ISLAB/CAISR - User contributions [en]

PRIME

2017-11-16T10:54:08Z

Yuantao:

{{ResearchProjInfo
|Title=PRIME
|ContactInformation=Pablo
|ProjectResponsible=Pablo
|ProjectStart=2017/01/01
|ProjectEnd=2020/12/30
|ApplicationArea=Biometrics
}}
__NOTOC__
{{ShowResearchProject}}

Swedish Research Council, Research Project No: 2016-03497

[[Image:VRlogo.png|150px]]


[[Image:Eye mobile.png|300px]]

PRIME

2017-11-16T10:53:27Z

Yuantao:

{{ResearchProjInfo
|Title=PRIME
|ContactInformation=Pablo
|LogotypeFile=Grid.jpg
|ProjectResponsible=Pablo
|ProjectStart=2017/01/01
|ProjectEnd=2020/12/30
|ApplicationArea=Biometrics
}}
__NOTOC__
{{ShowResearchProject}}

Swedish Research Council, Research Project No: 2016-03497

[[Image:VRlogo.png|150px]]

This project relates broadly to ocular biometrics in unconstrained sensing environments. Particularly, to methods for reliable detection/segmentation of ocular regions, and reconstruction of low-resolution images. The specific research objectives are:

*'''Detection of ocular region in unconstrained sensing environments''' under variations in scale, illumination, pose, low resolution, noise, etc. This is novel, since the vast majority of ocular recognition works have relied on manual annotation.

*'''Super-resolution reconstruction of ocular images'''. It may be used to iteratively improve detection (which may improve reconstruction further too), and ultimately to get better recognition accuracy thanks to enhanced image quality. Despite low resolution is frequent in relaxed environments, few ocular reconstruction works exist.

*'''Ocular recognition by case studies using data at a distance and on the move'''. Fundamental research contributions can be greatly benefited with practical applications in mind, since they enable to assess merits of the developments. We will concentrate on two cases: cooperative scenario with personal devices (smartphone), and uncooperative with surveillance cameras.

A primary consequence will be facilitated user interaction by enabling the use of data acquired in a wide range of operational conditions. More comfort and convenience can be achieved thanks to the use of own devices and natural interaction patterns with digital systems.


[[Image:Eye mobile.png|300px]]

PRIME

2017-11-16T10:53:10Z

Yuantao:

{{ResearchProjInfo
|Title=PRIME
|ContactInformation=Fernando Alonso-Fernandez
|ShortDescription='''Ocular Biometrics in Unconstrained Sensing Environments'''
|Description=This is a four-years project financed the Swedish Research Council. The project is concerned with ocular biometrics in unconstrained sensing environments. Attention will be paid to the periocular modality (the part of the face surrounding the eye), which has shown a surprisingly high discrimination ability, and is the facial-ocular modality requiring the least constrained acquisition.

One goal is to contribute with methods for efficient ocular detection and segmentation. This is still a challenge, with most works relying on manual image annotation, or on detecting the full face, which may not be reliable for example under occlusion. We will continue initiated work with symmetry filters, and will explore deep learning algorithms too, which are giving promising results in many computer vision tasks. Low resolution is another limitation. Thus, another goal will be super-resolution (SR) reconstruction of ocular images. With few works focused on iris, and none on periocular, adaptation of the many available SR methods to the particularities of ocular images is a promising avenue yet to be explored.

Ubiquitous biometrics has emerged as critical not only in light of current security threats (e.g. identifying terrorists in surveillance videos), but also due to the proliferation of consumer electronics (e.g. smartphones) in need of continuous personal authentication for a wide variety of applications. By our contributions, we expect to be able to handle a wide range of variations in biometric imaging from these scenarios.
|LogotypeFile=Grid.jpg
|ProjectResponsible=Pablo
|ProjectStart=2017/01/01
|ProjectEnd=2020/12/30
|ApplicationArea=Biometrics
}}
__NOTOC__
{{ShowResearchProject}}

Swedish Research Council, Research Project No: 2016-03497

[[Image:VRlogo.png|150px]]

This project relates broadly to ocular biometrics in unconstrained sensing environments. Particularly, to methods for reliable detection/segmentation of ocular regions, and reconstruction of low-resolution images. The specific research objectives are:

*'''Detection of ocular region in unconstrained sensing environments''' under variations in scale, illumination, pose, low resolution, noise, etc. This is novel, since the vast majority of ocular recognition works have relied on manual annotation.

*'''Super-resolution reconstruction of ocular images'''. It may be used to iteratively improve detection (which may improve reconstruction further too), and ultimately to get better recognition accuracy thanks to enhanced image quality. Despite low resolution is frequent in relaxed environments, few ocular reconstruction works exist.

*'''Ocular recognition by case studies using data at a distance and on the move'''. Fundamental research contributions can be greatly benefited with practical applications in mind, since they enable to assess merits of the developments. We will concentrate on two cases: cooperative scenario with personal devices (smartphone), and uncooperative with surveillance cameras.

A primary consequence will be facilitated user interaction by enabling the use of data acquired in a wide range of operational conditions. More comfort and convenience can be achieved thanks to the use of own devices and natural interaction patterns with digital systems.


[[Image:Eye mobile.png|300px]]

PRIME

2017-11-16T10:52:49Z

Yuantao: Created page with "{{ResearchProjInfo |Title=PRIME |ContactInformation=Fernando Alonso-Fernandez |ShortDescription='''Ocular Biometrics in Unconstrained Sensing Environments''' |Description=This..."

{{ResearchProjInfo
|Title=PRIME
|ContactInformation=Fernando Alonso-Fernandez
|ShortDescription='''Ocular Biometrics in Unconstrained Sensing Environments'''
|Description=This is a four-years project financed the Swedish Research Council. The project is concerned with ocular biometrics in unconstrained sensing environments. Attention will be paid to the periocular modality (the part of the face surrounding the eye), which has shown a surprisingly high discrimination ability, and is the facial-ocular modality requiring the least constrained acquisition.

One goal is to contribute with methods for efficient ocular detection and segmentation. This is still a challenge, with most works relying on manual image annotation, or on detecting the full face, which may not be reliable for example under occlusion. We will continue initiated work with symmetry filters, and will explore deep learning algorithms too, which are giving promising results in many computer vision tasks. Low resolution is another limitation. Thus, another goal will be super-resolution (SR) reconstruction of ocular images. With few works focused on iris, and none on periocular, adaptation of the many available SR methods to the particularities of ocular images is a promising avenue yet to be explored.

Ubiquitous biometrics has emerged as critical not only in light of current security threats (e.g. identifying terrorists in surveillance videos), but also due to the proliferation of consumer electronics (e.g. smartphones) in need of continuous personal authentication for a wide variety of applications. By our contributions, we expect to be able to handle a wide range of variations in biometric imaging from these scenarios.
|LogotypeFile=Grid.jpg
|ProjectResponsible=Fernando Alonso-Fernandez
|ProjectStart=2017/01/01
|ProjectEnd=2020/12/30
|ApplicationArea=Biometrics
}}
__NOTOC__
{{ShowResearchProject}}

Swedish Research Council, Research Project No: 2016-03497

[[Image:VRlogo.png|150px]]

This project relates broadly to ocular biometrics in unconstrained sensing environments. Particularly, to methods for reliable detection/segmentation of ocular regions, and reconstruction of low-resolution images. The specific research objectives are:

*'''Detection of ocular region in unconstrained sensing environments''' under variations in scale, illumination, pose, low resolution, noise, etc. This is novel, since the vast majority of ocular recognition works have relied on manual annotation.

*'''Super-resolution reconstruction of ocular images'''. It may be used to iteratively improve detection (which may improve reconstruction further too), and ultimately to get better recognition accuracy thanks to enhanced image quality. Despite low resolution is frequent in relaxed environments, few ocular reconstruction works exist.

*'''Ocular recognition by case studies using data at a distance and on the move'''. Fundamental research contributions can be greatly benefited with practical applications in mind, since they enable to assess merits of the developments. We will concentrate on two cases: cooperative scenario with personal devices (smartphone), and uncooperative with surveillance cameras.

A primary consequence will be facilitated user interaction by enabling the use of data acquired in a wide range of operational conditions. More comfort and convenience can be achieved thanks to the use of own devices and natural interaction patterns with digital systems.


[[Image:Eye mobile.png|300px]]

VBPM

2017-11-09T17:55:13Z

Yuantao:

{{ResearchProjInfo
|Title=VBPM
|ContactInformation=Sławomir Nowaczyk, Yuantao Fan
|ShortDescription=Volvo Bus Predictive Maintenance
|Description=The overall objective of this project is to improve uptime for Volvo buses as well as scheduling maintenance cost-effectively. Guaranteeing vehicle uptime is important since downtime caused by component failures are increasingly difficult to identify and being dealt with as the complexity of modern transport solution increases. The project is aiming at developing a framework, powered by machine learning technique, for predicting component failures in buses and providing fleet operator decision support for scheduling maintenance. Proposed machine learning models will be built, tested and validated based on real data. This project is a collaboration with Volvo Bussar AB and Volvo Truck Technology.
|LogotypeFile=Procedure.png
|ProjectResponsible=Sławomir Nowaczyk
|ProjectStart=2017/01
|ProjectEnd=2018/06
|ApplicationArea=Intelligent Vehicles
|Lctitle=No
}}
{{AssignProjPartner}}
{{AssignProjPartner
|projectpartner=Volvo Bussar AB
}}
__NOTOC__
{{ShowResearchProject}}

== Background and Objectives ==

The current paradigm for maintaining industrial equipments is a combination of reactive and preventive actions. Take commercial transportation vehicles as example, they are typically maintained after an equipment failure occurs or according to preplanned visits to the workshops based on mileage or calendar time. This mixture of maintenance strategy is not ideal: i ) it does not perform maintenance pro-actively well before the failure happens, i.e. severe component failures usually result in extra damage to the system and could be prevented; ii ) planned maintenance with fixed time intervals does not guarantee all routinely changed parts have used all their potentials. Therefore, a shift of current maintenance strategy to one with more predictive maintenance is required: to inspect and repair components (well) before they causes a breakdown or severe damage to the system.

Nowadays, with the development of electronic devices and the emergence of Internet of Things, huge amount of sensor data collected and transmitted remotely can be utilized for equipment monitoring, fault detection and prognostics. By processing sensor date during operations, condition of the equipment will be accessed and maintenance decision will be made. This project will improve the work carried out in VPMS project. The proposed framework is capable of extracting relevant data sources from LVD (Logged Vehicle Data) and VSR (Vehicle Services Records) for predicting component failure uses, finding useful features and generate prediction models for component of importance and interests.

== Expectations ==

VBPM will drive the development of predictive maintenance innovation at Volvo Bussar AB. Proposed novel machine learning techniques for predicting component failures will be able to model component degradation and identify faults in advance, giving enough time margin for performing maintenance service and fixing worn-out components. This project will lead the transition of the current maintenance paradigm to one with more predictive maintenance at Volvo Bussar AB. The project are expect to improve maintenance efficiency and safety while reduce cost.

{{ShowProjectPublications}}

VBPM

2017-11-09T17:22:17Z

Yuantao:

{{ResearchProjInfo
|Title=VBPM
|ContactInformation=Sławomir Nowaczyk, Yuantao Fan
|ShortDescription=Volvo Bus Predictive Maintenance
|Description=The overall objective of this project is to improve uptime for Volvo buses as well as scheduling maintenance cost-effectively. Guaranteeing vehicle uptime is important since downtime due to component failures are increasingly difficult to dealt with as the complexity of modern transport solution increases. The project is aiming at developing a framework, powered by machine learning technique, for predicting component failures in buses and providing fleet operator decision support for scheduling maintenance. Proposed machine learning models will be built, tested and validated based on real data. This project is a collaboration with Volvo Bussar AB and Volvo Truck Technology.
|LogotypeFile=Procedure.png
|ProjectResponsible=Sławomir Nowaczyk
|ProjectStart=2017/01
|ProjectEnd=2018/06
|ApplicationArea=Intelligent Vehicles
|Lctitle=No
}}
{{AssignProjPartner}}
{{AssignProjPartner
|projectpartner=Volvo Bussar AB
}}
__NOTOC__
{{ShowResearchProject}}

== Background and Objectives ==

The current paradigm for maintaining industrial equipments is a combination of reactive and preventive actions. Take commercial transportation vehicles as example, they are typically maintained after an equipment failure occurs or according to preplanned visits to the workshops based on mileage or calendar time. This mixture of maintenance strategy is not ideal: i ) it does not perform maintenance pro-actively well before the failure happens, i.e. severe component failures usually result in extra damage to the system and could be prevented; ii ) planned maintenance with fixed time intervals does not guarantee all routinely changed parts have used all their potentials. Therefore, a shift of current maintenance strategy to one with more predictive maintenance is required: to inspect and repair components (well) before they causes a breakdown or severe damage to the system.

Nowadays, with the development of electronic devices and the emergence of Internet of Things, huge amount of sensor data collected and transmitted remotely can be utilized for equipment monitoring, fault detection and prognostics. By processing sensor date during operations, condition of the equipment will be accessed and maintenance decision will be made. This project will improve the work carried out in

This project will utilise data from Volvo bus, including Logged Vehicle Data and Vehicle Service Records.

== Expectations ==

HEALTH will enhance current 100% uptime promise of Volvo Trucks by expanding the existing range of predictive maintenance solutions. Novel Machine Learning methods for representing lifelong histories of trucks will be used to precisely identify vehicles that are likely to fail soon, and corrective actions will be suggested based on the probable failure causes. Overall effects will include prolonging vehicle life, providing more timely and cheaper maintenance, and increasing traffic safety.

== Scheduled planning and implementation ==
The HEALTH project is planned for two years, starting October 2017. The work will be carried out in a close collaboration between Volvo Trucks Aftermarket and Halmstad University. The project is divided into five work packages focusing on data aggregation, fully and partially observable sequence modeling, causal analysis and the demonstrator. Implementation includes research and development of new machine learning methods, their deployment, and finally evaluation in real business setting.

{{ShowProjectPublications}}

VBPM

2017-11-09T15:40:20Z

Yuantao:

{{ResearchProjInfo
|Title= VBPM
|ContactInformation=Sławomir Nowaczyk, Yuantao Fan
|ShortDescription=Volvo Bus Predictive Maintenance
|Description=The aim of Active@Work is to explore if mobile technology including a personalized decision support system, can have any effect on physical activity level, health, work ability, quality of life, work productivity or sick leave among individuals with osteoarthritis (OA). We also aim to study if there is any difference in effect between using mobile technology and activity monitoring alone or when continuous feedback concerning physical activity is added.

|LogotypeFile=Procedure.png
|ProjectResponsible=Sławomir Nowaczyk
|ProjectStart=2017/01
|ProjectEnd=2018/06
|ApplicationArea=Intelligent Vehicles
|Lctitle=No
}}
{{AssignProjPartner
|projectpartner=Region Halland
}}
__NOTOC__
{{ShowResearchProject}}

== Purpose and goals ==
Unplanned downtime can be avoided by accurate prediction of the failure through continuously monitoring of vehicles’ health status. However, to reveal patterns behind failures in a system as complex as a modern truck, new methods for analysing the data need to be developed. The HEALTH project aims to create a sequence model capturing the lifelong history of a truck, and use it to explain relations between different events such as failures, repairs, fault codes - leading to better maintenance.

== Expected effects and results ==
HEALTH will enhance current 100% uptime promise of Volvo Trucks by expanding the existing range of predictive maintenance solutions. Novel Machine Learning methods for representing lifelong histories of trucks will be used to precisely identify vehicles that are likely to fail soon, and corrective actions will be suggested based on the probable failure causes. Overall effects will include prolonging vehicle life, providing more timely and cheaper maintenance, and increasing traffic safety.

== Scheduled planning and implementation ==
The HEALTH project is planned for two years, starting October 2017. The work will be carried out in a close collaboration between Volvo Trucks Aftermarket and Halmstad University. The project is divided into five work packages focusing on data aggregation, fully and partially observable sequence modeling, causal analysis and the demonstrator. Implementation includes research and development of new machine learning methods, their deployment, and finally evaluation in real business setting.

{{ShowProjectPublications}}

VBPM

2017-11-09T15:38:39Z

Yuantao: Created page with "{{ResearchProjInfo |Title= VBPM |ContactInformation=Sławomir Nowaczyk, Yuantao Fan |ShortDescription=Volvo Bus Predictive Maintenance |LogotypeFile=Procedure.png |ProjectResp..."

{{ResearchProjInfo
|Title= VBPM
|ContactInformation=Sławomir Nowaczyk, Yuantao Fan
|ShortDescription=Volvo Bus Predictive Maintenance
|LogotypeFile=Procedure.png
|ProjectResponsible=Sławomir Nowaczyk
|ProjectStart=2017/01
|ProjectEnd=2018/06
|ApplicationArea=Intelligent Vehicles
|Lctitle=No
}}
{{AssignProjPartner
|projectpartner=Region Halland
}}
__NOTOC__
{{ShowResearchProject}}

== Purpose and goals ==
Unplanned downtime can be avoided by accurate prediction of the failure through continuously monitoring of vehicles’ health status. However, to reveal patterns behind failures in a system as complex as a modern truck, new methods for analysing the data need to be developed. The HEALTH project aims to create a sequence model capturing the lifelong history of a truck, and use it to explain relations between different events such as failures, repairs, fault codes - leading to better maintenance.

== Expected effects and results ==
HEALTH will enhance current 100% uptime promise of Volvo Trucks by expanding the existing range of predictive maintenance solutions. Novel Machine Learning methods for representing lifelong histories of trucks will be used to precisely identify vehicles that are likely to fail soon, and corrective actions will be suggested based on the probable failure causes. Overall effects will include prolonging vehicle life, providing more timely and cheaper maintenance, and increasing traffic safety.

== Scheduled planning and implementation ==
The HEALTH project is planned for two years, starting October 2017. The work will be carried out in a close collaboration between Volvo Trucks Aftermarket and Halmstad University. The project is divided into five work packages focusing on data aggregation, fully and partially observable sequence modeling, causal analysis and the demonstrator. Implementation includes research and development of new machine learning methods, their deployment, and finally evaluation in real business setting.

{{ShowProjectPublications}}

Recurrent and Deep Learning for Machine Prognostics

2017-10-04T17:58:16Z

Yuantao:

{{StudentProjectTemplate
|Summary=Construct and optimise Recurrent Neural Networks for industrial applications on machine prognostics; Augmenting industrial data for supervised learning
|Keywords=Recurrent Neural Networks, time series forecasting, supervised learning, Prognostics and Health Management, fault detection
|TimeFrame=Winter 2016 - Spring 2017
|References=[1] Liu, Jie, et al. An adaptive recurrent neural network for remaining useful life prediction of lithium-ion batteries. NATIONAL AERONAUTICS AND SPACE ADMINISTRATION MOFFETT FIELD CA AMES RESEARCH CENTER, 2010.

[2] Malhi, Arnaz, Ruqiang Yan, and Robert X. Gao. "Prognosis of defect propagation based on recurrent neural networks." IEEE Transactions on Instrumentation and Measurement 60.3 (2011): 703-711.

[3] Heimes, Felix O. "Recurrent neural networks for remaining useful life estimation." Prognostics and Health Management, 2008. PHM 2008. International Conference on. IEEE, 2008.

[4] Marco Rigamonti et al., "Echo State Network for the Remaining Useful Life Prediction of a Turbofan Engine." Third European Conference of the Prognostics and Health Management Society 2016, Bilbao, Spain, 5-8 July, 2016. PHM Society, 2016.

[5] Mandic, Danilo P., and Jonathon A. Chambers. Recurrent neural networks for prediction: learning algorithms, architectures and stability. New York: John Wiley, 2001.

[6] Jaeger, Herbert. Tutorial on training recurrent neural networks, covering BPPT, RTRL, EKF and the" echo state network" approach. Vol. 5. GMD-Forschungszentrum Informationstechnik, 2002.

[7] Sun, Jianzhong, et al. "Application of a state space modeling technique to system prognostics based on a health index for condition-based maintenance." Mechanical Systems and Signal Processing 28 (2012): 585-596.

[8] Wang, Wilson Q., M. Farid Golnaraghi, and Fathy Ismail. "Prognosis of machine health condition using neuro-fuzzy systems." Mechanical Systems and Signal Processing 18.4 (2004): 813-831.
|Prerequisites=Artificial Intelligence and Learning Systems courses; good knowledge in machine learning and neural networks; programming skills for implementing machine learning algorithms
|Supervisor=Sławomir Nowaczyk, Sepideh Pashami, Yuantao Fan,
|Examiner=Antanas Verikas
|Level=Master
|Status=Internal Draft
}}
Background:

The current paradigm for maintaining industrial equipments is a combination of reactive and preventive actions. Take commercial transportation vehicles as example, they are typically maintained after an equipment failure occurs or according to preplanned visits to the workshops based on mileage or calendar time. This mixture of maintenance strategy is not ideal: i ) it does not perform maintenance pro-actively well before the failure happens, i.e. severe component failures usually result in extra damage to the system and could be prevented; ii ) planned maintenance with fixed time intervals does not guarantee all routinely changed parts have used all their potentials. Therefore, a shift of current maintenance strategy to one with more predictive maintenance is required: to inspect and repair components (well) before they causes a breakdown or severe damage to the system.

Nowadays, with the development of electronic devices and the emergence of Internet of Things, huge amount of sensor data collected and transmitted remotely can be utilized for equipment monitoring, fault detection and prognostics. By processing sensor data during operations, condition of the equipment will be accessed and maintenance decision will be made. A common approach for machine prognostics is to estimate Remaining Useful Life (RUL) of the system. Many researchers applied Recurrent Neural Networks (RNNs) for estimating RUL based on sensor measurements, e.g. see reference [1,2,3,4]. Another common approach for prognostics is to generate index that reflects health status of equipment, e.g. reference [7,8], based on it's condition.

A recurrent neural network (RNN) is a type of artificial neural network where connections between units form a directed cycle. It can use their internal memory to process arbitrary sequences of inputs and capture dynamic temporal behaviour, tutorials can be found in [5,6].

In this project, the student will construct RNNs for predicting RUL of machines in different domains, including simulated data and real data from industrial application. The industrial data collected from large amount of vehicles performing transportation tasks in the field. Vehicle’s configuration, aggregated sensor values collected at different time are available. Service record that contains assessments and repair actions is provided. Based on sensor data and service records, a machine learning method for predicting RUL of equipment is expected to be proposed and evaluated.

This project is programming oriented, the student will be working with some of the libraries that includes RNNs implementations, e.g. Theano and Keras.

Objectives:

1. Construct Recurrent Neural Networks, optimise it's architecture and training algorithms for predicting Remaining Useful Life of equipment.

2. Investigate and propose different scenarios augmenting industrial data for machine prognostics, e.g. generating targets/teaching signals for RNNs to learn?

3. Investigate what type of data representation technique can be employed for this application, e.g. histograms of aggregated values as input and use CNN-RNN model for the networks.

Research Questions:

1. How to construct and train RNNs for predicting RUL?

2. How to augment industrial data for this study?
- What type of representations can be used?
- How to generate targets/teaching sequences based on service records? Does imposing arbitary conditions based on application improves prediction performance?

3. How to augment the network architecture for different types of data representation (e.g. histograms with aggregated values as input)?

Recurrent and Deep Learning for Machine Prognostics

2017-10-04T17:52:27Z

Yuantao:

{{StudentProjectTemplate
|Summary=Construct and optimise Recurrent Neural Networks for industrial applications on machine prognostics; Augmenting industrial data for supervised learning
|Keywords=Recurrent Neural Networks, time series forecasting, supervised learning, Prognostics and Health Management, fault detection
|TimeFrame=Winter 2016 - Spring 2017
|References=[1] Liu, Jie, et al. An adaptive recurrent neural network for remaining useful life prediction of lithium-ion batteries. NATIONAL AERONAUTICS AND SPACE ADMINISTRATION MOFFETT FIELD CA AMES RESEARCH CENTER, 2010.

[2] Malhi, Arnaz, Ruqiang Yan, and Robert X. Gao. "Prognosis of defect propagation based on recurrent neural networks." IEEE Transactions on Instrumentation and Measurement 60.3 (2011): 703-711.

[3] Heimes, Felix O. "Recurrent neural networks for remaining useful life estimation." Prognostics and Health Management, 2008. PHM 2008. International Conference on. IEEE, 2008.

[4] Marco Rigamonti et al., "Echo State Network for the Remaining Useful Life Prediction of a Turbofan Engine." Third European Conference of the Prognostics and Health Management Society 2016, Bilbao, Spain, 5-8 July, 2016. PHM Society, 2016.

[5] Mandic, Danilo P., and Jonathon A. Chambers. Recurrent neural networks for prediction: learning algorithms, architectures and stability. New York: John Wiley, 2001.

[6] Jaeger, Herbert. Tutorial on training recurrent neural networks, covering BPPT, RTRL, EKF and the" echo state network" approach. Vol. 5. GMD-Forschungszentrum Informationstechnik, 2002.

[7] Sun, Jianzhong, et al. "Application of a state space modeling technique to system prognostics based on a health index for condition-based maintenance." Mechanical Systems and Signal Processing 28 (2012): 585-596.

[8] Wang, Wilson Q., M. Farid Golnaraghi, and Fathy Ismail. "Prognosis of machine health condition using neuro-fuzzy systems." Mechanical Systems and Signal Processing 18.4 (2004): 813-831.
|Prerequisites=Artificial Intelligence and Learning Systems courses; good knowledge in machine learning and neural networks; programming skills for implementing machine learning algorithms
|Supervisor=Sławomir Nowaczyk, Sepideh Pashami, Yuantao Fan,
|Examiner=Antanas Verikas
|Level=Master
|Status=Open
}}
Background:

The current paradigm for maintaining industrial equipments is a combination of reactive and preventive actions. Take commercial transportation vehicles as example, they are typically maintained after an equipment failure occurs or according to preplanned visits to the workshops based on mileage or calendar time. This mixture of maintenance strategy is not ideal: i ) it does not perform maintenance pro-actively well before the failure happens, i.e. severe component failures usually result in extra damage to the system and could be prevented; ii ) planned maintenance with fixed time intervals does not guarantee all routinely changed parts have used all their potentials. Therefore, a shift of current maintenance strategy to one with more predictive maintenance is required: to inspect and repair components (well) before they causes a breakdown or severe damage to the system.

Nowadays, with the development of electronic devices and the emergence of Internet of Things, huge amount of sensor data collected and transmitted remotely can be utilized for equipment monitoring, fault detection and prognostics. By processing sensor data during operations, condition of the equipment will be accessed and maintenance decision will be made. A common approach for machine prognostics is to estimate Remaining Useful Life (RUL) of the system. Many researchers applied Recurrent Neural Networks (RNNs) for estimating RUL based on sensor measurements, e.g. see reference [1,2,3,4]. Another common approach for prognostics is to generate index that reflects health status of equipment, e.g. reference [7,8], based on it's condition.

A recurrent neural network (RNN) is a type of artificial neural network where connections between units form a directed cycle. It can use their internal memory to process arbitrary sequences of inputs and capture dynamic temporal behaviour, tutorials can be found in [5,6].

In this project, the student will construct RNNs for predicting RUL of machines in different domains, including simulated data and real data from industrial application. The industrial data collected from large amount of vehicles performing transportation tasks in the field. Vehicle’s configuration, aggregated sensor values collected at different time are available. Service record that contains assessments and repair actions is provided. Based on sensor data and service records, a machine learning method for predicting RUL of equipment is expected to be proposed and evaluated.

This project is programming oriented, the student will be working with some of the libraries that includes RNNs implementations, e.g. Theano and Keras.

Objectives:

1. Construct Recurrent Neural Networks, optimise it's architecture and training algorithms for predicting Remaining Useful Life of equipment.

2. Investigate and propose different scenarios augmenting industrial data for machine prognostics, e.g. generating targets/teaching signals for RNNs to learn?

3. Investigate what type of data representation technique can be employed for this application, e.g. histograms of aggregated values as input and use CNN-RNN model for the networks.

Research Questions:

1. How to construct and train RNNs for predicting RUL?

2. How to augment industrial data for this study?
- What type of representations can be used?
- How to generate targets/teaching sequences based on service records? Does imposing arbitary conditions based on application improves prediction performance?

3. How to augment the network architecture for different types of data representation (e.g. histograms with aggregated values as input)?

Recurrent and Deep Learning for Machine Prognostics

2017-10-04T17:47:00Z

Yuantao: Created page with "{{StudentProjectTemplate |Summary=Construct and optimise Recurrent Neural Networks for industrial applications on machine prognostics; Augmenting industrial data for supervise..."

{{StudentProjectTemplate
|Summary=Construct and optimise Recurrent Neural Networks for industrial applications on machine prognostics; Augmenting industrial data for supervised learning
|Keywords=Recurrent Neural Networks, time series forecasting, supervised learning, Prognostics and Health Management, fault detection
|TimeFrame=Winter 2016 - Spring 2017
|References=[1] Liu, Jie, et al. An adaptive recurrent neural network for remaining useful life prediction of lithium-ion batteries. NATIONAL AERONAUTICS AND SPACE ADMINISTRATION MOFFETT FIELD CA AMES RESEARCH CENTER, 2010.

[2] Malhi, Arnaz, Ruqiang Yan, and Robert X. Gao. "Prognosis of defect propagation based on recurrent neural networks." IEEE Transactions on Instrumentation and Measurement 60.3 (2011): 703-711.

[3] Heimes, Felix O. "Recurrent neural networks for remaining useful life estimation." Prognostics and Health Management, 2008. PHM 2008. International Conference on. IEEE, 2008.

[4] Marco Rigamonti et al., "Echo State Network for the Remaining Useful Life Prediction of a Turbofan Engine." Third European Conference of the Prognostics and Health Management Society 2016, Bilbao, Spain, 5-8 July, 2016. PHM Society, 2016.

[5] Mandic, Danilo P., and Jonathon A. Chambers. Recurrent neural networks for prediction: learning algorithms, architectures and stability. New York: John Wiley, 2001.

[6] Jaeger, Herbert. Tutorial on training recurrent neural networks, covering BPPT, RTRL, EKF and the" echo state network" approach. Vol. 5. GMD-Forschungszentrum Informationstechnik, 2002.

[7] Sun, Jianzhong, et al. "Application of a state space modeling technique to system prognostics based on a health index for condition-based maintenance." Mechanical Systems and Signal Processing 28 (2012): 585-596.

[8] Wang, Wilson Q., M. Farid Golnaraghi, and Fathy Ismail. "Prognosis of machine health condition using neuro-fuzzy systems." Mechanical Systems and Signal Processing 18.4 (2004): 813-831.
|Prerequisites=Artificial Intelligence and Learning Systems courses; good knowledge in machine learning and neural networks; programming skills for implementing machine learning algorithms
|Supervisor=Sławomir Nowaczyk, Sepideh Pashami, Yuantao Fan,
|Examiner=Antanas Verikas
|Level=Master
|Status=Open
}}
Background:

The current paradigm for maintaining industrial equipments is a combination of reactive and preventive actions. Take commercial transportation vehicles as example, they are typically maintained after an equipment failure occurs or according to preplanned visits to the workshops based on mileage or calendar time. This mixture of maintenance strategy is not ideal: i ) it does not perform maintenance pro-actively well before the failure happens, i.e. severe component failures usually result in extra damage to the system and could be prevented; ii ) planned maintenance with fixed time intervals does not guarantee all routinely changed parts have used all their potentials. Therefore, a shift of current maintenance strategy to one with more predictive maintenance is required: to inspect and repair components (well) before they causes a breakdown or severe damage to the system.

Nowadays, with the development of electronic devices and the emergence of Internet of Things, huge amount of sensor data collected and transmitted remotely can be utilized for equipment monitoring, fault detection and prognostics. By processing sensor data during operations, condition of the equipment will be accessed and maintenance decision will be made. A common approach for machine prognostics is to estimate Remaining Useful Life (RUL) of the system. Many researchers applied Recurrent Neural Networks (RNNs) for estimating RUL based on sensor measurements, e.g. see reference [1,2,3,4]. Another common approach for prognostics is to generate index that reflects health status of equipment, e.g. reference [7,8], based on it's condition.

A recurrent neural network (RNN) is a type of artificial neural network where connections between units form a directed cycle. It can use their internal memory to process arbitrary sequences of inputs and capture dynamic temporal behaviour, tutorials can be found in [5,6].

In this project, the student will construct RNNs for predicting RUL of machines in different domains, including simulated data and real data from industrial application. The industrial data collected from large amount of vehicles performing transportation tasks in the field. Vehicle’s configuration, aggregated sensor values collected at different time are available. Service record that contains assessments and repair actions is provided. Based on sensor data and service records, a method, based on RNNs, for predicting RUL of equipment is expected to be proposed and evaluated.

This project is programming oriented, the student will be working with some of the libraries that includes RNNs implementations, e.g. Theano and Keras.

Objectives:

1. Construct Recurrent Neural Networks, optimise it's architecture and training algorithms for predicting Remaining Useful Life of equipment.

2. Investigate and propose different scenarios augmenting industrial data for machine prognostics, e.g. generating targets/teaching signals for RNNs to learn?

3. Investigate what type of data representation technique can be employed for this application, e.g. histograms of aggregated values as input and use CNN-RNN model for the networks.

Research Questions:

1. How to construct and train RNNs for predicting RUL?

1. How to augment industrial data for this study?
- What type of representations can be used?
- How to generate targets/teaching sequences based on service records? Does imposing arbitary conditions based on application improves prediction performance?

3. How to augment the network architecture for different types of data representation (e.g. histograms with aggregated values as input)?