F R A U N H O F E R I N S T I T U T E F O R I N T E G R AT E D C I R C U I T S I I S
D E S I G N A U T O M AT I O N D I V I S I O N E A S
Photo: hramovnick - Fotolia
SELF-LEARNING
CONDITION MONITORING
To run an efficient business, most companies
friendly solution, saving time, effort and
find it essential to have optimum availability
thereby costs as well.
of their machinery and systems at all times.
Unplanned downtimes or unnecessary
component exchanges create preventable
Your Benefits
costs and reduce productivity. Condition
Monitoring Systems (CMS) can provide a
■
resulting from wear
very valuable service in these areas.
■
service life to increase the operating
furnishes meaningful information about
hours between maintenance
the condition of wear on components like
Integrated Circuits IIS
motors, pumps and bearings. It facilitates
Design Automation Division EAS
a reliable determination of maintenance
■
Zeunerstrasse 38
a CMS is the enormous amount of data it
01069 Dresden, Germany
generates, consisting of sensor values and
sometimes specific process data. This can
Contact:
complicate using such systems and make
Dr. Olaf Enge-Rosenblatt
them rather difficult to control.
Avoiding consequential damage because of unrecognized wear situations
■
Shortening of maintenance related
downtimes
and repair needs. The downside of such
■
Reducing the company's spare parts
stock
■
CMS operation without the need for
extensive prior technical knowledge
■
Phone +49 351 4640-711
»Learning« CMS with automated
feature selection
The Fraunhofer IIS/EAS approach for an
■
intelligent and independent monitoring
www.eas.iis.fraunhofer.de/en.html
Full maximization of the component's
Early on, a predictive CMS approach
Fraunhofer Institute for
olaf.enge-rosenblatt@eas.iis.fraunhofer.de
Early detection of possible damage
of system components eliminates this
disadvantage. It provides a quick and user-
Automatic limit value selection or
tracking through trend identification
■
Cost reduction for commissioning a
CMS
Photo: industrieblick - Fotolia
1
The intelligent CMS
This is how the CMS expands its data-
■
Data analysis for CMS users
base step-by-step with measurements of
Examining existing data for abnor-
The key feature of the Fraunhofer
various wear and operating conditions
malities and automatic generation of
approach is its significantly simplified
and monitors all previous classification
appropriate differentiators
operation, linking automatic data analysis
features continuously. When they no
with a self-learning classification. To do
longer represent the current situation, the
■
Customized CMS solutions
this, the Condition Monitoring System
differentiators of the data fingerprints will
Development and implementation
must establish settings, such as the limit
be recalculated and an adjusted set of
of user specific methods to extract
values, automatically. This is achieved by
specific features is established.
features and integrate algorithms, e.g.
means of mathematical algorithms, which
in embedded systems
evaluate known operating conditions of a
Our Services
■
Any changes in this data fingerprint are
We offer the following services against
automatically detected. Outliers are filtered
the background of our learning CMS
out, to eliminate false alarms. Once enough
solution:
plant. This is how the CMS »learns« the
data characteristics of various conditions.
Software development for CMS
providers
Implementation of additional software
modules for automated data analysis
measurements are available, limit values
can be set automatically. After completing
■
Feasibility studies
the initial learning phase only minimum
Analysis of customer-specific require-
additional effort is required for operating
ments and recommendations for data
the CMS.
recording
1 An inexpensive and userfriendly CMS is especially important for electrical motors used in
The CMS executes regularly scheduled
industrial plant systems.
measurements of values which are
relevant to wear monitoring. This results in
a data pool, which is evaluated together
Example of Various Clusters in a 3D Feature Space
with the respective process data. The
ultimately relevant differentiators are
The features used
obtained automatically by targeted
were determined in an
selection from a wide range of statistical
automated process.
parameters. Each record represents a point
Because of the large
in a feature space, which can then be
distance of the cluster
captured visually. Multiple measurements
in relation to each
of operating conditions form clusters. They
other (scatter) the
initially represent different operating con-
conditions are easily
ditions. The first appearance of machine
separable. The features
wear will change the form and location of
are therefore suitable
the clusters. As a consequence, this can
for classification.
then be detected in the fingerprint and
the diagnosed change is reported.
Error-free
Gearbox Damage
Bearing Damage 1
Bearing Damage 2
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0
0.2
0.4
0.6
0.8
1