Holistic System-Analytics as an Alternative to Isolated Sensor Technology: A Condition Monitoring Use Case

Conference Paper (PDF Available) · January 2019with 201 Reads
DOI: 10.24251/HICSS.2019.124
Conference: Hawaii International Conference on System Sciences (HICSS-52), At Grand Wailea, Maui, Hawaii
Cite this publication
Abstract
Sensor technology has become increasingly important (e.g., Industry 4.0, IoT). Large numbers of machines and products are equipped with sensors to constantly monitor their condition. Usually, the condition of an entire system is inferred through sensors in parts of the system by means of a multiplicity of methods and techniques. This so-called condition monitoring can thus reduce the downtime costs of a machine through improved maintenance scheduling. However, for small components as well as relatively inexpensive or immutable parts of a machine, sometimes it is not possible or uneconomical to embed sensors. We propose a system-oriented concept of how to monitor individual components of a complex technical system without including additional sensor technology. By using already existing sensors from the environment combined with machine learning techniques, we are able to infer the condition of a system component, without actually observing it. In consequence condition monitoring or additional services based on the component's behavior can be developed without overcoming the challenges of sensor implementation.
This is the author’s version of a work that is published in the following source
Martin, D.; Kühl, N. (2019). Holistic System-Analytics as an Alternative to Isolated
Sensor Technology: A Condition Monitoring Use Case. In Proceedings of the Hawaii
International Conference on System Sciences (HICSS-52), Grand Wailea, Maui, Hawaii.
Please note: Copyright is owned by the author and/or the publisher. Commercial use is not
allowed.
Karlsruhe Service Research Institute
Kaiserstr. 89
76133 Karlsruhe
Germany
Holistic System-Analytics as an Alternative to Isolated Sensor Technology:
A Condition Monitoring Use Case
Dominik Martin
Karlsruhe Institute of Technology (KIT)
martin@kit.edu
Niklas K¨
uhl
Karlsruhe Institute of Technology (KIT)
kuehl@kit.edu
Abstract
Sensor technology has become increasingly
important (e.g., Industry 4.0, IoT). Large numbers
of machines and products are equipped with sensors
to constantly monitor their condition. Usually, the
condition of an entire system is inferred through sensors
in parts of the system by means of a multiplicity of
methods and techniques. This so-called condition
monitoring can thus reduce the downtime costs of a
machine through improved maintenance scheduling.
However, for small components as well as relatively
inexpensive or immutable parts of a machine, sometimes
it is not possible or uneconomical to embed sensors. We
propose a system-oriented concept of how to monitor
individual components of a complex technical system
without including additional sensor technology. By
using already existing sensors from the environment
combined with machine learning techniques, we are
able to infer the condition of a system component,
without actually observing it. In consequence condition
monitoring or additional services based on the
component’s behavior can be developed without
overcoming the challenges of sensor implementation.
1. Introduction and Motivation
With the rapid development of advanced sensor
technology and the decline in hardware costs an
increasing number of machines and products has been
equipped with sensors to acquire data on assets behavior
[1, 2, 3]. Condition monitoring uses this data in order
to facilitate the detection of machine malfunctions in
an early state, and, thus, minimizing consequential
damage, enhancing maintenance work scheduling as
well as reducing downtime costs [4].
Industry as well as academia face the challenge
of developing condition monitoring solutions for
machines, products, systems and components. Most
research focuses on the development of smart materials
equipped with sensors, or equipping machines or
components with a large number of sensors designed for
this particular purpose [5, 6, 7].
However, in many cases it is either not technically
feasible or uneconomic to monitor the components
or parts of interest with dedicated sensors [8, 9,
10]. For instance, it is not possible to incorporate
sensor technology into a part of interest—such as
i.e., a seal—due to, for instance, technical reasons or
unreasonable expenses.
Especially for very complex assets (i.e., aircraft
engines, production machines, wind turbines), the
reliability of each individual component is of enormous
importance. Such complex assets consist of a number
of components that often again are composed of
subcomponents produced by various manufacturers.
The manufacturing at the entire system level thus
requires only little knowledge about the behavior of all
parts down to the smallest components. Therefore, the
OEMs do not necessarily have the technical know-how
to develop high-quality condition monitoring solutions
for each individual component.
Nevertheless, it may be interesting to monitor even
small and seemingly negligible parts in order to gain
insights about the part’s behavior within a complex
environment, to differentiate from competitors or to be
able to offer additional services. Such services could
be, for instance, maintenance services, wear prediction
or analysis tools for a specific component or part.
Even condition monitoring as a service could be a
conceivable business model. Furthermore, condition
monitoring allows to prevent consequential damage to
the overall system due to a component malfunction.
However, especially for suppliers of very cheap or small
components, it is often not possible to offer their own
condition monitoring solutions for their products, since
an observation of these components in the context of
the overall system is too costly. Thus, for instance,
embedding special sensors into a part or even customize
the surrounding system of a component that cost a few
cents is generally not profitable.
In this work, we propose a concept to observe the
state of individual parts by a system-wide consideration
without actually being able to observe them directly.
By applying machine learning techniques, we show
how conclusions about individual components can be
drawn by considering the entire system. Thus, our
approach allows to make statements about the behavior
or condition of subordinate parts without additional
hardware being necessary.
We show that, despite complex interrelations within
a system, as described by General Systems Theory [11,
12] and the idea of holism [13, 14], condition monitoring
of components of a system can be accomplished by
means of machine learning.
2. Foundations and Related Work
We consider machines, products and assets
as systems and therefore use systems theory—an
interdisciplinary approach for describing and explaining
aspects, properties and principles of systems—to
describe interrelations and the behavior of systems and
its components. Subsequently, we introduce the concept
of condition monitoring and discuss related literature
regarding existing condition monitoring approaches.
2.1. Systems Theory
The aim of classical physics was to reduce natural
phenomena to an interaction of elementary units—such
as atoms—whose properties are independent of the
environment, and thus of adjacent units. Similarly, in
biology, the assumption was that phenomena of life can
be resolved into separate parts that can be considered in
isolation. Accordingly, an organism can be thought of
as the interplay of various elements—such as cells—that
function independently [11].
This traditional scientific theory for exploring single
components without regard to the surrounding units is
known as reductionism. Reductionism is a bottom-up
approach and tries to deduce the understanding of
the whole from the understanding of individual parts.
However, a whole aggregated of parts is not immediately
apparent from the individual parts [15].
This problem arises from complex interrelationships
between the individual parts, from which a whole
emerges. Only the whole gives the parts and their
interactions a meaning. For instance, only a living
organism gives meaning to the heart or lungs as well as
a family to the roles of husband, wife, son or daughter
[16].
According to this assumption, and contrary to the
idea of reductionism, modern biology has developed
the notion that not only the consideration of isolated
parts, but also of relations between parts and the
resulting dynamic interactions is essential. This results
in differences between individual parts considered
separately and parts within a whole organism [11, 12].
The biologist Ludwig von Bertalanffy proposed
systems theory in the 1940’s, in which he transferred and
generalized this idea to other scientific disciplines [17].
He defines systems as interaction contexts that differ
from their environment, which in turn consists of other
interaction contexts. Through their interaction with the
environment, they can evolve new properties. System
theory does not reduce the whole (i.e., the human body)
to the behavior or properties of its parts (i.e., organs), but
rather describes the relationships and arrangement of the
parts which build the whole [17].
Accordingly, holism is the idea that systems
(biological, physical, social, economic, etc.) and their
properties should be viewed as wholes, not just as a
collection of parts [13, 14].
Based on this, Ackoff [18] defines a system as a
set of two or more elements that satisfies following
conditions: The behavior of each element has an effect
on the behavior of the whole, where both the behavior
of the elements as well as their effects on the whole are
interdependent. Additionally, all subgroups of elements
have an effect on the behavior of the whole but none has
an independent effect on it.
2.2. Condition Monitoring
Observing the behavior of systems and their
elements poses a complex challenge in various
disciplines, such as biology, physics, medicine, but
also in computer science or mechanical engineering.
Especially in industry, the continuous observation of
machine or product conditions is an important factor for
smooth operations.
The concept of condition monitoring describes the
regular acquisition and analysis of physical parameters
for monitoring the condition of machines. The physical
machine parameters are collected by means of sensors
and include vibration [19, 20], temperature [21, 22],
acoustic emission [23, 24], electricity [25], and many
more.
By identifying significant changes in the sensor
readings, the occurrence of machine failures is
determined by a variety of methods [26]. This makes
condition monitoring an important basis for predictive
maintenance. By being aware of the machine conditions
at any time, maintenance work can be better planned
and accordingly machine downtimes reduced [26]. In
addition, further actions can be made to minimize
consequential damage as well as the effects of a failure
on surrounding components. Condition monitoring
allows the earlier detection of machine failures, before
their effects can be directly perceived by humans.
Thus, condition monitoring is an effective alternative to
periodic machine inspections [27].
Especially with rotating machines, such as motors,
pumps, and compressors, condition monitoring methods
can be used effectively [20]. However, interesting
research approaches are also available for translatory or
even static applications.
For instance, Nandi et. al. [26] describe different
components which indicate failures in electric motors.
In addition to the typical failure types occurring in
different components, common recognition and analysis
techniques are summarized. However, the focus relies
on the analysis of conditions of the entire electric motor
system. A targeted analysis of components such as the
bearing or even components of it, without dedicated
sensors, is not considered.
Also, Carden and Fanning [20] summarize different
methods and techniques which appear in literature,
which are capable of detecting failures based on the
analysis of vibrations. They argue that different methods
require a varying number of sensors and the results
improve as the number of sensors increases.
This work does neither want to treat the integration
of new sensors into a system, nor to consider the
behavior of the whole system. Our work focuses on
monitoring a single component of a system that does
not have integrated sensing capabilities by using already
available sensors from the component’s environment.
Thus, we aim to show the feasibility of inferring the
condition of such a system component even without
dedicated embedded sensors.
3. Research Design
Our goal is to present a concept to gain insights about
the behavior of a part of a system based on data gathered
from sensors in the part’s environment. Thus, our first
research question is:
RQ1: How can we design a concept that allows
us to derive the behavior of a part of a system based
on system-wide considerations, without being able to
observe this directly?
We follow the research paradigm of Design Science
Research, which is widely accepted in the field of
information systems [28, 29]. The central idea
of the Design Science Research approach is to
gain an understanding of the problem domain and
scientifically founded knowledge while at the same
time guaranteeing practical relevance [30]. This is
achieved by developing an artifact. Our artifact is
a construct [31], which is evaluated by a technical
experiment [28]. Thus, our knowledge contribution is
thereby an exaptation, since we extend already known
solutions (application of General Systems Theory) to
new problems (system-based condition monitoring for
components) [29].
By developing an artifact in a first design cycle, we
aim to show the feasibility of inferring the behavior of
a single component within a system without actually
observing it, using already available sensors.
Based on this concept, we aim to develop a condition
monitoring system artifact for hydraulic cylinders,
which is capable of detecting seal failures on the basis
of sensors within the cylinder, but, however, without
being able to observe the seal directly. Thus, we want
to evaluate the feasibility of our proposed concept by
means of a concrete technical experiment. Therefore,
the second research question is as follows:
RQ2: By applying a systemic view on the hydraulic
cylinder system, can we infer the state of individual
components without directly monitoring them itself?
After demonstrating the feasibility, we also provide
a prototype that demonstrates a possible use case.
4. Concept of Holistic System-Analytics
As described by Ackoff [18], systems consist of
several elements that are exposed to each other’s
interactions and thus influence the behavior of the
entire system. According to the idea of holism, the
total system is more than the sum of its parts. This
contradicts the approach of reductionism, which intends
to analyze all individual components of a system in
order to comprehensively describe the behavior of
the entire system. Based on these assumptions, the
behavior of a system from a holistic viewpoint can not
be fully explained by the behavior of the individual
elements. Thus, no direct inferences from the overall
system behavior to its elements are possible, since the
behavior of the entire system is influenced by complex
interdependent relationships of the system elements.
However, even if there is no direct relationship
between the behavior of the entire system and individual
elements of it, each of the effects of an element
influences the behavior of the entire system. Thus, by
observing these more complex relationships on a system
level, patterns can be recognized which are caused by
individual elements.
Typically, monitoring of technical systems is
achieved by incorporating sensors and observation
capabilities into its components. Individual components
in complex systems are analyzed separately. Based on
this information, conclusions about the behavior of the
entire system are derived [26].
Figure 1 visualizes this common approach of a
system-level analysis. It shows an open system, which
consists of a set of elements having interdependent
effects on the entire system. The system itself also
interacts with the environment. Individual elements of
the system can be observed by i.e., sensors (elements A
and C), while other elements are unobservable (element
B). Using the observable behavior of elements and
the observable interaction of the entire system with
the environment, conclusions can be drawn about the
behavior of the entire system.
Figure 1. System-level analysis
Our proposed concept describes a contradictory
approach. This concept describes a method to
analyze elements of a complex technical system in
which an element of interest on the one hand is not
directly observable and on the other hand has no
trivial connections between the behavior of itself and
the behavior of the entire system. Based on data
representing information about the behavior of the entire
system, machine learning is used to identify the effects
of individual components on the entire system. This
approach allows to observe the behavior and thus the
state of system components, which are not observable
directly.
This approach is visualized in figure 2. Therefore,
the behavior of an element within the entire system
is of interest. However, this behavior can not be
observed directly. Thus, by observing the interaction
of the system with the environment, the impact of the
element’s behavior (element B) on these interactions is
inferred.
This systemic top-down approach, in which the
behavior of the entire system serves as basis for the
analysis of system elements, requires a technique to
identify patterns in the interactions of the system
with the environment and, to isolate the corresponding
causers of these patterns.
We use machine learning techniques for tackling
Figure 2. Component-level analysis
this and demonstrate the feasibility of our approach
by means of an industrial use case. Thus, our
proposed concept of compensating missing sensors in
components by applying machine learning techniques to
infer component behavior represents an Design Science
Research artifact.
5. Evaluation and Discussion
This section evaluates the described artifact
(presented concept). Therefore, we use an industrial
use case to assess the feasibility of our approach. We
define a hydraulic cylinder as a system in which a seal
(component) is to be observed without equipping it with
sensors.
5.1. Technical Experiment
A hydraulic system is used to transfer power from
a hydraulic pump to a piston. By introducing a
pressurized hydraulic fluid into a cylinder, the piston
within the cylinder is moved. The medium necessary
for this is a hydraulic fluid, such as oil, which is usually
heavily pressurized. Depending on the application (i.e.,
an excavator) enormous forces act on the piston that are
transmitted from the hydraulic system. The seals which
seal the piston rod to the cylinder wall are thus essential
for the function of such a hydraulic system.
Seals, even if they seem unimportant and
inconspicuous, are an essential component of various
applications. The catastrophic incident of the NASA
space shuttle Challenger in 1986 caused by a damaged
seal, reveals that seal failures can lead to dramatic
effects [27].
As described in section 3, we develop a machine
learning model, which is able to infer the behavior, or
the state of a seal within a hydraulic cylinder based on
sensor time series obtained from the surrounding of the
seal itself.
The intention of this model is the real-time detection
of failure scenarios. Seals are sensitive components
of hydraulic applications. Thus, a malfunction of
sealing elements results in inadequate sealing efficiency
and thus—in extreme cases—in a failure of the entire
application. Due to the high sensitivity, a correct
installation of an undamaged seal must be ensured.
The model has to be capable to detect i.e., a possible
assembly failure or a damaged seal.
Thus, we define three failure classes: no failure,
assembly failure and damage. Data gained from
different tests conducted on hydraulic cylinder test rigs
serves as basis for the machine learning model, which is
able to recognize patterns in the data and derive a class
assignment accordingly. In order to obtain a wide range
of different failure scenarios, various assembly failures
and damages are simulated in several test series.
In order to train the classification model 400 hours of
data with a frequency of 20 Hz captured by 13 sensors
is available.
These sensors capture different pressures,
temperatures, and rod velocity of the hydraulic
cylinder. The data is cleaned and features are extracted.
The failure classes are used as labels.
We use a Random Forest Classifier, as it achieves
good performance in pre-tests with low training time.
For model training and validation by conducting a grid
search, we use 10-fold nested cross validation to avoid
overfitting [32].
By using the F1-score [33] as a performance metric,
the classification model achieves an average score of
0.971. In addition, we use neural networks, decision
trees, and support vector machines as classification
algorithms, all of which, however, yield lower
performance (below 0.94).
5.2. Prototype
As described in section 5.1 the condition of
seals can be classified by the application of machine
learning algorithms. This proves the general feasibility
of inferring the condition of seals without directly
implemented sensors. This opens up the possibility
of making this knowledge accessible and usable for
third parties (i.e., customers). Therefore, we deploy the
previously developed model as a web service, thus, it
can be accessed via internet.
The architecture of this prototype is shown in figure
3. The raw data collected at the hydraulic cylinder
is provided unprocessed to an IoT gateway. This
gateway allows secure and scalable communication
between IoT devices and the cloud, and makes the
data available to downstream processes, like stream
Figure 3. Architecture of condition monitoring
prototype
processing. Streaming allows data to be processed in
real-time and delivered to the deployed model realized
by a RESTful web service. The web service handles
the data cleaning and feature extraction and uses the
implemented classifier to determine the corresponding
failure class of the seal. This information is passed back
to the stream module as a response. Subsequently, the
raw data, extended to the determined failure information
is passed to the dashboard module. This allows the
visualization of real-time data in a browser. In addition,
the raw data, extended by the classification results,
is stored in a database. Optionally, the determined
classification results can be transmitted back to the
hydraulic cylinder’s control unit in order to initiate
appropriate actions there. For practical applications,
changing load conditions or switching off the affected
cylinder would be conceivable.
Figure 4. Screenshot of condition monitoring
dashboard prototype
Figure 4 shows a screenshot of the dashboard
showing the condition of the sealing elements in
real-time. Thus, visual monitoring of the seal condition
is possible with a simple graphical interface. The first
tile at the top left shows the current seal condition on a
virtual scale from 0 to 1. This is an indicator calculated
from the prediction probability and the moving average
of historical inferences. The second tile shows the
current classified seal condition and the third tile the
corresponding probability of a seal failure in the present
classification result. The last tile in the first row shows
the relative frequency of classified failure classes since
system ramp-up. The lower part of the dashboard shows
current sensor raw data in real time.
5.3. Discussion
This work shows how machine learning can be
used to obtain insights about the behavior of a system
component by data obtained from system-level sensors.
Thus, there are some practical implications.
Component manufacturers which previously provided
only physical products thus have the opportunity to
monetize their domain know-how and offer additional
services supporting their core products [34]. However,
collaboration with data owners and manufacturers of
the immediate environment of the part or component,
including the sensors, is required. Embedding
component services to the entire system level, however,
also creates additional value concerning the entire
system. Furthermore, failure localization within a
system is a complex problem, which could be solved
hereby. In general, failures are only analyzed at the
system level. Thus, an investigation on failure causes
needs to be conducted additionally. Knowing the exact
location of a failure within a complex system can
therefore lead to less time-consuming troubleshooting
[35].
6. Conclusion and Outlook
Due to embedded sensors, machines and products
produce large amounts of data which is often only
used for a specific purpose. The entire potential of
this already existing data is therefore far from being
exploited. Furthermore, some manufacturers face the
challenge that some components of complex technical
systems can not be monitored because no sensors can be
implemented for technical or economic reasons.
Therefore we propose a concept how this problem
can be resolved. By means of a system-wide analysis
of existing sensor data from the immediate environment
of the part of interest through machine learning, it is
possible to extract patterns that can be used to deduce
the behavior of parts of a system.
Based on an industrial use case, we show that the
condition of a seal within a hydraulic cylinder, even
though no sensors are integrated into it, can be inferred
from the sensors in the hydraulic cylinder surrounding
the seal. A classification technique analyzes the effects
of the different conditions of a seal on the behavior of the
entire system and, thus, the system-level sensor values.
The results show that patterns caused by the different
conditions can be detected within the data captured with
surrounding sensors.
In addition to these contributions, this work also
has limitations. We assume that the sensors in
the environment record patterns that affect a system
component. This can not be proven by the technical
experiment presented in this work. In addition, the
demarcation of a system in general is not clearly
defined. This requires a case-by-case analysis.
Furthermore, the application of machine learning also
has shortcomings compared to the analysis of physical
coherences. The analysis of cause-effect relationships
using mathematical models can also be feasible in
some cases, however, without the need for collecting
training data. However, as with the analysis of physical
relationships between components, the sensors used
must deliver reliable and stable values over time to
ensure the quality of the classification model. A
systematic shift in sensor readings over time or a
changing sensor configuration can result in decreasing
performance.
Nevertheless, we are convinced that this concept
provides an exciting research approach. In the next
steps, it remains to analyze whether the quality of
this approach is comparable to sensor-based condition
monitoring in use cases where sensors can be equipped.
Thus, our proposed concept could potentially replace
sensors within components, which could lead to a
decrease in overall system costs.
Furthermore, it has to be investigated which effects
the behavior of a system component has on the
behavior of other system components. Thus, the
concept could be extended if the observation of other
(observable) components provides additional insights on
the (unobservable) target component. This would be
a mixture of the traditional approach (observation of
individual components to analyze the overall system
conditions) and the concept presented in this work.
In addition, it can be analyzed how the quality of this
approach, compared with sensing the component itself,
changes for other use cases as well as for other system
boundaries.
Our presented use case could also be extended by
increasing the granularity of the failure classes leading
to more detailed insights.
Finally, we are convinced that our concept can
be used in many applications and leads to a broader
understanding about system components. This holds the
potential to make a whole range of applications more
transparent and reliable.
References
[1] F. Civerchia, S. Bocchino, C. Salvadori, E. Rossi,
L. Maggiani, and M. Petracca, “Industrial Internet of
Things monitoring solution for advanced predictive
maintenance applications,” Journal of Industrial
Information Integration, vol. 7, pp. 4–12, 2017.
[2] J. Gubbi, R. Buyya, S. Marusic, and M. Palaniswami,
“Internet of Things (IoT): A vision, architectural
elements, and future directions,” Future Generation
Computer Systems, vol. 29, no. 7, pp. 1645–1660, 2013.
[3] S. A. Macskassy and F. Provost, “A Brief Survey
of Machine Learning Methods and their Sensor
and IoT Applications,” International Conference on
Information, Intelligence, Systems & Applications
(IISA), pp. 172–175, 2017.
[4] A. K. Jardine, D. Lin, and D. Banjevic, “A review
on machinery diagnostics and prognostics implementing
condition-based maintenance,” Mechanical Systems and
Signal Processing, vol. 20, no. 7, pp. 1483–1510, 2006.
[5] O. Uluyol, G. Parthasarathy, W. Foslien, and K. Kim,
“Power Curve Analytic for Wind Turbine Performance
Monitoring and Prognostics,” Annual Conference of
the Prognostics and Health Management Society,
no. August, pp. 1–8, 2011.
[6] V. J. Hodge, S. O. Keefe, M. Weeks, and A. Moulds,
“Wireless Sensor Networks for Condition Monitoring in
the Railway Industry: A Survey,IEEE Transactions
on Intelligent Transportation Systems, vol. 16, no. 3,
pp. 1088–1106, 2015.
[7] B. Peeters, J. Maeck, and G. D. Roeck, “Vibration-based
damage detection in civil engineering: Excitation
sources and temperature effects,Smart Materials and
Structures, vol. June, no. October 2014, 2001.
[8] P. Kassanos, S. Anastasova, and G.-Z. Yang, “Electrical
and Physical Sensors for Biomedical Implants,” in
Implantable Sensors and Systems: From Theory to
Practice (G.-Z. Yang, ed.), pp. 99–195, Cham: Springer
International Publishing, 2018.
[9] P. Kassanos, H. Ip, and G.-Z. Yang, “Ultra-Low Power
Application-Specific Integrated Circuits for Sensing,
in Implantable Sensors and Systems: From Theory to
Practice (G.-Z. Yang, ed.), pp. 281–437, Cham: Springer
International Publishing, 2018.
[10] A. J. Thompson and G.-Z. Yang, “Tethered and
Implantable Optical Sensors,” in Implantable Sensors
and Systems: From Theory to Practice (G.-Z. Yang, ed.),
pp. 439–505, Cham: Springer International Publishing,
2018.
[11] L. v. Bertalanffy, “An Outline of General System
Theory,The British Journal For Philisophy Of Science,
vol. 1, no. 2, pp. 134–165, 1950.
[12] P. Checkland, “Systems thinking,” in Rethinking
Management Information Systems, pp. 45–56, Oxford
University Press, 1999.
[13] B. Oshry, Seeing Systems: Unlocking the Mysteries of
Organizational Life. Berrett-Koehler Publishers, Inc.,
2007.
[14] S. Y. Auyang, “Foundations of Complex-system
Theories,” in Economics, Evolutionary Biology, and
Statistical Physics, pp. 1–404, 1999.
[15] F. C. Fang and A. Casadevall, “Reductionistic and
holistic science,” Infection and Immunity, vol. 79, no. 4,
pp. 1401–1404, 2011.
[16] M. C. Jackson, Systems Thinking: Creative Holism for
Managers. John Wiley & Sons Ltd., 2007.
[17] K. E. Boulding, “General Systems Theory - The
Skeleton of Sciene,” Management Science, vol. 2, no. 3,
pp. 197–208, 1956.
[18] R. L. Ackoff, Creating the Corporate Future: Plan or be
Planned For. New York: Wiley, 1981.
[19] Z. K. Peng and F. L. Chu, “Application of the
wavelet transform in machine condition monitoring
and fault diagnostics: A review with bibliography,
Mechanical Systems and Signal Processing, vol. 18,
no. 2, pp. 199–221, 2004.
[20] E. P. Carden and P. Fanning, “Vibration Based Condition
Monitoring: A Review,Structural Health Monitoring:
An International Journal, vol. 3, no. 4, pp. 355–377,
2004.
[21] W. Zhou, T. G. Habetler, and R. G. Harley, “Bearing
Condition Monitoring Methods for Electric Machines: A
General Review,” 2007 IEEE International Symposium
on Diagnostics for Electric Machines, Power Electronics
and Drives, pp. 3–6, 2007.
[22] S. Bagavathiappan, B. B. Lahiri, T. Saravanan,
J. Philip, and T. Jayakumar, “Infrared thermography for
condition monitoring - A review,” Infrared Physics and
Technology, vol. 60, pp. 35–55, 2013.
[23] D. A. Dornfeld and M. DeVries, “Neural Network
Sensor Fusion for Tool Condition Monitoring, CIRP
Annals - Manufacturing Technology, vol. 39, no. 1,
pp. 101–105, 1990.
[24] M. S. H. Bhuiyan, I. A. Choudhury, M. Dahari,
Y. Nukman, and S. Z. Dawal, “Application of acoustic
emission sensor to investigate the frequency of tool wear
and plastic deformation in tool condition monitoring,”
Measurement: Journal of the International Measurement
Confederation, vol. 92, pp. 208–217, 2016.
[25] F. C. Trutt, J. Sottile, and J. L. Kohler, “Online condition
monitoring of induction motors,” IEEE Transactions on
Industry Applications, vol. 38, no. 6, pp. 1627–1632,
2002.
[26] S. Nandi, H. A. Toliyat, and X. Li, “Condition
monitoring and fault diagnosis of electrical motors -
A review,” IEEE Transactions on Energy Conversion,
vol. 20, no. 4, pp. 719–729, 2005.
[27] I. Alsyouf, “The role of maintenance in improving
companies’ productivity and profitability,” International
Journal of Production Economics, vol. 105, no. 1,
pp. 70–78, 2007.
[28] K. Peffers, M. Rothenberger, T. Tuunanen, and R. Vaezi,
“Design Science Research Evaluation,Design Science
Research in Information Systems. Advances in Theory
and Practice, pp. 398–410, 2012.
[29] S. Gregor and A. R. Hevner, “Positioning and Presenting
Design Science Research for Maximum Impact,” MIS
Quarterly, vol. 37, pp. 337–355, 12 2013.
[30] A. R. Hevner, S. T. March, J. Park, and S. Ram,
“Design Science in Information Systems Research,” MIS
Quarterly, vol. 28, no. 1, pp. 75–105, 2004.
[31] S. T. March and G. F. Smith, “Design and natural science
research on information technology,” Decision Support
Systems, vol. 15, no. 4, pp. 251–266, 1995.
[32] G. C. Cawley and N. L. C. Talbot, “On Over-fitting
in Model Selection and Subsequent Selection Bias in
Performance Evaluation,Journal of Machine Learning
Research, vol. 11, p. 20792107, 2010.
[33] D. Powers, “Evaluation: From Precision, Recall
and F-Measure To Roc, Informedness, Markedness
& Correlation,” Journal of Machine Learning
Technologies, vol. 2, no. 1, pp. 37–63, 2011.
[34] T. S. Baines, H. W. Lightfoot, S. Evans, A. Neely,
R. Greenough, J. Peppard, R. Roy, E. Shehab,
A. Braganza, A. Tiwari, J. R. Alcock, J. P. Angus,
M. Basti, A. Cousens, P. Irving, M. Johnson, J. Kingston,
H. Lockett, V. Martinez, P. Michele, D. Tranfield,
I. M. Walton, and H. Wilson, “State-of-the-art in
product-service systems,” Proceedings of the Institution
of Mechanical Engineers, Part B: Journal of Engineering
Manufacture, vol. 221, no. 10, pp. 1543–1552, 2007.
[35] A. Rytter, Vibrational Based Inspection of Civil
Engineering Structures. PhD thesis, Aalborg University,
Denmark, 1993.
This research hasn't been cited in any other publications.
  • Chapter
    Optical imaging and sensing modalities have been used in medical diagnosis for many years. An obvious example is endoscopy, which allows remote wide-field imaging of internal tissues using optical fibers and/or miniature charge-coupled device (CCD) cameras. While techniques such as endoscopy provide useful tools for clinicians, they do not typically allow a complete diagnosis to be made. Instead, physical biopsies may be required to confirm or refute the presence of disease. Furthermore, endoscopic procedures are both invasive and time-consuming. As such, much research is currently directed toward the development of devices that can provide a complete in vivo diagnosis without the requirement for a physical biopsy. Ideally, such devices should also be minimally or non-invasive, and they should provide immediate identification of disease at the point of care. Additionally, there is significant interest in the development of implantable diagnostic devices that can be left within patients’ bodies for extended periods of time (for several days or longer). Such systems could be used for automated disease diagnosis, and example applications include the detection of post-surgical infections as well as monitoring of the health status of patients undergoing chemotherapy. This chapter focuses on the development of optical instruments that can provide in situ diagnosis at the point of care, with an emphasis on progress towards miniature devices that may function as implants in the future.
  • Chapter
    In the quest for ever-reducing system size and increased integration and functionality, application-specific integrated circuit (ASIC) technology plays a pivotal role in modern implants, where custom circuits designed at transistor and device levels are replacing off-the-shelf commercial chips and bulky benchtop systems. Recently, commercial system-on-chip (SoC) devices encompassing digital microcontrollers, radio, and analog–digital converters, as well as reconfigurable amplifier circuits, are widely available. Despite this, further development of ASIC-specific implantable systems is required, particularly in the area of multi-channel array sensor interfaces, ultra-low power data acquisition, and circuits that work with specialized micro-sensors for implants. ASICs designed to focus on a particular application have given designers the freedom to optimize power consumption for a set task, unlike general-purpose SoCs that have to cater for a wide range of applications and hence typically consume more power. In this chapter, we begin with a survey on the latest development of ASICs and related integrated systems from literature. This is followed by an overview of technological trends in integrated circuit/sensor fabrication and fusion. The rest of the chapter focuses on a number of engineering aspects related to ultra-low power ASIC circuits appropriate for implantable sensors and sensor front-ends, covering bioimpedance, neural and electrochemical sensor measurement circuits, as well as low-power analog-to-digital converter design and architectures.
  • Chapter
    In addition to the electrochemical sensors discussed in Chap. 2, a range of other sensing modalities are also important for biomedical and implantable applications. The frequency-dependent electrical properties of tissues are essential for assessing various physiological parameters. This, for example, can be quantified via electrical bioimpedance measurements, which can be combined and corroborated with electrochemical sensors. The human body is a dynamic system in constant motion; therefore, sensors for the measurement of physical properties such as strain and pressure are also important. Sensors for these applications rely on the measurement of resistance, capacitance, and sometimes inductance, and these will also be discussed in this chapter for completeness. Temperature is an important health marker for various applications, and consequently the current state of the art in temperature sensors is also discussed, in terms of both monolithic integration and discrete sensor solutions. Monitoring of the electrical response of the nervous system and the delivery of stimuli represent an important family of applications for neuroscience research and neuroprosthetic devices. These will be addressed in this chapter, along with various application scenarios. Other aspects to be discussed include sensor metrics, such as sensitivity, limit of detection, stability, linear range, selectivity, and specificity.
  • Article
    Internet of Things (IoT) solutions in industrial environments can lead nowadays to the development of innovative and efficient systems aiming at increasing operational efficiency in a new generation of smart factories. In this direction the article presents in detail an advanced Industrial IoT (IIoT) solution, the NGS-PlantOne system, specially designed to enable a pervasive monitoring of industrial machinery through battery-powered IoT sensing devices, thus allowing the development of advanced predictive maintenance applications in the considered scenario. To the end of evaluating the performance of the developed IIoT system in a real environment, the NGS-PlantOne solution has been first installed and then set in operation in a real electricity power plant. The deployed testbed, based on 33 IoT sensing devices performing advanced temperature and vibration monitoring tasks, has been kept in operation for two months while evaluating transmission delays and system operating life through power consumption measures. Performance results show as the developed IIoT solution benefits from all the advantages provided by the adopted IoT protocols, guaranteeing that each node is reachable through simple IP-based techniques with an acceptable delay, and showing an estimated average life of 1 year in case of each IoT smart device is configured to send collected and elaborated data every 30 minutes.
  • Article
    Full-text available
    The metal cutting process initiates with the occurrence of plastic deformation of workmaterial and is followed by tear and removal of material from the workpiece. This process ultimately damages cutting tool and causes tool wear. An acoustic emission (AE) sensor has been employed to measure the signal frequency in machining. The AE signal component of tool wear and plastic deformation in turning are separated by simulating the process of tool wear by a grinding test where the workpiece of grinding test is the same tool-insert for tuning test, and the process of tool wear in turning is replicated by the process of material removal in grinding. The frequency of tool wear for this particular investigation is found to lie between 67 kHz and 471 kHz whereas for plastic deformation of workmaterial, it has a fluctuation within the range starting from 51 kHz to some value within 471 kHz.
  • Article
    Full-text available
    In recent years, the range of sensing technologies has expanded rapidly, whereas sensor devices have become cheaper. This has led to a rapid expansion in condition monitoring of systems, structures, vehicles, and machinery using sensors. Key factors are the recent advances in networking technologies such as wireless communication and mobile ad hoc networking coupled with the technology to integrate devices. Wireless sensor networks (WSNs) can be used for monitoring the railway infrastructure such as bridges, rail tracks, track beds, and track equipment along with vehicle health monitoring such as chassis, bogies, wheels, and wagons. Condition monitoring reduces human inspection requirements through automated monitoring, reduces maintenance through detecting faults before they escalate, and improves safety and reliability. This is vital for the development, upgrading, and expansion of railway networks. This paper surveys these wireless sensors network technology for monitoring in the railway industry for analyzing systems, structures, vehicles, and machinery. This paper focuses on practical engineering solutions, principally, which sensor devices are used and what they are used for; and the identification of sensor configurations and network topologies. It identifies their respective motivations and distinguishes their advantages and disadvantages in a comparative review