Monitoring of slowly progressing deterioration
of computer numerical control machine axes
E Uhlmann
1,2
, C Geisert
1
*,and E Hohwieler
2
1
Fraunhofer Institute for Production Systems and Design Technology (IPK), Berlin, Germany
2
Department of Manufacturing Technology, Technische Universita
¨t Berlin, Berlin, Germany
The manuscript was received on 1 November 2007 and was accepted after revision for publication on 18 March 2008.
DOI: 10.1243/09544054JEM1040
Abstract: The feed axes of computer numerical control (CNC) grinding machine tools are
among the most mechanically stressed components of machine tools owing to the high process
forces and rough manufacturing environment which they encounter. The resulting wear and
tear depends strongly on the product range and the manner of machine operation. To counter-
act a functional deficiency of these central machine units, the current usual approach is preven-
tive maintenance. The manual inspection of feed axes is complex and time consuming. A
complicating matter is that the deterioration normally progresses very slowly and depends on
the position of the stress along the axis. Existing approaches to automated estimation of the
‘health status’ of feed axes do not take this factor into account.
This paper presents a procedure that addresses this gap. During simple test routines, the
drive current, axis position, and feed rate are recorded. With the help of additional machine
data, characteristic values are computed directly at the computer of the human–machine inter-
face (HMI). The results are then transferred to and stored on a database server at the machine
manufacturer. This approach enables the service technician to trace the progression of the axes’
‘health status’ over a long time. This approach makes it possible to detect trends in the charac-
teristic values at an early point in time. This leads to a better planning of necessary mainte-
nance actions adapted to the remaining lifetime of the wearing component.
Keywords: condition monitoring, availability, maintenance, wear and tear, feed axes, life-cycle
1 INTRODUCTION
Applications of centreless external cylindrical grinding
are exceedingly appropriate to mass production. One
of the main components of such computer numerical
control (CNC) machine tools to experience wear and
tear is the feed axes. How fast and where along the
axis deterioration occurs depends on factors such as:
(a) the tool–workpiece combination;
(b) the grinding method;
(c) the condition of lubrication;
(d) the quality of maintenance procedures.
Grinding methods such as plunge-cut grinding and
throughfeed grinding (see Fig. 1) in particular involve
a very limited effective workspace.
This leads to stress on the feed axis, which is highly
location-specific. At these locations along the axis,
wearout can be expected to occur more rapidly and
to a greater extent. Another factor that influences
the progress of deterioration is the contamination of
the guidance by grinding debris. This happens if the
wiper does not function properly. The result is a
loss of accuracy or, in the worst case, a breakdown
of the feed axis [1]. As explained in reference [2],
non-working wipers are responsible for 90 per cent
of all failures in linear guidance.
Owing to the influence of the axis controller, the
deterioration of a feed axis only becomes obvious
when its function fails and an alarm message is
generated by the programmable logic controller
(PLC). In this article, a procedure for tracing changes
in the dynamic behaviour of feed axes is described.
Using statistical signal processing methods, charac-
teristic values are computed. These can be used
to determine the axis ‘health status’. A history of
*Corresponding author: Department of Manufacturing Tech-
nology, Pascalstrasse 8-9, Berlin, Germany. email: claudio.
JEM1040 ÓIMechE 2008 Proc. IMechE Vol. 222 Part B: J. Engineering Manufacture
SPECIAL ISSUE PAPER 1213
position-dependent characteristic values is generated
and stored in a relational database.
2 STATE OF THE ART
Much effort concerning condition monitoring sys-
tems focuses on the research and development of
IT-frameworks with the ability to integrate external
sensor signals respectively to use control-integrated
signals [3–8]. However, even though most of these sys-
tems provide various methods for signal processing
and data analysis, it is difficult to adapt them to the
system that needs to be monitored [9]. Given a parti-
cular recorded pattern, expert know-how and the
results of multiple case studies are necessary to ensure
that the deterioration status is interpreted correctly.
This section considers a selection of commercial
products and findings in current research on feed
axis monitoring.
2.1 ePS Network Services
With its ePS Network Services, the Siemens Corpora-
tion offers a tool for the acquisition and documenta-
tion of the current machine axis state [10]. This
feature is called eP-Performance and includes the
following tests:
(a) the circularity test;
(b) the synchronized axes test;
(c) the universal axis test.
During the tests, control internal data are sampled
and used to calculate certain parameters. The para-
meters calculated in the universal axis test are char-
acteristic quantities to define the friction, the
moment of inertia and a torque offset [11]. These
are obtained using complex mathematical models
[12] in which the dynamic system is stimulated by a
special axis motion profile.
Repetitions of single measurements are used to
generate trend curves from characteristics in mea-
surement series. It is suggested to use these trend
curves as a basis for the optimization of maintenance
activities (see Fig. 2).
There are two additional features implemented,
called NC monitor and PLC monitor. The NC monitor
can be used to estimate the life-cycle loads of machine
axes by counters, while the PLC monitor enables long
time records of arbitrary PLC variables [11].
2.2 LoeWe – life cycle oriented machine tool
The project LoeWe [13] was funded by the German
Federal Ministry of Education and Research (BMBF)
within the framework of the ‘Research for Tomorrow’s
Production’ programme (project duration 2004–2007).
One objective of this research project was to monitor
the deterioration status of components of particular
functional relevance. The ball screw drive was identi-
fied as mainly responsible for the axis positioning
accuracy. Increasing wear of the ball screw leads to a
decrease of the ball screw pre-load. Therefore a condi-
tion monitoring strategy for the detection of loss of this
pre-load was chosen [14]. The implemented methods
for measuring the wear of the ball screw are based on
investigations that are described in reference [15]. In
the first method the change of the natural frequency
of the axis in dependency to pre-load conditions is
analysed. Figure 3 shows that the estimated eigenfre-
quency falls with a decrease of the ball screw pre-load.
The second method uses the spring characteristics
of ball screws that differ for unworn and worn ball
Fig. 1 (a) Plunge-cut grinding and (b) throughfeed grind-
ing (copyright Studer Mikrosa GmbH)
Fig. 2 Trend generation and monitoring with ePS Network
Services [10] (copyright Siemens AG 2006)
Fig. 3 Correlation between pre-load and eigenfrequency
[14] (copyright IFW)
Proc. IMechE Vol. 222 Part B: J. Engineering Manufacture JEM1040 ÓIMechE 2008
1214 E Uhlmann, C Geisert, and E Hohwieler
screws. Further wear measurements for vertical feed
axes are planned using a specific velocity profile [14].
2.3 REACH – development of a method to improve
the reliability and availability of
machine tools
A modular system for the monitoring of machine tool
component state was developed by the European
Brite/EuRam project REACH (project duration
1998–2001) [16]. Current control internal signals of
open CNCs were used to describe the behaviour of
the components. If necessary, external sensors were
integrated into the system [17].
With the aid of long-running tests, it was shown
that the use of the drive current instead of the
displacement force signal is sufficient for detecting
mechanical disturbances. Validation of the realized
monitoring system was done at a test bench for two
typical types of feed axis failure:
(a) backlash;
(b) pitting on guideways.
Backlash detection was analysed by building the
difference between the directly measured position
and the position calculated from the motor encoder.
For the detection of pitting the drive current signal
was chosen [18].
In the context of this project, a doctoral thesis war-
rants mentioning [19]. A detailed enumeration of the
analytical methods used to detect various feed axis
disturbances is provided in this work.
2.4 Machine and process diagnostics
At the Institute for Control Engineering of Machine
Tools and Manufacturing Units (ISW) in Stuttgart, the
usability of control-integrated signals for the condition
monitoring of feed axes is currently undergoing re-
search. The investigations, which where carried out in
this context were funded by the Research Association
for Machine Tools and Manufacturing Technologies
(FWF) and the German Research Foundation (DFG).
With the help of adapted test signals the ‘Stribeck
curve’ can be estimated [20]. Furthermore, methods
for the condition monitoring of ball screws, commonly
used components within feed axes, were developed.
Algorithms for the estimation of characteristic values
were implemented at the ISW [21]for:
(a) change of lead screw pitch error;
(b) friction;
(c) stiffness;
(d) backlash error.
2.5 e-Industrial Services
One aspect of the Fraunhofer Institute’s research
project ‘e-Industrial Services’ (2000–2003) was the
development of electronic services for the analysis
and prediction of machine health status using enhan-
ced diagnostic algorithms [22].
In this approach, data are recorded during a speci-
fically designed test-NC-procedure under specifically
defined conditions. The signals recorded and used
are the drive current and the rotation speed of the
axis or the spindle drives. A linear mathematical
model of this electromechanical unit was used to
estimate the condition of a feed axis with controlled
rotation speed. From this system of differential
equations characteristic diagnostic features were
generated. These features represent the static and
sliding friction parameters and the moment of iner-
tia. The drive current, rotation speed and accelera-
tion derived from the rotation speed are used as
input for the least squares method. This approach is
used for parameter identification in the case of an
over-determined set of linear equations [23].
In addition to the diagnostic tests, the load profile
of the machine tool is continuously logged during
normal machine operation. From the load profile
experience-based statements about the expected
‘health state’ can be made. Figure 4 shows an exam-
ple of an internet based service that indicates the
point in time at which service activities are expected
to be necessary [24].
3 MONITORING CONCEPT
As mentioned in section 1, in the case considered, the
machine tool axes underlay a special load profile that
is specific to this kind of machining. The challenge is
to develop a monitoring system which exhibits on
the one hand sufficient sensitivity in order to detect
Fig. 4 Presentation of life cycle data (here: cumulated
rotation speed) as a Web Service (copyright
Fraunhofer IPK)
JEM1040 ÓIMechE 2008 Proc. IMechE Vol. 222 Part B: J. Engineering Manufacture
Monitoring of slowly progressing deterioration of CNC-machine axes 1215
progressive, position-dependent deterioration, and on
the other hand sufficient robustness to endure the
rough manufacturing environment. Last but not least,
the monitoring system must fulfil the requirement
that additional auxiliary process time be avoided.
3.1 Premises
Centreless grinding machines are mainly used for mass
production in the automotive industry. The machine
tools are characterized by a great number of feed
axes owing to the centreless grinding technology. A
typical axes plan is shown in Fig. 5. The grinding wheel
made of corundum is mounted on the X1-axis. The
weight of a new, unused wheel is about 400 kg (with-
out mounting device) and slowly decreases as a result
of abrasion.
The machine tools are equipped with an open CNC
system of type SINUMERIK 840D made by Siemens.
The open CNC architecture enables data acquisition
of control internal sensor signals and machine data
via OPC (OLE for Process Control).
3.2 Structure of the monitoring system
The proposed monitoring system is capable of the
following principal tasks:
(a) automated execution of axis tests;
(b) data logging;
(c) signal pre-processing (sensor signals only);
(d) data compression;
(e) data transmission;
(f) data storage;
(g) data analysis;
(h) report generation.
Figure 6 shows the system’s scheme. A decentra-
lized architecture was chosen with the multiplicity
of the tasks in mind. While data logging, signal pre-
processing, and data compression take place at the
machine tool, storage, analysis and visualization
take place at the location of the machine manufac-
turer. This separation of location has the advantage
that a large amount of data, even coming from sepa-
rate machine tools, can easily be used for compara-
tive analyses. Long-term trend analyses of archived
axis test results support the detection of hidden
trends in the dynamic behaviour of feed axes. Trend
analysis is an important tool owing to the controller’s
influence on the dynamics of feed axis. The axis
function will be correct as long as control variables
are regulated within their limitations. Trend monitor-
ing of characteristic values, which are generated from
implemented axis tests can be used to uncover small
trends, caused by for example deterioration of feed
axes, even before the effects become obvious.
The aspect of data transmission will not be
addressed in this paper. It belongs to the area of infor-
mation and communication technologies and secure
IT infrastructure build-up. A comprehensive discus-
sion of these subjects can be found in reference [25].
Fig. 5 Axes plan of a MIKROSA centreless grinding
machine tool of type KRONOS L (copyright Studer
Mikrosa GmbH)
Fig. 6 Global structure of the monitoring system
Proc. IMechE Vol. 222 Part B: J. Engineering Manufacture JEM1040 ÓIMechE 2008
1216 E Uhlmann, C Geisert, and E Hohwieler
3.3 Axis test and life cycle data
If wear and tear of a feed axis increases, the dynamic
behaviour of the feed drive system changes. For exam-
ple, as a result of modifications to the tribological sys-
tem, friction rises and sluggishness occurs. With the
increased friction the drive current also rises [26].
By moving the axis with constant feed rate across
its entire range (in both directions), it is possible to
detect changes in the dynamical behaviour. During
the test, the drive current, position and feed are
sampled. Figure 7 shows the sampled position and
drive current data of an axis test. The feed signal is
used automatically to detect the two intervals in
which the positive or negative axis motion is con-
stant. The phases of acceleration are masked because
they would interfere with the analysis.
Auxiliary data providing information on the life
cycle of the machine tool are logged continuously.
This includes the alarm-log history and information
on previous maintenance activities. Such data facili-
tate failure analysis in detailing the events leading
to a failure during operation [27].
3.4 Pre-processing
The following statistical moments are calculated
from the drive current signal and used to generate
key indicators for condition monitoring:
(a) mean: m¼1
NX
N
i¼1
Xi
(b) variance: VarðXÞ¼EððXmÞ2Þ
(c) skewness: vðXÞ¼EððXEðXÞÞ3Þ
VarðXÞ3=2
(d) excess kurtosis: g2¼EððXEðXÞÞ4Þ
VarðXÞ23
The meaning of significant changes in the mean
and variance is evident. The smoothness of the sys-
tem is characterized by the variance of the signal,
while the drive current is proportional to the driving
torque. Therefore, difficulties experienced by the feed
axis are reflected in respective trends in the mean.
An exemplary trend of the mean current of the X1-
axesofagrindingmachineoftypeKRONOSLispre-
sented in Fig. 8. The data represent the dynamic state
of the axes within a small area (position: 383–403 mm).
The general trend in the characteristic value is easy
to see. Less power is needed to traverse the feed axis.
The machine is still working, and no maintenance
activities have been necessary up until now. A possi-
ble reason for this phenomenon is the shake down
effect of new feed axes. After a certain period of usage
the axes motion becomes smoother. An explication
for peak in the centre of the trend cannot be given,
owing to the fact that this machine is not located at
the test field of MIKROSA.
The skewness indicates on which side of the mean
the weight of the data falls. It can be used to detect
onsetting failures that depend on the direction of
traversing. The excess kurtosis is an indicator for
variations of the distribution referring to normal dis-
tribution. Distributed local damages along the axes
lead to a distribution with many peaks, for example
pitting on guideways. In this case, the excess kurtosis
is greater than zero [28].
As mentioned before, the deterioration does not
proceed homogeneously at all points along the axis.
For this reason, the analysis of the signal is divided
into small ‘windows’. Every window pertains to a
certain position along the axis.
To avoid additional auxiliary process time, the axis
test is combined with a periodically recurring traver-
sing of the axes in which the guidance is provided
with lubricant.
3.5 Data management
In order to detect trends in the recorded characteris-
tic values, an ‘initial state’ must first be defined. The
characteristic values in the initial state are in accor-
dance with a ‘good’ feed axis state of health and pro-
vide the basis for trend detection.
XML (eXtensible Markup Language) was chosen
as the data interface between the different compo-
nents of the monitoring system. XML is a universal
data exchange format that can be used for machine-
Fig. 7 (a) Axis position and (b) drive current
Fig. 8 Trend of feed drive current signal (copyright
Fraunhofer IPK)
JEM1040 ÓIMechE 2008 Proc. IMechE Vol. 222 Part B: J. Engineering Manufacture
Monitoring of slowly progressing deterioration of CNC-machine axes 1217
to-machine communication. In the work presented
the XML-logfiles are parsed by an XSLT-processor.
The processor generates files in SQL-format (Struc-
tured Query Language) as it imports the data into a
relational database (see Fig. 9).
The scheme of the database shows the relations
between the data, the components, and the machine
tool. Visualization of the data is provided by prede-
fined reports. It is possible to present visually the
stored information for various time intervals in
various ways.
4 CONCLUSION AND OUTLOOK
Using the proposed method of monitoring the condi-
tion of feed axes, it is possible to detect trends and
slowly progressing changes in the dynamic system.
Information on the machine tool’s load history that
is additionally logged during the machine tool’s life
cycle can be used to support the search for causes
of failure. The database-based approach facilitates
comparative analyses and long-term investigations.
It is still an open task to provide evidence, which
kind of axis deterioration and failures are detectable
by observing the proposed characteristic values. At
the moment, this is done by collecting experience in
the field from operating machines.
Ongoing work at the Fraunhofer IPK deals with the
development of model based machine diagnosis. Sev-
eral mathematical models were implemented. The
evaluation of the models is done using different
motion profiles for the axis movement to stimulate
the system’s dynamic behaviour. The aim of this
investigation is to generate further robust and sensitive
characteristic values for the monitoring of feed axes.
With the implementation of intelligent machine
tool components that are able to analyse and store
their wear status the field of condition monitoring
applications will increase.
ACKNOWLEDGEMENTS
Parts of the presented work have been carried out
within the research project DYNAPRO funded by the
‘Stiftung Deutsche Industrieforschung’. It was done
in cooperation with Studer Mikrosa GmbH, a manu-
facturer of centreless external cylindrical grinding
machine tools.
The Fraunhofer Institute for Production Systems
and Design Technology is a partner of the EU-funded
FP6 Innovative Production Machines and Systems
(I*PROMS) Network of Excellence.
REFERENCES
1Available online at www.ima.unistuttgart.de/dich
tungstechnik/abgeschlossene_projekte/wzm/wzm.en.
html. Institute of Machine Components, University of
Stuttgart, 2002.
2 Jansen, M. Abstreifer fu
¨r Werkzeugmaschinenfu
¨hrungen.
Dissertation, Universita
¨t Stuttgart, URN: urn:nbn:de:bsz:
93-opus-24403, available online at URL: http://elib.uni
stuttgart.de/opus/volltexte/2005/2440/, 2005.
3 Lee, J., Abujamra, R., Jardine, A. K. S., Lin, D., and
Banjevec, D. An integrated platform for diagnostics,
prognostics and maintenance optimization. Proceed-
ings of the Intelligent Maintenance Systems 2004,
15–27 July 2004, Arles, France.
4 Lee, J. and Ni, J. Smart prediction to prevent downtime.
Innovation, 2004, 4(3), 41–42.
5 Djurdjanovic, D., Lee, J., and Ni, J. Watchdog Agent –
an infotronics-based prognostics approach for product
performance degradation assessment and prediction.
Advd Engng Inf., 2003, 17(3–4), 109–125.
6 Grudzien, W. and Seliger, G. Life cycle unit in product
life cycle – tool for improved maintenance, repair and
recycling In Proceedings of 33rd CIRP International
Seminar on Manufacturing Systems, Stockholm 5–7
June 2000, pp. 143–148 (International Institution for
Production Engineering Research).
7 Hirschmann, J. Fault detection and diagnosis of
the electromechanical drive units in the automation
technology. Conference Proceedings: E-Manufacturing
and E-Business Integration, Milwaukee, Wisconsin,
USA, 9–11 September, 2002, pp. 179–181.
8 Middendorf, A., Buchholz, A., Hefer, J., Reichl, H.,
Schrank, K., and Seliger, G. Life-cycle information units
for monitoring and identification of product use condi-
tions. Proceedings of Global Conference on Sustainable
Product Development and Life Cycle Engineering 2004,
29 September–1 October, 2004, Berlin, pp. 91–96.
9 Walter, K.-D. Wireless fu
¨r Pra
¨ventiv-Aufgaben. Compu-
ter & Automation, 2006, 01, 36–39.
10 Heinz, J. ePS Network Services – internet based condi-
tion monitoring of CNC machines. Lecture held at the
‘International Machine Tool Forum’ at IMTEX 2007,
Bangalore, India, 20 January 2007, Copyright ÓSiemens
AG 2006.
11 ePS Network Services, Function Manual (FH)-Release
04/2007, Copyright ÓSiemens AG 2007.
12 Barth, R. Produktiver mit Condition Monitoring. In:
WB Werkstatt þBetrieb 09/2005, pp. 183–185 (Carl
Hanser Verlag, Munich).
13 Available online at www.projekt-loewe.com
Fig. 9 Data transformation from XML to SQL
Proc. IMechE Vol. 222 Part B: J. Engineering Manufacture JEM1040 ÓIMechE 2008
1218 E Uhlmann, C Geisert, and E Hohwieler
14 Denkena, B., Harms, A., Jacobsen, J., M€
ohring, H.-C.,
Lange, D., and Noske, H. Life-cycle oriented develop-
ment of machine tools. In Proceedings of the 13th
CIRP International Conference on Life Cycle Engineer-
ing, LCE 2006, 31 May–2 June 2006, Katholieke Universi-
teit Leuven, Belgium, pp. 693–698.
15 Imiela, J. Verfu
¨gbarkeitssicherung von Werkzeugmaschi-
nenachsen mit Kugelgewindetrieb durch modellbasierte
Verschleibu
¨berwachung. Doctoral Thesis, PZH-Verlag,
Hannover, Germany, 2006.
16 Available online at http://www.cordis.lu/data/PROJ_
BRITE/ACTIONeqDndSESSIONeq23541200595ndDOCeq
115ndTBLeqEN_PROJ.htm. (accessed 1 November 2007).
17 Available online at http://www.cordis.lu/data/RESU_
BRITE/ACTIONeqDndSESSIONeq23864200598ndDOCeq
152ndTBLeqEN_RESU.htm. (accessed 1 November 2007).
18 Plapper, V. and Weck, M. Sensorless machine tool
condition monitoring based on open NCs. Proceedings
of the 2001 IEEE International Conference on Robo-
tics and Automation, Seoul, Korea 21–26 May 2001,
pp. 3104–3108.
19 Plapper, V. Steuerungsintegrierte U
¨berwachung von
Werkzeugmaschinen. Doctoral Thesis, Aachen, RWTH,
2004.
20 Dietmair, A., Walther, M., and Pritschow, G. Antriebs-
basierte Maschinendiagnose – Fortschritte bei der
Nutzung von Antriebssignalen zur Maschinendiagnose.
wt Werkstattstechnik online, 2005, 95(5), 351–356.
21 Walther, M. and Verl, A. Maschinendiagnose auf Basis
von Antriebssignalen. In Zuverla
¨ssigkeit und Diagnose
in der Produktion, Fortschritt-Berichte VDI, Vol. No.
663, VDI Verlag, Du
¨sseldorf, Germany, 2007, pp. 39–55.
22 Uhlmann, E., Hohwieler, E., and Berger, R. Providing
online services for machine operation and mainte-
nance. In Proceedings of the International IMS Forum
2004, Vol. 2: Global challenges in manufacturing, Villa
Erba, Cernobbio, Italy, 17–19 May 2004, pp. 1120–1127.
23 Hohwieler, E., Berger, R., and Geisert, C. Con-
dition monitoring services for e-Maintenance. In
Proceedings of the 7th IFAC Symposium on Cost
Oriented Automation, 7–9 June 2004, Gatineau/Ottawa,
Canada.
24 Hohwieler, E. and Geisert, C. Intelligent machines offer
condition monitoring and maintenance prediction
services. In Proceedings of the 4th CIRP International
Seminar on Intelligent Computation in Manufacturing
Engineering (CIRP ICME ’04), (Ed. R. Teti), 30 June–2
July, 2004, Sorrento, Italy, pp. 599–604.
25 Berger, R. Sicherheitszentrierte Architektur fu
¨r Internet-
basierte Dienste im Maschinen und Anlagenbau. Doc-
toral Thesis, Technische Universita
¨t, Berlin, 2005.
26 Kru
¨ger, J. Methoden zur Verbesserung der Fehlererken-
nung an Antriebsstrecken. Doctoral Thesis, Technische
Universita
¨t, Berlin, 1998.
27 Hohwieler, E. and Geisert, C. Holistic approach for
condition monitoring and prediction of machine axis.
In Proceedings of First I*Proms Virtual Conference on
Intelligent production machines and systems, 4–15 July
2005 (Eds D. T. Pham, E. E. Eldukhri, and A. J. Soroka),
Elsevier, 2005, pp. 67–72.
28 Geisert, C. and Uhlmann, E. Steuerungsbasierte
Erfassung zustandsrelevanter Life-Cycle Daten an CNC-
Werkzeugmaschinen. In Zuverla
¨ssigkeit und Diagnose
in der Produktion, Fortschritt-Berichte VDI, Vol. 2, No.
663, VDI Verlag, Du
¨sseldorf, Germany, 2007, pp. 1–14.
JEM1040 ÓIMechE 2008 Proc. IMechE Vol. 222 Part B: J. Engineering Manufacture
Monitoring of slowly progressing deterioration of CNC-machine axes 1219
