ArticlePDF Available

Abstract and Figures

Emergence of uniquely addressable embeddable devices has raised the bar on Telematics capabilities. Though the technology itself is not new, its application has been quite limited until now. Sensor based telematics technologies generate volumes of data that are orders of magnitude larger than what operators have dealt with previously. Real-time big data computation capabilities have opened the flood gates for creating new predictive analytics capabilities into an otherwise simple data log systems, enabling real-time control and monitoring to take preventive action in case of any anomalies. Condition-based-maintenance, usage-based-insurance, smart metering and demand-based load generation etc. are some of the predictive analytics use cases for Telematics. This paper presents the approach of condition-based maintenance using real-time sensor monitoring, Telematics and predictive data analytics.
Content may be subject to copyright.
International Journal of Mobile Network Communications & Telematics ( IJMNCT) Vol. 3, No.3, June 2013
DOI : 10.5121/ijmnct.2013.3303 19
Gopalakrishna Palem
Symphony-Teleca Corporation, Bangalore, India
Emergence of uniquely addressable embeddable devices has raised the bar on Telematics capabilities.
Though the technology itself is not new, its application has been quite limited until now. Sensor based
telematics technologies generate volumes of data that are orders of magnitude larger than what operators
have dealt with previously. Real-time big data computation capabilities have opened the flood gates for
creating new predictive analytics capabilities into an otherwise simple data log systems, enabling real-time
control and monitoring to take preventive action in case of any anomalies. Condition-based-maintenance,
usage-based-insurance, smart metering and demand-based load generation etc. are some of the predictive
analytics use cases for Telematics. This paper presents the approach of condition-based maintenance using
real-time sensor monitoring, Telematics and predictive data analytics.
Telematics, Preventive Maintenance, Predictive Maintenance, Sensor arrays
Maintenance, considered often as a non-value add function, is always under a constant pressure
from the top management to contribute more for costs reduction, keep the machines in excellent
working condition, all the while satisfying the stringent safety and operational requirements.
Towards this end manufacturers and operators usually employ various maintenance strategies, all
of which and can be broadly categorized as below:
Corrective Maintenance
Preventive Maintenance
Predictive Maintenance
Corrective maintenance is the classic Run-to-Failure reactive maintenance that has no special
maintenance plan in place. The machine is assumed to be fit unless proven otherwise.
High risk of collateral damage and secondary failure
High production downtime
Overtime labour and high cost of spare parts
Machines are not over-maintained
No overhead of condition monitoring or planning costs
Preventive maintenance (PM) is the popular periodic maintenance strategy that is actively
employed by all manufacturers and operators in the industry today. An optimal breakdown
International Journal of Mobile Network Communications & Telematics ( IJMNCT) Vol. 3, No.3, June 2013
window is pre-calculated (at the time of component design or installation, based on a wide range
of models describing the degradation process of equipment, cost structure and admissible
maintenance etc.), and a preventive maintenance schedule is laid out. Maintenance is carried-out
on those periodic intervals, assuming that the machine is going to break otherwise.
Calendar-based maintenance: Machines are repaired when there are no faults
There will still be unscheduled breakdowns
Fewer catastrophic failures and lesser collateral damage
Greater control over spare-parts and inventory
Maintenance is performed in controlled manner, with a rough estimate of costs well-
known ahead of time
Predictive Maintenance, also known as Condition-based maintenance (CBM) is an emerging
alternative to the above two that employs predictive analytics over real-time data collected
(streamed) from parts of the machine to a centralized processor that detects variations in the
functional parameters and detects anomalies that can potentially lead to breakdowns. The real-
time nature of the analytics helps identify the functional breakdowns long before they happen but
soon after their potential cause arises.
Unexpected breakdown is reduced or even completely eliminated
Parts are ordered when needed and maintenance performed when convenient
Equipment life is maximized
High investment costs
Additional skills might be required
Condition-based maintenance differs from schedule-based maintenance by basing maintenance
need on the actual condition of the machine rather than on some pre-set schedule.
For example, a typical schedule-based maintenance strategy demands automobile operators to
change the engine oil, say after every 3,000 to 5,000 Miles travelled. No concern is given to the
actual condition of vehicle or performance capability of the oil.
If on the other hand, the operator has some way of knowing or somehow measuring the actual
condition of the vehicle and the oil lubrication properties, he/she gains the potential to extend the
vehicle usage and postpone oil change until the vehicle has travelled 10,000 Miles, or perhaps
pre-pone the oil change in case of any abnormality.
In the following sections we look into what constitutes a condition-based maintenance solution,
the steps involved in implementing one and an overall solution methodology.
International Journal of Mobile Network Communications & Telematics ( IJMNCT) Vol. 3, No.3, June 2013
Predictive analytics in combination with sensor based telematics provides deep insights into the
machine operations and full functionality status giving rise to optimal maintenance schedules
with improved machine availability.
Underlying schedule-based maintenance is the popular belief that machine failures are directly
related to machine operating age, which studies indicate not to be true always. Failures are not
always linear in nature. Studies indicate that 89% of the problems are random with no direct
relation to the age [2]. Table 1 showcases some of these well-known failure patterns and their
conditional probability (Y-axis) with respect to Time (X-axis).
Table 1. Failure Conditional Probability Curves
Age-Related = 11%
Random = 89%
Wear out
Type A = 2%
Infant Mortality
Type B = 68%
Type E = 4%
Initial Break-in Period
Type C = 7%
Type F = 5%
Relatively Constant
Type D = 14%
Complex items frequently demonstrate some infant mortality, after which their failure probability
either increases gradually or remains constant, and a marked wear-out age is not common.
Considering this fact, the chance of a schedule-based maintenance avoiding a potential failure is
low, as there is every possibility that the system can fail right after a scheduled maintenance.
Thus, preventive maintenance imposes additional costs of repair. Condition-based maintenance
reduces such additional costs by scheduling maintenance if and only when a potential breakdown
symptom is identified.
Preventive Maintenance
Predictive Maintenance
Figure 1. Condition-based maintenance schedules are flexible and cost-saving
International Journal of Mobile Network Communications & Telematics ( IJMNCT) Vol. 3, No.3, June 2013
However, the costs of monitoring equipment and monitoring operations should not exceed the
original asset replacement costs; lest the whole point of condition-based maintenance becomes
moot. Internal studies conducted with our customers have estimated that a properly functioning
CBM program can provide savings of 8% to 12% over the traditional maintenance schemes. They
indicated the following industrial average savings resultant from initiation of a functional
predictive maintenance program:
Reduction in maintenance costs: 25% to 30%
Elimination of breakdowns: 70% to 75%
Reduction in equipment or process downtime: 35% to 45%
Increase in production: 20% to 25%
Apart from the above, improved worker and environment safety, increased component
availability, better product quality etc. are making more and more manufacturers and operators
embrace CBM based management solutions.
A Condition-based maintenance management (CBMM) solution is enabled by three major
technology enhancements over a traditional maintenance solution:
1. Remote Sensor Monitoring & Data Capturing
2. Real-time Stream Processing of Sensor Data
3. Predictive Analytics
CBMM solutions essentially operate by having sensors attached to remote assets (mobile or
stationary) that send continuous streams of data about the assets’ operational conditions to a
monitoring station that then analyses them in real-time using predictive analytic models and
detects any problems in the behaviour or state of the asset. Once a problem is detected,
appropriate pre-configured action is taken to notify the operator or manufacture for corrective
action. The monitoring station in question can be on the same network as that of the sensors or it
could be in a remote location far away from them, connected through wide area networks or
satellite networks.
Figure 2. Condition-based maintenance using sensor arrays and telematics
International Journal of Mobile Network Communications & Telematics ( IJMNCT) Vol. 3, No.3, June 2013
Nature of the sensors being monitored, frequency of the data getting collected and precision of the
analytic models being used all affect the quality of the prediction results. Thus, it is imperative
that manufacturers and operators define all these parameters with utmost care while deploying a
CBMM system. This, however, entails a thorough understanding of the system under operation
and expects a clear-cut answer as to what is being monitored and what is expected out of such
monitoring. Some of the questions that can help manufacturers and operators along those lines
For monitoring:
Which parts of the system or asset are expected to be monitored?
What type of data is expected to be collected and which type of sensors give such
data? For example, visual data, thermal data etc.
What is the expected frequency for the data collection?
How should any failures in the sensors be handled?
For real-time stream collection and processing at the monitoring station:
What is the acceptable data processing latency?
How to deal with imperfections in the received data? For example, a faulty sensor
sending incorrect data
What should be done with the collected data after processing?
For Analytics sub-system:
Which analysis technique accurately models the asset/system behaviour?
What is the definition of acceptable behaviour and anomaly?
What should be the response in case of any anomaly detection?
What should be the reasonable timeframe between anomaly detection and corrective
How to deal with situations where there are multiple anomalies detected at the same
Generic and security related questions:
Who should be allowed to access the collected data and analysis results?
What is the change management process required in case one wants to tune the
tracking and analysis parameters?
The following section briefly summarizes some of the standard methods we use in the CBMM
systems we deploy for our customers and can help in answering above questions.
The primary component in a condition-based maintenance management solution is a sensor array
and the measurements it provide. Some of the widely used measurement techniques in the
industry are:
Temperature Measurement: Thermal indicators, such as temperature-sensitive paint,
thermography etc., help detect potential failures arising out of temperature changes in the
equipment. Excessive mechanical friction, degraded heat transfer, poor electrical
connections are some of the problems that can be detected with this type of measurement.
A thermocouple or RTD
Can be used on all accessible
IR radiation measured from a
surface, often with laser sight
Good for walk around temperature
checks on machines and panels
International Journal of Mobile Network Communications & Telematics ( IJMNCT) Vol. 3, No.3, June 2013
Chemical indicators calibrated
to change colors at specified
Works great for inspection rounds
Handheld still or video camera
sensitive to emitted IR
Best for remote monitoring.
Requires good training
Dynamic Monitoring: Spectrum analysis, shock pulse analysis are some of the dynamic
monitoring methods that measure and analyse energy emitted from mechanical equipment
in the form of waves, vibration, pulses and acoustic effects. Wear and tear, imbalance,
misalignment and internal surface damage are some of the problems that can be detected
with this type of measurement.
ISO Filtered
2Hz-1kHz filtered velocity
A general condition indicator
Carpet and Peak related to
demodulation of sensor
resonance around 30kHz
Single value bearing indicator
Acoustic Emission
Distress & dB, demodulates
a 100kHz carrier sensitive to
stress waves
Better indicator than ISO
velocity, without the ISO
comfort zone
Vibration Meters
Combine velocity, bearing
and acceleration techniques
ISO Velocity, envelope and
high frequency acceleration
give best performance
4-20mA sensors
Filter data converted to
DCS/PLC compatible signal
Useful to home-in on specific
problems by special order
Fluid Analysis: Ferrography, particle counter testing are some of the fluid analysis
methods performed on different types of oils, such as lubrication, hydraulic, insulation oil
etc., to identify any potential problems of wear and tear in the machines. Machine
degradation, oil contamination, improper oil consistency, oil deterioration are some of the
problems that can be detected with this method. The main areas of analysis in this are:
Fluid physical properties: Viscosity, appearance
Fluid chemical properties: TBN, TAN, additives, contamination, % water
Fluid contamination: ISO cleanliness, Ferrography, Spectroscopy, dissolved
Machine health: Wear metals associated with plant components
Corrosion Monitoring: Methods such as Coupon testing, corrometer testing help identify
the extent of corrosion, corrosion rate and state (active/passive corrosion) for the
materials used in the asset.
Non-destructive Testing: Involves using non-destructive methods, such as X-Rays,
ultrasonic etc., to detect any potential anomalies arising internal to the asset structure.
Most of these tests can be performed while the asset is online and being used.
Electrical testing and Monitoring: High potential testing, power signal analysis are some
of the prominent electrical condition monitoring mechanisms that try to identify any
changes in the system properties, such as resistance, conductivity, dielectric strength and
potential. Electrical insulation deterioration, broken motor rotor bars and shorted motor
International Journal of Mobile Network Communications & Telematics ( IJMNCT) Vol. 3, No.3, June 2013
stator lamination etc. are some of the problems that can be detected with this type of
Observation and Surveillance: Visual, audio and touch inspection criteria are some of the
surveillance condition monitoring techniques based on the human sensory capabilities.
They act as supplement to other condition-monitoring techniques and help detect
problems such as loose/worn parts, leaking equipment, poor electrical and pipe
connections, stream leaks, pressure relief valve leaks and surface roughness changes etc.
Once the appropriate measurement mechanisms are in place, the next step is the event definition
phase: to define what constitutes acceptable system behaviour and what is to be considered as
anomaly. It is useless to put costly monitoring equipment in place, without knowing what to
expect out of it. Expert opinion and judgment (such as manufacturer’s recommendations),
published information (such as case studies), historical data etc. are some of the good sources that
can help in this task. The definition of anomaly should be unambiguous and easy to detect. If the
cost of anomaly detection far exceeds the costs of consequences of that anomaly, then it is not a
valid scenario for implementing CBMM system.
The next step that follows the event definition phase is, determining event inspection frequency.
Frequency of any of form of condition-based-maintenance is based on the fact that most failures
do not occur instantaneously, and that it is often possible to detect them during their final stages
of deterioration. If evidence can be found that something is in the final stages of failure, it is
possible to take action for preventing it from failing completely and/or avoid the consequences.
Figure 1. Measurement Frequencies can be determined with P-F Interval
The failure behaviour typically exhibited by majority of the systems in operation is as showcased
in Figure 3. During operation, over a period of time, the systems enter a phase of potential failure
(P), and start displaying few early signs of wear & tear and other stressful behaviours that if
neglected finally lead to full functional failure (F). For most of the systems the time interval
between the potential failure point (P) and full function failure (F) is large enough to allow
detection and prevention of the failure.
This time gap between P and F is what is popularly known as the P-F Interval in the literature,
and any cost-effective maintenance strategy should try to maximize on it. The P-F Interval could
be in hours or days, or even weeks or months, based on the complexity of the system and the unit
of measurement - for it is not uncommon to see the P-F Interval being measured in non-time
International Journal of Mobile Network Communications & Telematics ( IJMNCT) Vol. 3, No.3, June 2013
units, such as stop-start cycles or units of output etc. Based on the failure mode and the unit of
measurement, the P-F Interval can end up varying from fractions of a second to several decades
(on temporal scale).
Be whatever the unit of measurement and the P-F interval, a successful CBMM system should be
capable of detecting the early signals after P and respond to them long before F. The response
action typically consists of multiple steps (as laid out below) and should all be accompanied
within the P-F interval.
1. Analysing the root-cause based on detected early signals
2. Planning corrective action based on the analysed root-cause
3. Organizing the resources to implement the laid out plan
4. Actual implementation of the corrective action plan
The amount of time needed for these response actions usually vary, from a matter of hours (e.g.
until the end of operating cycle or end of shift), minutes (e.g. to clear people from a failing
building), to weeks or even months (e.g. until a major shutdown).
Thus, it is a common practice to use the inspection interval to be half the P-F interval. This will
ensure that there is at-least half the P-F interval remaining after the potential-failure detection for
corrective action plan. However, it should be noted that most of the times earlier the corrective
action plans are implemented, lower the cost in which cases, some other smaller fraction of P-F
interval can be used as the inspection interval, so that potential problems can be detected as early
as possible and rectified.
However, an important point to be remembered is P-F interval is not an easy metric to be
computed. It varies from asset type to asset type, environment to environment and even from one
asset to another with in the same asset type (based on its previous fault history and working
conditions). Understanding the failure patterns, and identifying the class of pattern to which the
asset belongs, its past fault history, manufacturer’s recommendations, operating conditions, expert
judgment etc. are some of the sources that can help in arriving at an accurate P-F interval for any
given asset/system.
At each inspection interval, the CBMM system collects data from sensors and uses one of the
following methods to determine the condition of the asset being monitored:
Trend Analysis: Reviews the data to find if the asset being monitored is on an obvious
and immediate downward slide toward failure. Typically a minimum of three monitoring
points are recommended for arriving at a trend accurately as a reliable measure to find if
the condition is deprecating linearly.
Pattern recognition: Decodes the causal relations between certain type of events and
machine failures. For example, after being used for a certain product run, one of the
components used in the asset fails due to stresses that are unique to that run
Critical range and limits: Tests to verify if the data is within a critical range limit (set by
professional intuition)
Statistical process analysis: Existing failure record data (retrieved from warranty claims,
data archives and case-study histories) is driven through analytical procedures to find an
accurate model for the failure curves and the new data is compared against those models
to identify any potential failures.
International Journal of Mobile Network Communications & Telematics ( IJMNCT) Vol. 3, No.3, June 2013
Based on the failure mode and asset class the right method for the prediction can vary. For
example, assets that fall into type E class (bathtub pattern) usually benefit from Weibull
distribution, while split system approach is used for complex systems with multiple sub-systems.
Stream processing the arriving data can help build the trend analysis and critical range limits, but
to accurately process pattern recognition and statistical model building methods, past history data
is as important as the new arriving data. Thus, typically CBM management systems should keep
record of old data for some reasonable amount of time before they are archived or destroyed. This
time period varies from domain to domain and may even be regulated by local country laws. For
example, financial fraud records may need to be kept active for longer time, in the range of 7 to
15 years per se, while flight records generated from airplane internal sensors are typically
discarded after the journey completion (primarily due to it being voluminous, though this trend
could soon change as the big data warehousing gets more prominent).
Another reason the old data streams become important is to identify any potential outliers in the
streamed-in data from the sensors. While monitoring for the faults in the assets, it is possible that
the sensor that is taking the readings, being a machine itself, could fail and start sending faulty
records. Intelligent CBM management systems capable of detecting such outliers will try to
isolate these faulty sensors and notify the appropriate personal for corrective action, or substitute
it with proper estimated data based on previous records. In either case, human inspection is as
much necessary as a completely automated monitoring system for automation only
complements the human surveillance efforts, not replace them. Thus, many automated monitoring
systems provide a way for manual override for configurable parts of their functionality.
Figure 2. Condition-based maintenance management system architecture
Typically a CBM management solution will have an administrative console that lets the operators
define and update various parameters, such as critical limits, response notifications, default
corrective actions etc. Advanced management systems also allow the administrators to access the
monitoring solution functionality remotely through web UI and mobile UI, capable of sending
periodic digests weekly or monthly for pre-configured stake-holders reporting the status of the
asset being monitored on a regular basis. With architectures like the one shown in Figure 4 and
common interface standards such as IEEE 1451, IEEE 1232, MIMOSA and OSA-CBM,
International Journal of Mobile Network Communications & Telematics ( IJMNCT) Vol. 3, No.3, June 2013
advanced management systems integrations become possible among disparate software and
hardware components from different vendors, all working hard together just for one single
purpose to provide the operators maximum usage out of their assets.
Condition-based maintenance is based on the principle of using real-time data to prioritize and
optimize maintenance resources. Such a system will determine the equipment's health, and act
only when maintenance is actually necessary. Telematics development (such as IPV6, 3G and 4G
LTE) in the recent times in combination with big-data real-time stream analytics is opening new
opportunities for manufacturers and asset owners to save costs and optimize resource usage in
innovative ways. Condition-based maintenance management systems built around real-time
sensor monitoring and telematics technologies offer flexibility and cost-savings in terms of
providing greater control over when to perform the maintenance, which parts to pre-order and
how the optimally schedule the labour.
[1] Weibull, W, (1951) "A statistical distribution function of wide applicability", Journal of Applied
Mechanics-Trans. ASME, Vol. 18, No. 3: pp. 293297.
[2] Nowlan, F. Stanley, and Howard F. Heap, (1978) Reliability-Centred Maintenance. Department of
Defense, Report Number AD-A066579 Washington, D.C. 1978.
[3] Pérez, Angel Torres; Hadfield, Mark. (2011). "Low-Cost Oil Quality Sensor Based on Changes in
Complex Permittivity." Sensors 11, no. 11: 10675-10690.
[4] J. CUENA, M. MOLINA, (2000) The role of knowledge modelling techniques in software
development: a general approach based on a knowledge management tool, International Journal
of Human-Computer Studies, Vol. 52, No. 3, pp. 385-421
[5] G. Abdul-Nour, H. Beaudoin, P. Ouellet, R. Rochette, S. Lambert, (2000) A reliability based
maintenance policy; a case study, Computers & Industrial Engineering, Vol. 35, No. 3-4, pp. 591-
[6] Hansen, T., Dirckinck-Holmfeld, L., Lewis, R., & Rugelj, J. (1999). Using telematics to support
collaborative knowledge construction. Collaborative learning: Cognitive and computational
approaches, 169-196.
Gopalakrishna Palem is a Corporate Technology Strategist specialized in Distributed Computing
technologies and Cloud operations. During his 12+ year tenure at Microsoft and Oracle, he helped many
customers build their high volume transactional systems, distributed render pipelines, advanced
visualization & modelling tools, real-time dataflow dependency-graph architectures, and Single-sign-on
implementations for M2M telematics. When he is not busy working, he is actively engaged in driving
open-source efforts and guiding researchers on Algorithmic Information Theory, Systems Control and
Automata, Poincare recurrences for finite-state machines, Knowledge modelling in data-dependent
systems and Natural Language Processing (NLP).
... Maintenance consists in ensuring that the building keeps fulfilling the designed function for as long as possible (Ilter & Ergen, 2015). There are three types of maintenance (Rabatel et al., 2011, Motawa & Almarshad, 2013, Palem, 2013, Bilal et al., 2016, Shalabi & Turkan, 2017. The first type is preventive or planned maintenance, usually instructed by the manufacturer. ...
... The third type is predictive maintenance: detecting anomalies with the help of sensors to intervene before the failure happens. Palem (2013) states that both preventive and corrective maintenance are inefficient. The first one because it is costly and might happen when the system is still functioning, and the second one because it happens after the system has failed, therefore extra time is needed to fix the issue. ...
... Palem (2013) recommends using predictive maintenance as it reduces both the costs and the time spent. In order to apply predictive maintenance, remote sensor monitoring and real-time processing are fundamental (Palem, 2013). ...
Full-text available
Purpose At the end of a building’s lifecycle, there are several limitations to the decision-making process (DMP). There is a lack of data available from the building’s history, the difficulty in assessing the condition of a building and the variety of stakeholders’ needs that have to be satisfied. The purpose of this paper is to answer the question: how would end-of-life (EOL) DMP change when buildings will have been digitally built? The answer will be illustrated through a conceptual framework. Design/methodology/approach A qualitative analysis of the existing literature has been performed to identify the elements within building information modelling (BIM) and advanced digital technologies that could be of support to the DMP. The findings have been collected and summarised in a conceptual framework that has been validated and enhanced through online interviews with industry experts. Findings The enhanced framework has identified that BIM technology would bring the benefit of providing the initial digital data source, from which machine learning and data analytics would then extract the relevant data needed to measure accurately the criteria during the analysis of the EOL options put on the table. Originality/value The findings of this research could contribute to developing the software modules making the bridge between BIM and machine learning technologies, to implement them in the EOL DMP.
... The former strategy usually induces additional costs [2] and an important unavailability of the system [3]. Besides, a failure can raise safety issues, as the systems can no longer perform all or part of the functions it was designed for. ...
... In order to overcome these drawbacks, a new form of maintenance has emerged: predictive maintenance [3]. It consists of a regular surveillance of the organs of the system [5], which allow evaluating their health state and their proper functioning. ...
... -Corrective maintenance involves waiting for a failure before repairing or replacing the affected components -Preventive maintenance involves replacing the components at regular intervals, according to a fixed planning The former strategy usually induces additional costs (Lee, Ni, Djurdjanovic, Qiu, & Liao, 2006) and an important unavailability of the system (Palem, 2013). Besides, a failure can raise safety issues as the systems can no longer perform all or part of the functions it was designed for. ...
... Therefore, some components are replaced whereas they are still in a nominal operating state or slightly degraded (Lee, Ni, Djurdjanovic, Qiu, & Liao, 2006) (Le, 2016). In order to overcome these drawbacks, a new form of maintenance has emerged: predictive maintenance (Palem, 2013). It consists of a regular surveillance of the organs of the system (Bartelds, et al., 2004), which allow evaluating their health state and their proper functioning. ...
... Even if predictive maintenance provides more advantages than other types of maintenance (such as corrective maintenance or preventive maintenance) [11,[20][21][22], it remains a challenging task for industry to find the optimal time for predictive maintenance. The existing literature provides a large panel of methods to optimize the predictive maintenance strategy [23]. ...
Full-text available
Predictive maintenance can be efficiently improved by studying the sensitivity of the maintenance decisions with respect to changes in the proposed model parameters (costs, duration of reparation, etc.). To address this issue, we first propose an original approach that includes both maintenance costs and maintenance risks in the same objective function to minimize. This approach uses the RUL as an indicator of the health state of the system and supposes that the system is under regular inspections and can only be replaced by a new system in case of serious deterioration or failure. Then, we present a process of human decision making under uncertainty based on several criteria. Finally, we study and analyze the influence of the model parameters and their implications on the obtained maintenance policies. The study will lead to some recommendations that can improve the predictive maintenance decisions and help experts better handle maintenance costs.
... La deuxième, appelée maintenance curative, a pour objectif de restaurer toutes les fonctions requises du système. Ces types de maintenance engendrent généralement des surcoûts [LND + 06, LBB + 15] et une indisponibilité [Pal13]. [Bay13] sur la détection d'état dégradé pour les systèmes équipés de HUMS de Thales, on suppose que les composants du pod ont 3 états (stable, dégradé, panne) parfaitement observables. ...
On présente un problème d'optimisation pour la maintenance d'un système multi-composants conçu par Thales. Ce dernier est sujet à des détériorations et défaillances aléatoires de ses composants, au cours de missions pour lesquelles il est requis, entraînant l'évolution de son état et une pénalité d'indisponibilité en cas d’échec. L’enjeu est alors de définir une politique optimale d'emploi et de maintenance du système, afin de garantir le bon déroulement des missions, tout en minimisant ses coûts de gestion. Il s’agit de déterminer un compromis entre intervenir trop tôt, générant des coûts de maintenances inutiles, et intervenir trop tard, menant à la panne du système et à payer des pénalités et des opérations plus coûteuses.Une des spécificités de ce travail est de considérer une prise de décision séquentielle sur le système. Ensuite, il s’agit de différencier les opérations de maintenance selon l’état de chacun de ses composants. L'idée principale de ce travail est alors de proposer un modèle mathématique pour l'évolution du système via le formalisme d'un Processus Markovien Décisionnel (MDP). Ainsi, l’objectif est de résoudre le problème d’optimisation associé, à savoir déterminer pour chaque date de décision et chaque état du système, l’action qui minimise la somme des coûts générés sur tout l’horizon. C’est ce qu’on appelle une politique. On propose ensuite plusieurs politiques de référence préventives et correctives et on compare leurs performances en termes de coûts et de statistiques de pannes, par simulations de Monte-Carlo. Ceci illustre l'intérêt de regrouper les maintenances lors des passages en atelier, et de considérer les temps de fonctionnement des composants pour les prises de décision, afin de réduire à la fois les coûts et le taux de pannes.Le modèle spécifié pour cette problématique industrielle donne lieu à un problème d’optimisation non standard dans le cadre des MDP, car l'espace d’états du système est continu et le noyau de transition n'est pas explicite analytiquement, mais seulement simulable. Afin de pouvoir rechercher une politique optimale sur un ensemble fini de politiques admissibles, on discrétise la règle de décision du MDP. Ceci permet de prendre des décisions sur un nombre fini d’états, sans discrétiser la dynamique du MDP. Les coûts des politiques de référence sont utilisés pour calibrer cette discrétisation. Il s’agit de déterminer un compromis entre précision, conduisant à considérer un très grand nombre d’états, et une complexité numérique, conduisant à un nombre d'états le plus petit possible. Enfin, le noyau de transition n’étant toujours pas explicite, on implémente et on compare deux méthodes d'optimisation stochastiques basées sur les simulations afin d'approcher et d’expliciter une politique optimale. Une attention particulière est portée à l’identification de l’origine des gains, afin d’interpréter la politique optimale déterminée.
... While corrective maintenance may imply high costs and significant system downtime [2][3] [4], preventive maintenance, in the other hand, does not allow an optimal Manuscript received June 21, 2019; revised April 1, 2020. exploitation of the system [5]. ...
... Additionally, given that the FM are coming under pressure to keep the machines in excellent working condition while satisfying the stringent safety and operational requirements [45], BDA can be used for the development of various kinds of predictive and financially optimized maintenance strategies, suitable to streamline the labor intensive maintenance functions [15]. Applying BDA in maintenance provides substantial efficient management capabilities, as the development prognostic models can help avoid complete failures in critical equipment, such as chillers, boilers, fans, and pumps, reduce maintenance costs, and avoid disruptions in building services [42]. ...
Full-text available
The recent advances in Internet of Things (IoT), computational analytics, processing power, and assimilation of Big Data (BD) are playing an important role in revolutionizing maintenance and operations regimes within the wider facilities management (FM) sector. The BD offers the potential for the FM to obtain valuable insights from a large amount of heterogeneous data collected through various sources and IoT allows for the integration of sensors. The aim of this article is to extend the exploratory studies conducted on Big Data analytics (BDA) implementation and empirically test and categorize the associated drivers and challenges. Using exploratory factor analysis (EFA), the researchers aim to bridge the current knowledge gap and highlight the principal factors affecting the BDA implementation. Questionnaires detailing 26 variables are sent to the FM organization in the U.K. who are in the process or have already implemented BDA initiatives within their FM operations. Fifty-two valid responses are analyzed by conducting EFA. The findings suggest that driven by market competition and ambitious sustainability goals, the industry is moving to holistically integrate analytics into its decision making. However, data quality, technological barriers, inadequate preparedness, data management, and governance issues and skill gaps are posing to be significant barriers to the fulfillment of expected opportunities. The findings of this study have important implications for FM businesses that are evaluating the potential of the BDA and IoT applications for their operations. Most importantly, it addresses the role of the BD maturity in FM organizations and its implications for perception of drivers.
... Hardware for condition monitoring of a blowout preventer require sensors attached to it which collect and send data about its operational condition and performance to a monitoring station. The monitoring station performs real-time analysis of the data collected and detects if there are any problems with the performance of the BOP [33]. The monitoring station can be part of the same network as the sensors, an example of which is a rig-based engineering work station (EWS) described by Mckay et al. [25] or it can be in a remote location a great distance away from them. ...
Conference Paper
With the steadily growing demand for energy in the world, oil and gas companies are finding themselves facing increasing capital and operating costs. To ensure the economic viability of investments and improve the safety of operations, oil and gas companies are promoting their asset integrity management (AIM) systems. In the past, the oil and gas industry adopted reactive maintenance regimes, which involved recertification, testing and repair of faulty equipment while trying to achieve minimum downtime. As technology becomes more affordable, operators have been able to carry out improved fault diagnosis, prognosis and maintenance optimisation. As a result of this, condition-based maintenance (CBM) is being adopted more and more as the preeminent maintenance regime for oil and gas equipment. The blowout preventer (BOP) is one of the most expensive and safety critical drilling equipment in the oil and gas industry. However, there have been very few studies and best practices about how to develop a CBM policy and what specific monitoring techniques and devices will be required to implement it for the BOP system. This paper proposes a V-model based architecture for designing a CBM policy in BOP systems. As a result of the model proposed, gaps in implementation are identified and all the hardware, software and training requirements for implementing the CBM solution in BOP systems will be outlined in detail. Our proposed CBM framework will help BOP operators and maintenance personnel make cost savings through less repairs and replacements and minimal downtime.
La mise en place d’une politique de maintenance prévisionnelle est un défi majeur dans l’industrie qui tente de réduire le plus possible les frais relatifs à la maintenance. En effet, les systèmes sont de plus en plus complexes et demandent un suivi de plus en plus poussé afin de rester opérationnels et sécurisés. Une maintenance prévisionnelle nécessite d’une part d’évaluer l’état de dégradation des composants du système, et d’autre part de pronostiquer l’apparition future d’une panne. Plus précisément, il s’agit d’estimer le temps restant avant l’arrivée d’une défaillance, aussi appelé Remaining Useful Life ou RUL en anglais. L’estimation d’une RUL constitue un réel enjeu car la pertinence et l’efficacité des actions de maintenance dépendent de la justesse et de la précision des résultats obtenus. Il existe de nombreuses méthodes permettant de réaliser un pronostic de durée de vie résiduelle, chacune avec ses spécificités, ses avantages et ses inconvénients. Les travaux présentés dans ce manuscrit s’intéressent à une méthodologie générale pour estimer la RUL d’un composant. L’objectif est de proposer une méthode applicable à un grand nombre de cas et de situations différentes sans nécessiter de modification majeure. De plus, nous cherchons aussi à traiter plusieurs types d’incertitudes afin d’améliorer la justesse des résultats de pronostic. Au final, la méthodologie développée constitue une aide à la décision pour la planification des opérations de maintenance. La RUL estimée permet de décider de l’instant optimal des interventions nécessaires, et le traitement des incertitudes apporte un niveau de confiance supplémentaire dans les valeurs obtenues.
Full-text available
When engaging in collaborative knowledge construction, geographically distributed groups are found to have additional problems in comparison with groups working face-to-face. Telematic services have been created to address these problems, but so far they have not resolved all the difficulties. One of several reasons may be that too little consideration is given to the match between the tool and the process it is intended to support. Thus, this chapter first describes some functionalities of telematic tools, then offers two typical present-day examples of collaborative knowledge construction by groups that use such tools (one a case of project pedagogy distance learning, another a case of international research collaboration). A promising framework for understanding collaborative and tool-mediated human effort is the dialectically systemic approach of Activity Theory, which originated with Vygotsky's cultural-historical school of psychology in the 1920s and is presently being further developed by Yrjö Engeström, amongst others. Aspects of this theory are outlined and used as a framework to further the understanding of some processes and problems that were reported by the case studies.
Full-text available
The aim of the paper is to discuss the use of knowledge models to formulate general applications. First, the paper presents the recent evolution of the software field where increasing attention is paid to conceptual modelling. Then, the current state of knowledge modelling techniques is described where increased reliability is available through the modern knowledge-acquisition techniques and supporting tools. The knowledge structure manager (KSM) tool is described next. First, the concept of knowledge area is introduced as a building block where methods to perform a collection of tasks are included together with the bodies of knowledge providing the basic methods to perform the basic tasks. Then, the CONCEL language to define vocabularies of domains and the LINK language for methods formulation are introduced. Finally, the object-oriented implementation of a knowledge area is described and a general methodology for application design and maintenance supported by KSM is proposed. To illustrate the concepts and methods, an example of system for intelligent traffic management in a road network is described. This example is followed by a proposal of generalization for reuse of the resulting architecture. Finally, some concluding comments are made regarding the feasibility of using the knowledge modelling tools and methods for general application design.
Full-text available
Real time oil quality monitoring techniques help to protect important industry assets, minimize downtime and reduce maintenance costs. The measurement of a lubricant's complex permittivity is an effective indicator of the oil degradation process and it can be useful in condition based maintenance (CBM) to select the most adequate oil replacement maintenance schedules. A discussion of the working principles of an oil quality sensor based on a marginal oscillator to monitor the losses of the dielectric at high frequencies (>1 MHz) is presented. An electronic design procedure is covered which results in a low cost, effective and ruggedized sensor implementation suitable for use in harsh environments.
Several years ago the Department of Defence in the United States engaged United Airlines to develop detailed documentation to describe and justify airline practices for developing preventative maintenance programmes, such as those exemplified by the Air Transportation Association publication: MSG-2 Airline/Manufacturer Maintenance Programme Planning Document. The result of that effort was a book entitled Reliability-Centred Maintenance, that described a logical discipline for developing preventative maintenance programmes which are matched to specific, identified, inherent reliability characteristics of the equipment that they support.
This book explains basic concepts, principles, definitions, and applications of a logical discipline for development of efficient scheduled (preventive) maintenance programs for complex equipment, and the on-going management of such programs. Such programs are called reliability-centered maintenance (RCM) programs because they are centered on achieving the inherent safety and reliability capabilities of equipment at a minimum cost. A U.S. Department of Defense objective in sponsoring preparation of this document was that it serve as a guide for application to a wide range of different types of military equipment. There are essentially only four types of tasks in a scheduled mainenance program: (1) Inspect an item to detect a potential failure; (2) Rework an item before a maximum permissible age is exceeded; (3) Discard an item before a maximum permissible age is exceeded; (4) Inspect an item to find failures that have already occurred but were not evident to the equipment operating crew. A central problem addressed in this book is how to determine which types of scheduled maintenance tasks, if any, should be applied to an item and how frequently assigned tasks should be accomplished. The use of a decision diagram as an aid in this analysis is illustrated. The net result is a structured, systematic blend of experience, judgment, and operational data/ information to identify and anlayze which type of maintnence task is both applicable and effective for each significant item as it relates to a particular type of equipment.
This paper discusses the applicability of statistics to a wide field of problems. Examples of simple and complex distributions are given.
Reliability centered maintenance (RCM) incorporates sound guidance for managers who wish to attain high standards of maintenance at their operating plants. Since the amount and type of maintenance which is applied depend strongly on the machine or components age (DFR, CFR or IFR), on its replacement cost as well as on the cost and safety consequences of system failure, a careful analysis of the system components based on their reliability data should be done in order to optimize the maintenance program. This paper describes the methodology which was used at an aluminum plant in order to select critical machines and to develop an optimal maintenance policy based on reliability data of each machine, safety consequences of system failure, lead time and repair time, and components criticality.
Low-Cost Oil Quality Sensor Based on Changes in Complex Permittivity The role of knowledge modelling techniques in software development: a general approach based on a knowledge management tool
  • Angel Pérez
  • Mark Torres Hadfield
Pérez, Angel Torres; Hadfield, Mark. (2011). "Low-Cost Oil Quality Sensor Based on Changes in Complex Permittivity." Sensors 11, no. 11: 10675-10690. [4] J. CUENA, M. MOLINA, (2000) " The role of knowledge modelling techniques in software development: a general approach based on a knowledge management tool ", International Journal of Human-Computer Studies, Vol. 52, No. 3, pp. 385-421