Transactions on Engineering Sciences vol 34, 2001 WIT Press, www.witpress.

com, ISSN 1743-3533

3D gear measurement by CMM

W. ~ o t z e &
' F. ~ a e r t i ~ *
1

Scientzjic Adviser, Car1 Zeiss, Dresden, Germany

Physikalisch-Technische Bundesanstalt, Braunschweig, Germany

Abstract
Gear checking by means of co-ordinate measuring machines is becoming more
universal and precise on the basis of a full surface gear model consisting of
substitute helical involute flanks. The substitute helical involute flank is
equivalent to the best fit evaluation model of any other features in co-ordinate
measuring techniques. This model allows us to pick up measuring points
anywhere on the flank and to calculate gear parameters beyond common gear
standards. Thus there is no need to scan profile, lead and pitch exactly at the
defmed lines. A further advantage of the universal substitute gear model is the
improved measuring and evaluation accuracy. It allows us: furthermore, to
ensure the traceability of involute gear measurement.

1 Background
Modem co-ordinate measuring machines (CMM) and software allow the quality
of gears to be checked at any place using models and algorithms similar to those
used for any other workpieces such as gear boxes. The measurement of gears is
not more difficult than that of common workpieces comprising various
geometric elements such as planes, cylinders or sculptured surfaces, and even
helical involute gear flanks.
The paper deals with the evaluation of gear measuring points by means of a
3D substitute gear model covering separate geometric elements of the gear
flanks. The advantage of this 3D evaluation model and the respective software is
that measuring points may be picked up anywhere on the gear flanks, without
any restrictions on defined cross-sections or measuring planes. Thus the new
mathematical model and the software allow gears to be measured as easily as
any other geometric elements by means of any NC- or manually operated CMM
equipped with a mechanical or optical probe.

Transactions on Engineering Sciences vol 34, 2001 WIT Press, www.witpress.com, ISSN 1743-3533

The basic principle of conventional gear measurement is mainly based on

two characteristic lines on the flank, i.e. the involute in a transverse plane and
the helix at the pitch cylinder. From the point of view of mathematical treatment
both lines can be interpreted as lines in two dimensions. The conventional gear
model may therefore be referred to as the 2D h i e rnodel of a gear. This is also
the reason why the conventional gear measuring devices have been designed in
such a way that the gear errors can be measured exactly at the lines
corresponding to profile and flank line, and in a transverse plane when other
errors are measured, e.g. errors of pitch, tooth thickness etc.. At the time when
CMM application in gear testing was still in its infancy, the principle was to
simulate the conventional gear measuring devices for measurement of profile
and flank line as well as pitch and tooth thickness. Large measuring errors
caused by gear misalignment and by probe stylus deformation in the case of
large helix angles were unavoidable. Manually controlled CMM could not be
used in this case either.
The conventional line model does not provide a sufficient basis for 3D gear
measurement because CMM do not allow gear flanks to be measured alongside
defined spatial lines on the gear flanks. The general principle of CMM is to pick
up a sample of measuring points of the geometric elements of a workpiece.
From these measuring points the mathematical expression of the geometrical
element, referred to as substitute element, must be calculated (parameterised) by
means of best fit algorithms. Consequently, any gear measurement by means of
CMM must be modelled in the same way as described in the following:

1.
2.

3.

4.

Measurement of the spatial surface point co-ordinates {X, y, z)

of the gear flanks by means of a spherical probe tip.
The probe head must be able to pick up measuring points
independent of the spatial orientation of the gear flank (spur
and helical gear or worm gear).
Evaluation of the measuring points by means of best fit
algorithm according to the principles of orthogonal distance
regression (ODR) for the gear flanks to be determined (helical
involute surfaces).
Calculation of the searched sizes (length and angular) as
defined in the drawing from the parameters of the substitute
elements.

The complete and precise performance of a gear measurement by CMM thus

requires a full 3D model and 3D measurement, independent of any defined
cross-sections, lines or points. This new model is therefore referred to as 3 0
surface model for gear testing by CMM. The paper gives a short overview of the
new mathematical model, the measuring principle and the application of the
INVOLUTE software package in spur and helical gear testing. Despite the new
evaluation kernel the result output is oriented to the gear errors as defined in the
existing standards [3, 4, 51.

Transactions on Engineering Sciences vol 34, 2001 WIT Press, www.witpress.com, ISSN 1743-3533

2 The 3D model of the helical gear

It is the state of the art in the design, manufacture and measurement of
workpieces to split them into individual simple geometric elements and to use
suitable mathematical representations in the various software packages (CAD,
CAM, CAQ). For measuring purposes, a distinction must be made between the
real geometric elements of a machined workpiece and the substitute geometric
elements represented by a set of geometrical equations for computer
representation. The real geometric elements of a workpiece in fact errors of
position, size and form, as well surface roughness. The substitute geometric
elements are more or less mathematical approximations to the real workpiece,
which are to make possible the calculations of the actual workpiece errors of
size, position and form. Such 3D models must even be used for gear flanks and
complete gears in order that the same scope of workpiece information as for any
other workpieces is obtained [6].
A common gear consists of two sets of geometric elements, such as:
The reference elements for the gear axis. e.g. the bore or shaft of a gear
(except the axis of splines). These elements determine the gear axis and
the co-ordinate system for gear evaluation.
The substitute gear flanks of spur and helical gears as a set of involute
flanks in space as shown in fig. 1.
f

base cylinder

gear flank

Reference
element for
gear axis

flank orig~n

measuring

point

Figure 1 : Substitute gear flank and measuring points

The substitute gear flank [l, 21 is determined by the flank origin as the point of
intersection of base helix and xy-plane. The mathematical parametric equation
for the helical involute flank contains only three coefficients as given below:

base radius rb of the flank origin;

polar angle cpb of the origin;
base helix angle P,,.

Transactions on Engineering Sciences vol 34, 2001 WIT Press, www.witpress.com, ISSN 1743-3533

336

Laser Iletr-olo~and Uachuw Performance

[,X ,
seca c o s [ C 6 - h Eva
F(a, 6) = ,v = r, seca sm[p,, + C

6
with

C = tan(pb)lrb
helix coefficient
Ev(a) = tan(u)-a involute function
ro
base radius
a, C
Gaussian surface parameters
(pressure angle, axis co-ordinate)
flank direction (h=*1 for left- and right-hand flank)
/7

Further coefficients may be used for describing flank modifications (e.g.

crowning). When this single flank model is used, the whole substitute gear with
z teeth consists of 2 z substitute left- and right-hand flanks (see fig. 2) defined
by a set of 6*2 coefficients ( z = number of teeth).
Which are the advantages of the full 3D model of the real gear flanks with
individual coefficients to be calculated from measuring points selected on a
sample or even on each flank:
use of complete computerized gear model of the whole gear is possible
for any further calculations;
calculation of any cross-section with transverse plane, pitch cylinder,
straight lines or circles for the determination of profiles and further
geometric elements as is usual in evaluations in co-ordinate measuring
technique;
calculation of any lengths and angles for determination of pitch, tooth
thickness, base length, radial and diametral size over balls, run-out,
etc.;
determination of any form deviation in defined cross-sections for
profile and lead errors.
One of the main advantages of the measuring point evaluation by means of the
3D surface model is that for profile and lead measurement the measuring points
must not be picked up exactly in the transverse plane for profile testing or on the
pitch cylinder for lead testing.

Transactions on Engineering Sciences vol 34, 2001 WIT Press, www.witpress.com, ISSN 1743-3533

gear axis

-,

gear flanks
/

Figure 2: Substitute model of the whole gear consisting of 2 z substitute gear

flanks
All misalignment errors of the gear are exactly compensated by means of the 3D
flank evaluation. There is also no need to clamp the probe in a predefined
measuring plane. A touch trigger probe or a 3D analogue probe with three
degrees of freedom allows measuring points on the gear flank to be picked up
without any error even in the case of very large helix angles (e.g. worm gears).
The only condition is that the gear axis must be measured at first and the
measuring points must be transferred into the workpiece coordinate system with
the gear axis serving as the z-axis.

Figure 3: Measuring point and flank point

The coefficients of each flank must be calculated from the measuring points by
best fit routines based on orthogonal distance regression (ODR) [6]. Thus the

Transactions on Engineering Sciences vol 34, 2001 WIT Press, www.witpress.com, ISSN 1743-3533

distance f of the measuring points P{x, y, z} from the best fit helical involute
flank must be calculated in direction normal to the corresponding flank points
F{a, 0 as shown in fig 3. The corresponding flank points are also referred to as
foot points. Therefore the surface parameters { a , of the foot points must be
determined as well. The ODR best fit evaluation in the normal direction, taking
the probe tip radius into consideration, is the only way of correct evaluation
although it seems in contradiction to the conventional way of determining
profile deviations in the transverse plane.

c}

3 Flank parameterisation by means of best fit calculation

To evaluate the gear measuring points (picked up using probe balls of different
diameter anywhere on the flank) and depending on the parameters searched,
different types of best fit routines with different constraints and objective
functions must be used together with probe radius compensation. Furthermore,
the distinction between gears and splines must be taken into account. As with
geometric elements the following objective functions may be used:
least squares method (Gaussian method), so far preferred in gear
measurement
(2)
Q,
f j 2 3 Min
minimum zone method (Tsebyshev method), widely used to determine
of the form deviation according to I S 0 1101 of any geometric
elements, available but not yet used in gear measurement
Q ? = Max (1 f,l)
Min
(3)
L1 approximation as a robust best fit method with low sensitivity to
outliers,
also
not
commonly
used
today
Q,=clfjla
Mi~z
(4)
full adjacent gear profile optirnisation in the case of splines based on
tooth thickness and axis optirnisation; also not yet used today
Q , = Max ( S , ) = Min ( 5 )

=C

The deviationJ of measuring points must be calculated as distance orthogonal

to the flank surface, and the best fit evaluation must be carried out as the socalled orthogonal distance regression (ODR). This distance can be calculated as
the distance between the measuring point itself and the respective foot point on
the flank. The calculation of the Gaussian parameters {a. 4)of the foot points F
therefore is always part of the best fit evaluation of measuring points P(x, 2: z).
It follows for the scalar and signed distance f of a measuring point P 5 X =
{X, c,, z l T picked up by a probe with tip radius I.T. and foot point parameters {a,
<} that
(6)
J = ,lT
(X - (
a . )rT

Transactions on Engineering Sciences vol 34, 2001 WIT Press, www.witpress.com, ISSN 1743-3533

and the normal vector n

In the case of single flank evaluation by means of any of the best fit methods
mentioned (except spline evaluation), the following constraints must be
distinguished:
full flank best fit with three degrees of freedom for the determination of
the coefficients rh, pb, and Pb;
constrained flank best fit with locked helix angle Ph (for the evaluation
of the profile error in relation to the nominal profile);
constrained flank best fit with locked base radius rb (for the evaluation
of the lead error in relation to the nominal profile);
constrained flank best fit with locked base radius rb and helix angle Pb
(for the evaluation of the pitch and tooth thickness error).
In the case of splines, a full best fit can be carried out simultaneously all teeth to
determine the spline axis and the effective tooth thickness (similar to a full gogauge). All the best fit routines mentioned are implemented as a software kernel
for gear evaluation by means of the powerful INVOLUTE software package.
The overview of the various best fit methods shows many possibilities and
ways of gear measurement evaluation, beyond the evaluation strategies referred
in the conventional gear standards. The new methods based on the full 3D
feature model of gears must be taker into account in future in order to avoid
contradictions and misunderstandings as regards the results of gear testing. On
the other hand, the new parameters and results provide more information for
machining and quality control in gear production.
It should be mentioned that the ODR best fit evaluation converts the {X,y, Z}
co-ordinates of the measuring points into involute co-ordinates { a , 4,h},
f;,
being the deviation in the normal direction of the flank. In order to determine
the common gear errors in the transverse plane, this value must be converted
using the following equation
l
(8)
cos a,
It is yet another advantage that the same evaluation procedure can be used
for worm gears as well. The best fit evaluation of measuring points works even
in the case of helix angles close to 90.
Besides the individual point deviationsf, a set of 6 z (six times the number of
teeth) coefficients will be available for further evaluation. These coefficients do
not directly describe the gear parameters and gear deviations as used for the gear

f,. =f,-

Transactions on Engineering Sciences vol 34, 2001 WIT Press, www.witpress.com, ISSN 1743-3533

tolerances stated in various standards. But these normal parameters can be

calculated by the substitute gear model.

Figure 4: Intersection of involute flanks with other geometric elements

Fig. 4 shows a few examples of the calculation of intersections of a flank with
other geometric features for the determination of profile and lead errors, pitch
errors, etc.. As an example the calculation of the radial single-ball size is
demonstrated by the geometric model shown in fig. 5. Two adjacent flanks may
be reconstructed from the actual coefficients. To determine the size over a single
ball, the ball is moved into the gap between two adjacent flanks. The position of
the measuring ball can be calculated by iterative calculation as the point of
intersection of a transverse plane and the two equidistant helical involute flanks.
l

Flanke 1

I
I
Achse

Meflkugel
Anlagepunkt

Figure 5: Calculation of the radial single-ball size

The radial size over the ball is calculated as the distance of the midpoint from
the gear axis and the measuring ball radius. From this size, the run-out as well as
the size over two balls can be determined for each gap. Further gear parameters
can be determined by similar geometric models and calculations. After
determination of the whole substitute gear model (including crowning etc., if
necessary) all known gear parameters and deviations can be calculated.
The above mentioned 3D surface model and software kernel, which is strictly
based on the rules of CMM measurement and modelling, offers yet another
advantage. The full mathematical 3D gear model allows the measuring
uncertainty of the individual gear parameters to be calculated or estimated as for
any other geometric elements such as cylinders etc.. Furthermore, if the CMM
has taken fully calibrated, the traceability of gear measurements by means of

Transactions on Engineering Sciences vol 34, 2001 WIT Press, www.witpress.com, ISSN 1743-3533

CMM and certified software can be guaranteed. The only gap 1s the certification
of the software by means of certified test data. A group of experts (preferably
from national metrology institutes) should work out a number of xyz datasets
for the testing and certification of gear evaluation software as it was done for
other geometric elements a few years ago [7].

4 INVOLUTE 97 software package

Several years ago, the INVOLUTE software package was developed for gear
measurements by means of manually or CNC-controlled co-ordinate measuring
machines. It is meanwhile used by a number of CMM manufacturers as standard
package for gear measurement. It has been redesigned now for the Windows
NTl2000 operating system, and it uses the INVOLUTE kernel for data
evaluation based on the full 3D surface model described above. It can also be
used as a two-dimensional (2D) package for the evaluation of optical gear
measurements by means of image processing systems. The measuring
parameters of the program are shown in table 1.
Table 1: Parameter limits of gears to be tested by the INVOLUTE program
max
l Symbol l min
Parameter
number of teeth
1z
13
1 500
normal module
I m,
10.1
pressure angle
Ia
1 10"
1 40"
i hehx angle
ph
1 0"
1 85"
number of teeth in segment
I z,
Il
Iz

The standard version of the program has been compiled for the English and
German language. It can be run in the stand-alone mode as well as linked to the
CMM software (via COM). Touch trigger probes as well as analogue 3D probes
or optical probes can be used. The user may define the individual evaluation
tasks such as profile testing, lead testing, pitch testing or flank testing, the
number of measuring points per flank and even the number of fully tested teeth.

a. profile

b. lead

c , flank

d. pitch

Figure 6: Measuring strategy for the testing of profile, lead, topography and
pitches
A single involute profile can be tested as well as a full gear with all profiles,
leads and pitches. The measuring points may be picked up point by point with a
touch trigger probe or by scanning with analog probes. In this case, low-pass

Transactions on Engineering Sciences vol 34, 2001 WIT Press, www.witpress.com, ISSN 1743-3533

342

Laser. \letr.olog~~
and .\taclzlt~ePerformance

filters such as the Gaussian filter can also be activated for profile and lead
measurement with different cut-off.
Evaluation of the measuring points requires that the nominal gear parameters
be available as a nominal data file. The parameters, tolerances, designations and
additional information can be defined by a menu for nominal parameters. A
universal system for K-charts has been implemented to make checking of profile
modification possible. Options are available for the testing of special
modifications such as crowning and relief.
A very flexible and language-independent system has been implemented for
result output. It allows special output forms to be designed in any language for
text and graphics output. The profile, flank and pitch error charts may be printed
in any orientation, size and number. The flank topography can also be drawn
when a regular or random grid of flank measuring points is measured (about 100
up to more than 1000 points per flank). The program is also equipped with
powerful error-detecting routines for the nominal data and the measurement
data.

5 Reducing measuring uncertainty

Taking an approach which treats the flank of gears as a surface, the German
national metrological institute, the Physikalisch-Technische Bundesanstalt is
developing new measuring methods to ensure the traceability of involute gears.
This is a must in order that the ever more exacting accuracy requirements for the
measurement of standards can be met. To date, gears have been traced back via
one standard each for profile, helix slope and pitch, a special standard
calibration facility being available for each of these standards. These facilities
allow the national primary standards to be traced back very accurately. The
substitution method is applied to pass the measure on within the traceability
chain, from industrial standards and master gears to the product [8, 91. The main
drawback of this method, apart from the time and effort to be spent, is the
accuracy loss at each stage of the traceability chain, due in particular to the fact
that the dimensions of standards and product may differ considerably. An
essential aim therefore is to clearly shorten the traceability chain. This aim can
be reached by making it possible to calibrate standards, which closely
approximate the work piece (product-like standards), already at the level of the
national institutes (Fig. 7).
In fuhire, the gears will be traced back using a coordinate measuring machine
whose axes are monitored interferometrically, and a highly precise rotary table.
As is common practice in the link-up of prismatic work pieces such as circle,
cone or sphere, the flanks of a gear will in future be treated as areas. This
approach offers two essential advantages:
The evaluation in the classical way by profile, helix slope and pitch
remains possible.

Transactions on Engineering Sciences vol 34, 2001 WIT Press, www.witpress.com, ISSN 1743-3533

The flank is treated as a surface. It is represented on the basis of its

characteristic parameters, such as base circle, lead and position.
Traceability to measurement results along the classical lines is an important
prerequisite which makes a comparison with the evaluation guidelines possible
which have been followed for decades. The description of the involute as a
surface bridges the gap to modem 3D measuring techniques which need not put
up with the technical limitations of the past.

today

future

prof~le

types:
- spurlhel~calgeas
- ~nternallexternalgears

helix

parameters:

pitch

- prof~le
- hehx

- rnodlflcat~ons
p~tch
...

Figure 7: Gear artefacts today and in future

6 Final remarks
The traditional quality control of gears has been increasingly turned into a field
of application of the co-ordinate measuring technique. Because of the very
conventional gear standards and testing recommendations the evaluation
software commonly used simulates only the conventional and more or less twodimensional testing procedures. When the more general 3D analytic modelling
of geometric elements is applied, gear evaluation is based on substitute helical
involute flanks and finally on a full 3D substitute gear model. This gives much
more flexibility for measurement and evaluation of gears by CMM. There is no
need anyrnore to measure profile, lead or pitch exactly at predefined flank lines
or flank points. The new INVOLUTE 97 software package for the Windows
NTl2000 operating system is a very powerful tool for the testing of helical and
spur gears by means of manually and CNC-controlled CMM based on the full
3D substitute gear model.

Transactions on Engineering Sciences vol 34, 2001 WIT Press, www.witpress.com, ISSN 1743-3533

Globig, A.: Anwendung der rechnergestiitzten Koordinatenrnessung zur

Paarungspriifung und zur Beurteilung von Stirnradverzahnungen.
Dissertation, TU Dresden, 1975
Lotze, W., Rauth, H.-H. und Ertl, F.: Neue Wege und Systeme fur die
wirtschaftliche 3D-Verzahnungspriifung. VDI-Berichte Nr. 1230, 1996,
S.1021-1030
DIN 3960: Begriffe und Bestimmungsgrol3en fur Stirnrader und
Stirnradpaare rnit Evolventenverzahnung. Beuth-Verlag GmbH, Koln
U. Berlin 1987
VDIIVDE-2612: VDIIVDE-Richtlinie: Priifung von Stirnradern mit
Evolventenprofil. Blatt
1: Profilpriifung, 1978; Blatt 2:
Flankenlinienpriifung, 1980. VDINDE-Gesellschaft fiir Me& und
Regelungstechnik.
VDIIVDE-2607:
VDIIVDE-Richtlinie:
Rechnerunterstiitzte
Auswertung von Verzahnungsmessungen an Zylinderradern rnit
Evolventenprofil, Entwurf 1995. VDIIVDE-Gesellschaft fur Me& und
Regelungstechnik.
ISOITR 10 360-1: Co-ordinate metrology - Part 1: Definitions and
applications of the fundamental geometrical principles. ISO, TC 3
BCR Information: Testing Co-ordinate Measuring Machine
Algorithms; Phase 11; EUR 13417 EN; 1991
Beyer W.: Traceable Calibration of Master Gears at PTB. AGMA
Technical Meeting, Cincinnati, Ohio, 1996
Haertig, F. Assessment of Uncertainty in Gear Measurement using the
"Virtual CMM Method"; Proceedings of the 2-nd. International
Conference on Machining and Measurements of Sculptured Surfaces;
Krakow; ISBN 83-912887-3-0; 2000