Ömer Faruk Eker
Dr. Eker works as a senior data scientist at Tarentum AI:
- Over ten years of data science experience
- Four years of international business development and product management experience
- Involved in data science projects from various disciplines such as predictive maintenance, autonomous drive, churn prediction, risk analysis & credit scoring and marketing analytics, NLP, recommender systems
- Skilled in programming with Python
- Also contributed to the scientific knowledge with a book, eight journal papers, and numerous conference papers
Phone: +905326505186
Address: Artesis Technology Systems, Gebze, Kocaeli / TURKEY
- Over ten years of data science experience
- Four years of international business development and product management experience
- Involved in data science projects from various disciplines such as predictive maintenance, autonomous drive, churn prediction, risk analysis & credit scoring and marketing analytics, NLP, recommender systems
- Skilled in programming with Python
- Also contributed to the scientific knowledge with a book, eight journal papers, and numerous conference papers
Phone: +905326505186
Address: Artesis Technology Systems, Gebze, Kocaeli / TURKEY
less
InterestsView All (32)
Uploads
Papers by Ömer Faruk Eker
system/component by calculating its Remaining Useful Life (RUL) under degradation process. This paper presents a state-based prognostic
method using state duration information to predict the fatigue life or fatigue crack growth in structures. The process of implementing the
proposed prognostics algorithm consists of the following two stages: The first stage is identifying the health state of the system, while the
second stage is calculating the expected RUL. The presented prognostic method is applied on the Virkler fatigue crack growth dataset and
results show that the method efficiently predicts the remaining useful life of aluminium test specimens under constant amplitude fatigue load
cycles.
obtain the desired level of purification for liquids or gas.
Air, fuel, and oil filters are the most common examples in
industrial systems. Filter clogging is the main failure mode
leading to filter replacement or undesired outcomes such as
reduced performance and efficiency or cascading failures.
For example, contaminants in fuel (e.g. rust particles, paint
chips, dirt involved into fuel while tank is filling, tank
moisture rust) may lead to performance reduction in the
engine and rapid wear in the pump. Prognostics has
potential to avoid cost and increase safety when applied to
filters. One of the main challenges of prognostics is the lack
of failure degradation data obtained from industrial systems.
This paper presents the process of design and building of an
experimental rig to obtain prognostics data for filter
clogging mechanism and data obtained from the rig. Two
types of filters have been used during the accelerated filter
clogging and 23 run-to-failure data have been collected.
Flow rate and pressure sensors are used for condition
monitoring purposes. The filter clogging has been recorded
through a camera to evaluate the findings with pressure and
flow sensors. The data collected is very promising for
development of prognostics methodologies.
—The importance of railway transportation has beenincreasing in the world. Considering the current and future es-timates of high cargo and passenger transportation volume inrailways, prevention or reduction of delays due to any failure isbecoming ever more crucial. Railway turnout systems are one of the most critical pieces of equipment in railway infrastructure.When incipient failures occur, they mostly progress slowly fromthe fault-free to the failure state. Although studies focusing onthe identification of possible failures in railway turnout systemsexist in literature, neither the detection nor forecasting of fail-ure progression has been reported. This paper presents a simplestate-basedprognostic(SSBP)methodthataimstodetectandfore-cast failure progression in electromechanical systems. The methodis compared with Hidden-Markov-Model-based methods on realdata collected from a railway turnout system. Obtaining statisti-cally sufficient failure progression samples is difficult, consider-ing that the natural progression of failures in electromechanicalsystems may take years. In addition, validating the classificationmodel is difficult when the degradation is not observable. Datacollection and model validation strategies for failure progressionare also presented
—
Turnout systems on railways play critical role onreliability of railway infrastructure. Identification and predictionof failures on mechanical systems have been attractingresearchers and industry in recent years. Condition basedmaintenance focuses on failure identification and predictionusing sensory information collected real-time from sensorsembedded on electro-mechanical systems. This paper presentsneural network based failure prediction algorithm on railwayturnouts.
paper aims to develop a method to identify „drive
-rod out-of-
adjustment‟ failure mode, one of the most frequently observed
failure modes. Support Vector Machine with Gaussian kernel isused for classification. In addition, results of feature selectionwith statistical t-test and feature reduction with principalcomponent analysis are compared in the paper"
system/component by calculating its Remaining Useful Life (RUL) under degradation process. This paper presents a state-based prognostic
method using state duration information to predict the fatigue life or fatigue crack growth in structures. The process of implementing the
proposed prognostics algorithm consists of the following two stages: The first stage is identifying the health state of the system, while the
second stage is calculating the expected RUL. The presented prognostic method is applied on the Virkler fatigue crack growth dataset and
results show that the method efficiently predicts the remaining useful life of aluminium test specimens under constant amplitude fatigue load
cycles.
obtain the desired level of purification for liquids or gas.
Air, fuel, and oil filters are the most common examples in
industrial systems. Filter clogging is the main failure mode
leading to filter replacement or undesired outcomes such as
reduced performance and efficiency or cascading failures.
For example, contaminants in fuel (e.g. rust particles, paint
chips, dirt involved into fuel while tank is filling, tank
moisture rust) may lead to performance reduction in the
engine and rapid wear in the pump. Prognostics has
potential to avoid cost and increase safety when applied to
filters. One of the main challenges of prognostics is the lack
of failure degradation data obtained from industrial systems.
This paper presents the process of design and building of an
experimental rig to obtain prognostics data for filter
clogging mechanism and data obtained from the rig. Two
types of filters have been used during the accelerated filter
clogging and 23 run-to-failure data have been collected.
Flow rate and pressure sensors are used for condition
monitoring purposes. The filter clogging has been recorded
through a camera to evaluate the findings with pressure and
flow sensors. The data collected is very promising for
development of prognostics methodologies.
—The importance of railway transportation has beenincreasing in the world. Considering the current and future es-timates of high cargo and passenger transportation volume inrailways, prevention or reduction of delays due to any failure isbecoming ever more crucial. Railway turnout systems are one of the most critical pieces of equipment in railway infrastructure.When incipient failures occur, they mostly progress slowly fromthe fault-free to the failure state. Although studies focusing onthe identification of possible failures in railway turnout systemsexist in literature, neither the detection nor forecasting of fail-ure progression has been reported. This paper presents a simplestate-basedprognostic(SSBP)methodthataimstodetectandfore-cast failure progression in electromechanical systems. The methodis compared with Hidden-Markov-Model-based methods on realdata collected from a railway turnout system. Obtaining statisti-cally sufficient failure progression samples is difficult, consider-ing that the natural progression of failures in electromechanicalsystems may take years. In addition, validating the classificationmodel is difficult when the degradation is not observable. Datacollection and model validation strategies for failure progressionare also presented
—
Turnout systems on railways play critical role onreliability of railway infrastructure. Identification and predictionof failures on mechanical systems have been attractingresearchers and industry in recent years. Condition basedmaintenance focuses on failure identification and predictionusing sensory information collected real-time from sensorsembedded on electro-mechanical systems. This paper presentsneural network based failure prediction algorithm on railwayturnouts.
paper aims to develop a method to identify „drive
-rod out-of-
adjustment‟ failure mode, one of the most frequently observed
failure modes. Support Vector Machine with Gaussian kernel isused for classification. In addition, results of feature selectionwith statistical t-test and feature reduction with principalcomponent analysis are compared in the paper"