand.
In diesem Artikel wird der Einfluss der thermischen Ausdehnung auf Abweichungen der Verzahnungsgeometrie diskutiert.
Theoretische Ansätze werden mit den Ergebnissen von Simulationen mit der Finite-Elemente-Methode sowie tatsächli-
chen Messwerten verglichen. Hierzu wurde das Großverzahnungsnormal der PTB bei verschiedenen Temperaturzuständen
gemessen. Die Ergebnisse können entweder dazu verwendet werden, den Einfluss auf die Messunsicherheit richtig abzu-
schätzen oder die Abweichungen, die durch den thermischen Zustand entstehen, zu korrigieren.

1 Introduction meter must always be accompanied by a statement on the

measurement uncertainty, as it accounts for the reliability
For a reliable assessment of manufactured gears by means of the measurement. This task-specific measurement uncer-
of dimensional metrology, it is of utmost importance to en- tainty can be regarded as an indication of the quality of
sure traceability. Measurements traced back to the SI unit the entire measurement process. Small measurement uncer-
tainties can be reached only if all influences that affect the
calibration process are well understood, thus allowing them
 Ann-Kathrin Wiemann
[email protected]
to be either controlled or compensated for.
In dimensional metrology, one of the most important
1
Physikalisch-Technische Bundesanstalt, Braunschweig, contributors to the measurement result is the ambient and
Germany workpiece temperature, which is especially important when

K
684 Forsch Ingenieurwes (2019) 83:683–690

measuring large parts. For the ambient temperature of cali- out this paper, let ˛ in K–1 denote the coefficient of thermal
bration laboratories during calibration measurements, 20 °C expansion, T in K the actual workpiece temperature and
is defined in standards as the reference temperature [1]. T = T − 20°C. All geometrical quantities affected by
However, the temperature of in-plant measurement stations thermal expansion will be labeled with an apostrophe, e.g.
used in the manufacturing industry often differs markedly rb0 .
from this reference.
In general, quantifying the influence of temperature
changes on the geometry of the workpiece to be measured 2 Computational studies with finite element
is a complex task. For elementary geometries such as analysis (FEA)
spheres, cylinders and cones, closed formulas can easily be
derived. For complex geometries, the temperature response In this section, the thermal expansion of large involute he-
is most often studied by means of finite element analysis lical gears is investigated with the help of computer aided
(FEA). This method is preferred when the temperature simulation which is the typical approach. A CAD model
distribution inside the workpiece is not homogeneous. of PTB’s large gear ring measurement standard (see Fig. 1
In Sect. 2 of this paper, the influence of thermal ex- left) serves as the input geometry [2]. The measurement
pansion is studied by means of FEA simulations using the standard is defined by the gear parameters given in Table 1.
example of PTB’s large involute gear measurement stan- The CAD model has the same characteristics as the in-
dard, which is 2 m in diameter [2]. In Sect. 3, we derive volute gear measurement standard, including a six-point
closed formulas that express the relation between the ho- support. Therefore, all supporting points are examined with
mogeneous temperature change and the workpiece defor- a fixed z-plane on their lower support plate. The move-
mation of cylindrical involute gears based on a linear ex- ment of the measurement standard is possible in all other
pansion model. In addition, we discuss the influence of the directions regarding this fixed plane.
temperature change on the most important evaluation pa- In this example, the measurement standard is virtually
rameters for the quality assessment of gear manufacturing heated to a temperature of T = 25 °C. The coefficient of
processes defined in international standards [3–5]. Sect. 4 thermal expansion is considered for the specific material
provides results from tactile measurements under various with ˛ = 11.1  10−6 K −1 . One profile line at b=2 and one
temperature conditions; these results are used as a means helix line on the V-cylinder were investigated on each flank
of verification. In Sect. 5, the contribution of temperature of the external and internal gears. For FEA, the measure-
change to the uncertainty of involute gear measurements is ment lines are implemented as geometrical elements with
estimated before Sect. 6 summarizes the article. Through- a high node density.

Fig. 1 Large gear ring measurement standard 2 m in diameter showing three different external and internal gears (left) and the overall movement
of the inhomogeneous temperature distribution with FEA (right)

Table 1 Geometrical parameters of PTB’s large gear ring measurement standard

External gear Internal gear
Helix angle β in ° 0 10 r 20 l 0 10 l 20 r
Facewidth b in mm 420 420 420 424 424 424
Profile evaluation length LAE in mm 85.83 83.00 80.59 90.66 89.24 86.65
Diameter of base circle db in mm 1809.848 1800.146 1804.075 1606.874 1594.415 1598.730
Base helix angle βb in ° 0.000 9.391 18.747 0.000 9.391 18.747

K
Forsch Ingenieurwes (2019) 83:683–690 685

Table 2 Comparison of the gear

Values from ... ... calculation ... evaluation of ... evaluation of ... evaluation of
deviations for profile and helix
using Eqs. 6 generated points points from FEA points from FEA
slope evaluation at 25 °C from
and 8 based on Eq. 5 (homog. temp.) (inhomog. temp.)
the four approaches described
External gear
0
fH˛ in μm
0° 4.764 4.765 4.752 4.747
10° right-handed 4.607 4.607 4.603 4.590
20° left-handed 4.473 4.474 4.466 4.454
0
fHˇ in μm
0° 0.000 0.000 0.022 0.050
10° right-handed –3.855 –3.855 –3.850 –3.797
20° left-handed –7.911 –7.911 –7.958 –7.827
Internal gear
0
fH˛ in μm
0° 5.032 5.033 5.106 5.023
10° left-handed 4.953 4.953 5.005 4.926
20° right-handed 4.809 4.809 4.875 4.807
0
fHˇ in μm
0° 0.000 0.000 0.020 0.070
10° left-handed –3.892 –3.892 –3.908 –3.785
20° right-handed –7.986 –7.986 –8.062 –8.071

FEA was performed for two different cases. The first the involute function
case considers a homogeneous temperature distribution at
25 °C, while the second case deals with an inhomogeneous inv ˛ .r/ = tan˛.r/ − ˛.r/
  r  r  (2)
distribution (see Fig. 1 right) based on real temperature b b
= tan cos−1 − cos−1 ;
measurements. Twelve temperature sensors monitored the r r
inner temperature and four sensors measured the surface and the constants
temperature. The mean value was 24.97 °C, while the over- 8
all range was about 0.5 K. The deflected geometries of the < −1 W left-handed
lines were extracted pointwise in the workpiece coordi- hand = 0 W spur ; and
:
nate system and evaluated by two independent commercial +1 W right-handed
8 (3)
software products used on coordinate measuring machines <−1 W left flank
in gear metrology. The results are depicted in Table 2 in flank = ;
:
Sect. 3. +1 W right flank

a 3D parametrization as defined in [6] is given by

3 A linear model for thermal expansion of 0 1
involute gears x
@ y A = FE .hand; flank; rb ; cI 'b ; r; z/
In addition to the typical approach described above in which z
0 1 (4)
simulations are used, this section examines theoretical con- r  cos .'b + hand  c  z + flank  inv ˛ .r//
siderations based on geometrical relations. = @ r  sin .'b + hand  c  z + flank  inv ˛ .r// A
The shape of involute helical gear flanks is determined z
exclusively by the actual values of the base radius rb and
the base helix angle ˇb . With the helix coefficient A linear thermal expansion is modeled as an isotropic
scaling whose scaling factor is  = 1 + ˛  T and whose
ˇb scaling center is in the coordinate origin, i.e. the point lo-
c= ; (1)
rb cated on the gear axis and on the bottom face of the gear.
Each point .x; y; z/ on the gear flanks shifts to œ  .x; y; z/.

K
686 Forsch Ingenieurwes (2019) 83:683–690

The resulting geometry described by 3.2 Helix deviations

0 0 1 0 1
x x Following Eq. (6.12) in [7], the helix slope deviation fHˇ
@ y0 A =   @ y A [3–5] of a perfect gear with a facewidth b at a temperature
z0 z T is
=   FE .hand; flank; rb ; cI 'b ; r; z/  0 
 
0 1 0 1
  r  cos .'b + hand  c  z + flank  inv ˛ .r// fHˇ = b  rb  c − c = b  rb  c  −1

= @   r  sin .'b + hand  c  z + flank  inv ˛ .r// A   (8)
(5) 1
z = b  tanˇb  −1
0    1 
  r  cos 'b + hand  c  z + flank  inv ˛ r
B     C
=B@   r  sin ' b + hand  c
 z + flank  inv ˛ r C
A
A gear with imperfect geometry has an actual helix coef-
  ficient cact that is different from its nominal value cnom = c =
z tan .ˇb / =rb . When this gear is measured at a temperature
0
= FE .hand; flank; rb ; c=I 'b ; r; z/ T , the value of helix slope deviation fHˇ is related to the
0
helix coefficient c = cact =. The actual value of the helix
is again an involute helical gear whose base radius is rb0 = slope deviation fHˇ;act is calculated from the measurement
rb and whose base helix angle is ˇb 0 = ˇb , hence c 0 = c=. value as
Note that ˇb 0 is related to the base radius rb0 while ˇb  
is related to rb . With this result, the influence of thermal fHˇ;act = b  rb  .cact − c/ = b  rb  c 0 − c
0
!
expansion on the gear’s profile, helix, and pitch deviations fHˇ
can be derived. = b  rb  + c − c (9)
brb
0
3.1 Profile deviations =   fHˇ + b  tanˇb  . − 1/

The profile slope deviation fH˛ [3–5] is directly propor-

tional to the gear’s base radius deviation. More precisely, if 3.3 Pitch deviations
LAE denotes the reference length for profile slope deviation,
a perfect gear at temperature T shows Pitch deviations are calculated from the actual positions of
0
the gear teeth represented by the starting angle 'b of the
0 rb –rb involute at the base circle. As these angles are unaffected
fH˛ = LAE = LAE  .–1/ = LAE  ˛  ΔT (6)
rb by thermal expansion in this linear model (see Eq. 5), there
is no impact of the temperature on the pitch deviations.
A gear with imperfect geometry has an actual base radius The outcome from Sect. 2 and 3 is depicted in Table 2
rb;act that is different from its nominal value rb;nom = rb . using the example of T = 25 °C. All geometry parameters
When this gear is measured at temperature T , the value of used as input quantities are taken from Table 1.
0
the profile slope deviation fH˛ is related to the base radius The values in the second column of Table 2 are based
0
rb = rb;act . The actual value of the profile slope deviation on Eqs. 6 and 8. The third column shows the results from
fH˛;act is calculated from the measurement value as synthetic measurement lines based on Eq. 5. These data sets
  have been imported to and evaluated by two independent
rb;act − rb LAE rb0 commercial software products used on coordinate measur-
fH˛;act = LAE = − rb
rb rb  ing machines in gear metrology and produced identical re-
 0 
LAE rb  fH˛ rb sults. The fourth and fifth columns show the outcome of
= + − rb (7)
rb   LAE  the FEA simulations described in Sect. 2.
 
f0 1 The differences between the two approaches from Sect. 3
= H˛ + LAE  −1 (direct calculation with Eqs. 6 and 8 and points generated
 
via Eq. 5) are not significant at 0.001 μm. The parameters
evaluated for the points generated and the homogenous FEA
lead to maximum absolute differences of 0.073 µm for the
profile slope deviation and 0.076 µm for the helix slope de-
viation. Homogenous and inhomogeneous temperature dis-
tribution during FEA produces the largest absolute differ-
ences. The maximum discrepancy occurs at 0.083 µm for

K
Forsch Ingenieurwes (2019) 83:683–690 687

profile slope deviation and at 0.131 µm for the helix slope surement. Only the results for the internal gear teeth are
deviation. shown in Fig. 2, which depicts the differences from the
reference measurement on the ordinate.
A linear relation between the temperature change and
4 Verification measurements using a large the measured gear parameters can be observed. Further-
gear ring measurement standard more, the temperature influence on helix slope deviations
clearly depends on the nominal helix angle (as described by
Verification measurements were conducted with a preheated Eqs. 8 and 9). According to the “Guide to the Expression
large gear ring measurement standard. During the measure- of Uncertainty in Measurement” (GUM) [8], separate and
ments, twelve inner temperature and seven additional sur- known systematic errors must be eliminated by means of
face temperature sensors monitored the workpiece temper- corrections for the measurement uncertainty to be as small
ature. Furthermore, eight temperature sensors distributed as possible.
around the measurement volume recorded the ambient tem- The correction is performed using Eqs. 7 and 9 for mea-
perature continuously. Throughout the cooling process of surements at a temperature that differs from the reference
this large gear ring, the geometry deviations were measured measurement temperature (20.3 °C). The results are shown
with a large coordinate measuring machine. in Fig. 3. The residual error of the compensated profile
Three tooth gaps each for the external and internal gears slope deviation increases to an absolute maximum size of
were measured by means of helix and profile sampling. 0.3 µm, while the helix slope deviation reaches an absolute
Both measurements conducted in scanning mode were per- maximum size of 1.8 µm after compensation. The residual
formed on the left and right flanks. In total, this amounted error does not seem to depend on the helix angle. The lin-
to 24 measurement lines. One complete measurement run ear response that corresponds to the temperature change is
including the measurements for the workpiece coordinate mainly eliminated for profile slope deviation. However, for
system lasted about 90 min. The range of the temperature helix slope deviations, a slight temperature dependency is
change in the workpiece during the measurement starts at still observable. One possible explanation is the inhomo-
1 K and ends at 0.1 K because of an exponential tempera- geneous temperature distribution during the measurement,
ture cooling progression. The mean value of all workpiece while the correction process considers only a mean tem-
temperature sensors during the first measurement was about perature. Secondly, the model is based on a pure linear
27.4 °C within a range of 0.4 K. Mean temperature values expansion and may reach its limit with increasing temper-
during the measurement of all external and internal gear ature. Moreover, the measuring machine’s characteristic is
teeth were considered separately. After the cooling process, not considered in the correction formula, which affects the
a measurement close to the reference temperature (20.3 °C) measurands as well. Nevertheless, the correction process
was conducted to ensure stability in the calibration mea-

Fig. 2 Differences of slope

deviations against the temper-
ature according to a reference
measurement at 20.3 °C for in-
ternal gear teeth, including linear
regression lines for each gear

K
688 Forsch Ingenieurwes (2019) 83:683–690

Fig. 3 Differences in profile

(top) and helix (bottom) slope
deviations against temperature
corrected with Eqs. 7 and 9
for the reference temperature
(20.3 °C) for internal gear teeth

allows the differences in the helix slope deviations to be re- uw and the temperature influence as a systematic error ub
duced by a factor of 4 for the example temperature (25 °C). sum to
q
U.k = 2/ = 2  u2cal + u2p + u2b + u2w (10)
5 Influence of temperature deviations on
measurement uncertainties If the temperature during the substitution measurement is
unknown, the measurement uncertainty is much higher be-
The large gear ring measurement standard was calibrated at cause the systematic error ub is derived from the difference
the metrological reference temperature (20.0 °C). All cal- between the calibration and measurement values. Hence,
ibrated parameters including the measurement uncertainty without compensating the temperature influence, all values
refer to this reference temperature (see Table 3). in Table 2 have to be considered for ub . Here, the qual-
For measurements with the substitution method [9], tem- ity inspection is sufficient only for lower quality classes
peratures that deviate from the reference temperature influ- in this case. By contrast, the correction process described
ence the measurement result, making it necessary to con- above, in which small residual errors u0b are one contributor
sider them as a contributor to the measurement uncertainty considered in the measurement uncertainty estimation, pro-
(see Eq. 10). The uncertainty contributors of the calibration duces smaller measurement uncertainties of the complete
process ucal , the measurement process up , the workpiece inspection process.
The correction can be calculated if the actual tempera-
ture is known. Here, the systematic error is calculated as
described in Sect. 2 and 4, i.e. by interpreting the calibra-
tion values for the deviant temperature. However, the resid-

Table 3 Measurement uncertainties compared to tolerances of profile and helix slope deviations of the large gear ring measurement standard
β Uncertainty from calibration Systematic error for Residual error contribution First grade tolerances
T = 5 K
ucal for fH˛ in μm ub for fH˛ in μm u0b for fH˛ in μm T ol of fH˛ in μm
0°/10°/20° 2.4 5.1 0.6 3.3
ucal for fHˇ in μm ub for fHˇ in μm u0b for fHˇ in μm T ol of fHˇ in μm
0° 2.6 0.0 1.2 3.3
10° 2.8 3.9 1.4 3.3
20° 3.2 8.0 1.8 3.3

K
Forsch Ingenieurwes (2019) 83:683–690 689

ual systematic error needs to be considered as an influence 20 °C, the deviation will have a positive or negative sign,
on the measurement uncertainty. respectively.
The relation between the measurement uncertainty U of
0
the complete inspection process and the tolerance T ol is fH˛ > fH˛ if T − 20°C > 0 (12)
recommended in VDA 5 [10] and ISO 22514-7 [11] (see
also [12]) and described in Eq. 11. For the helix slope deviation fHˇ , the facewidth b and
the base helix angle ˇb are the quantities that influence the
2  U.k = 2/=T ol  0.3 (11) measurand due to the temperature difference (see Eqs. 8
and 9). The sign of the deviation is related to the tempera-
Table 3 shows the calibration uncertainty, systematic er- ture difference in the following way:
rors ub (based on the third column in Table 2), contribution
0
of residual errors u0b and tolerance values for involute gears fHˇ < fHˇ if T − 20°C > 0 (13)
with a geometry similar to that of the large gear ring mea-
surement standard. The tolerance values are based on the The pitch measurement is independent of temperature.
formulas given in ISO 1328-1 [3]. The conclusions derived from this paper provide the
As manufacturing tolerances are tight, each influence on basis for calculating measurement deviations on gears due
the measurement process must be controlled to the greatest to temperature conditions and for correcting and thus mini-
extent possible. For profile measurements, the contribution mizing this influence. This is particularly relevant for large
of the compensation of the temperature influence to the gears of higher quality, for which measurements under
uncertainty decreases by a factor of almost 10. harsh production conditions are often favorable due to the
weight and complicated handling of the gears.
To simplify the measurement uncertainty estimation in
6 Summary and outlook future, the results of this work on the characteristic response
of large involute internal and external gears under different
This paper has investigated the effect of temperature on the thermal conditions will be included in the Virtual Coordi-
measurement of large gear deviations. First, the tempera- nate Measuring Machine (VCMM) software program [13].
ture dependence of the gear deviations was determined by This program enables the automatic estimation of task-spe-
means of FEA. Subsequently, the relationships were de- cific measurement uncertainties based on a virtual mea-
scribed by means of closed formulas. Both methods were surement process; its scope will be expanded regarding the
compared and produced equivalent results for each param- specific requirements of large-scale dimensional metrology.
eter at different temperatures. In this context, an additional issue is the investigation of
In this way, it was shown that the temperature influence mass effects from large workpieces and their uncertainty
is of great importance for the quality inspection of large contribution [2]. The VCMM is an easy-to-use tool that
workpieces. Temperature must be considered as an influ- provides task-specific measurement uncertainties in coordi-
ence in the measurement uncertainty budget according to nate metrology, thus allowing traceable quality assurance
valid standards. If the extent of the temperature influence to be performed within industrial production processes.
is known, it must be corrected. For this purpose, the basis
Acknowledgements This research project was funded by the BMWi.
for such corrections for gear deviations has been given in The authors would like to thank Thomas Cailloux from the University
this article in a mathematical description. However, if the of Besançon and Philipp Wortmann from PTB for their support in CAD
influence of temperature is unknown, the total difference modeling and FEA simulations.
between the measured value and the calibration value must
be considered in the measurement uncertainty budget as References
a systematic error.
Using a large gear ring measurement standard as an ex- 1. ISO 1:2016 – Geometrical product specifications (GPS) – Standard
reference temperature for the specification of geometrical and di-
ample, the geometrical deformations at 25 °C were investi- mensional properties (2016)
gated. Manufacturing tolerances for high-quality gears re- 2. Wiemann A, Stein M, Kniel K (2019) Traceable metrology for large
quire measurement uncertainties to be as small as possible involute gears, Precision Engineering vol 55. Elsevier, Amsterdam,
and temperature conditions to be known, as this allows cor- pp 330–338
3. ISO 1328-1:2013-09 – Cylindrical gears – ISO system of flank tol-
rections to be made. erance classification – Part 1: Definitions and allowable values of
For the profile slope deviations fH˛ , the roll length of deviations relevant to flanks of gear teeth (2013)
the active tooth profile LAE determines the magnitude of 4. VDI/VDE 2612-1:2018 – Measurement and testing of gears – Eval-
the thermal expansion (see Eqs. 6 and 7). Depending on uation of profile and helix measurements on cylindrical gears with
involute profile (2018)
whether the actual temperature is larger or smaller than

K
690 Forsch Ingenieurwes (2019) 83:683–690

5. DIN 21772:2012 – Gears – Cylindrical involute gears and gear pairs 10. VDA (2011) Volume 5. Capability of Measurement Processes, 2nd
– Definition of deviations (2012) completely revised edition 2010. Verband der deutschen Automo-
6. Härtig F, Stein M (2019) 3D involute gear evaluation – Part I: bilindustrie e. V., Frankfurt, Germany
Workpiece coordinates, Measurement vol 134. Elsevier, Amster- 11. ISO 22514-7. Statistical methods in process management – Capa-
dam, pp 569–573 bility and performance – part 7: Capability of measurement pro-
7. Lotze W (2005) Zahnradmessung mit Koordinatenmessgeräten. cesses (2012)
Eigenverlag, Dresden 12. ISO 14253-1:2017 – Geometrical product specifications (GPS) –
8. ISO/IEC Guide 98-3:2008 – Uncertainty of measurement – Part 3: Inspection by measurement of workpieces and measuring equip-
Guide to the expression of uncertainty in measurement (GUM:1995) ment – Part 1: Decision rules for verifying conformity or noncon-
(2008) formity with specifications (2017)
9. ISO 15530-3:2011 – Geometrical product specifications (GPS) – 13. Heißelmann D, Franke M, Rost K, Wendt K, Kistner T, Schwehn C
Coordinate measuring machines (CMM): Technique for determin- (2018) Determination of measurement uncertainty by Monte Carlo
ing the uncertainty of measurement – Part 3: Use of calibrated simulation. Adv Math Comput Tools Metrol Test XI:192–202
workpieces or measurement standards (2011)