Pub Date : 2023-06-05DOI: 10.1109/ICPHM57936.2023.10194225
Brett S. Sicard, Quade Butler, Youssef Ziada, S. Gadsden
The manufacturing world has advanced to the fourth industrial revolution (4IR). Machine tools, especially computer numerical control (CNC) machine tools are an essential part of manufacturing. An important part of the 4IR is predictive maintenance (PM). PM is key in ensuring the availability and high quality of parts produced by machine tools. An important part of CNC machine tools is their feed drives. It is essential to implement PM to keep these components in good working order. Often PM methods will need to be developed and tested on experimental setups before they can be implemented in production. This work examines the literature on experimental setups for feed drive condition monitoring, fault detection and PM and seeks to disseminate and organize information about methods and equipment used in these setups. Three primary factors were analyzed from these papers: the methods used to implement wear and faults, the external loading methods, and which sensors were used and where the sensors were installed. This work seeks to aid others who wish to create their own experimental setup to easily access information about the experimental setups of previous works on linear feed drive PM. A few trends were observed after examining the literature. A large quantity of experimental setups studied faults in ball screws, specifically preload in ball screws. A wide variety of sensors were used, the most popular being accelerometers. There was a lack of methods to implement external loading, with most papers using adjustable worktable weights or magnetic brakes.
{"title":"Experimental Setups for Linear Feed Drive Predictive Maintenance: A Review","authors":"Brett S. Sicard, Quade Butler, Youssef Ziada, S. Gadsden","doi":"10.1109/ICPHM57936.2023.10194225","DOIUrl":"https://doi.org/10.1109/ICPHM57936.2023.10194225","url":null,"abstract":"The manufacturing world has advanced to the fourth industrial revolution (4IR). Machine tools, especially computer numerical control (CNC) machine tools are an essential part of manufacturing. An important part of the 4IR is predictive maintenance (PM). PM is key in ensuring the availability and high quality of parts produced by machine tools. An important part of CNC machine tools is their feed drives. It is essential to implement PM to keep these components in good working order. Often PM methods will need to be developed and tested on experimental setups before they can be implemented in production. This work examines the literature on experimental setups for feed drive condition monitoring, fault detection and PM and seeks to disseminate and organize information about methods and equipment used in these setups. Three primary factors were analyzed from these papers: the methods used to implement wear and faults, the external loading methods, and which sensors were used and where the sensors were installed. This work seeks to aid others who wish to create their own experimental setup to easily access information about the experimental setups of previous works on linear feed drive PM. A few trends were observed after examining the literature. A large quantity of experimental setups studied faults in ball screws, specifically preload in ball screws. A wide variety of sensors were used, the most popular being accelerometers. There was a lack of methods to implement external loading, with most papers using adjustable worktable weights or magnetic brakes.","PeriodicalId":169274,"journal":{"name":"2023 IEEE International Conference on Prognostics and Health Management (ICPHM)","volume":"4 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-06-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130291764","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-06-05DOI: 10.1109/ICPHM57936.2023.10194013
Vincent Mendoza, John V. Morgan, Marlene Haag
The following paper discusses a digital Reliability, Availability, Maintainability and Safety (RAMS) ecosystem that provides model-based insight into the core elements of RAMS in order to enable CBM+ and the real-time health state of a system. The case focuses on using the combined capabilities of various Digital Twin models for data-driven, informed decision support and technical analysis. With reliability regularly an afterthought during design and development we argue that if taken into account during these phases systems will be more predictable, reliable and will limit the users' exposure to consequences. The example is based on a fictional system – the DeLorean Time Machine from Back to the Future – to illustrate that the solution is system agnostic and may be applied to any type of complex, technical system and any industry space.
{"title":"Using Digital Twins for CBM+ and RAMS Decision Support","authors":"Vincent Mendoza, John V. Morgan, Marlene Haag","doi":"10.1109/ICPHM57936.2023.10194013","DOIUrl":"https://doi.org/10.1109/ICPHM57936.2023.10194013","url":null,"abstract":"The following paper discusses a digital Reliability, Availability, Maintainability and Safety (RAMS) ecosystem that provides model-based insight into the core elements of RAMS in order to enable CBM+ and the real-time health state of a system. The case focuses on using the combined capabilities of various Digital Twin models for data-driven, informed decision support and technical analysis. With reliability regularly an afterthought during design and development we argue that if taken into account during these phases systems will be more predictable, reliable and will limit the users' exposure to consequences. The example is based on a fictional system – the DeLorean Time Machine from Back to the Future – to illustrate that the solution is system agnostic and may be applied to any type of complex, technical system and any industry space.","PeriodicalId":169274,"journal":{"name":"2023 IEEE International Conference on Prognostics and Health Management (ICPHM)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-06-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130295743","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-06-05DOI: 10.1109/ICPHM57936.2023.10194159
Josí Joaquín Mendoza Lopetegui, Gianluca Papa, M. Tanelli
Ground handling maneuvers in aircraft are strongly affected by the operational condition of the system. In particular, the shock absorbers present in the Main Landing Gear may have an incorrect amount of oil and/or gas, which deteriorates their performance and can pose a safety hazard for the pilot. In this paper, different methods are proposed to automatically assess the shock absorber status during ground braking maneuvers while the anti-skid system is active. To study the problem, a validated multibody aircraft simulator in a MATLAB/Simulink environment is used. Different data-driven algorithms and sensor placements for the data collection are proposed and evaluated, leveraging the simulator by conducting braking maneuvers over the operational envelope of the system. It is found that a Gaussian Process Regression model preprocessed by a Principal Component Analysis projection based on measurements of the vertical acceleration of the aircraft's body yields promising results.
{"title":"Data-driven Health Monitoring and Anomaly Detection in Aircraft Shock Absorbers","authors":"Josí Joaquín Mendoza Lopetegui, Gianluca Papa, M. Tanelli","doi":"10.1109/ICPHM57936.2023.10194159","DOIUrl":"https://doi.org/10.1109/ICPHM57936.2023.10194159","url":null,"abstract":"Ground handling maneuvers in aircraft are strongly affected by the operational condition of the system. In particular, the shock absorbers present in the Main Landing Gear may have an incorrect amount of oil and/or gas, which deteriorates their performance and can pose a safety hazard for the pilot. In this paper, different methods are proposed to automatically assess the shock absorber status during ground braking maneuvers while the anti-skid system is active. To study the problem, a validated multibody aircraft simulator in a MATLAB/Simulink environment is used. Different data-driven algorithms and sensor placements for the data collection are proposed and evaluated, leveraging the simulator by conducting braking maneuvers over the operational envelope of the system. It is found that a Gaussian Process Regression model preprocessed by a Principal Component Analysis projection based on measurements of the vertical acceleration of the aircraft's body yields promising results.","PeriodicalId":169274,"journal":{"name":"2023 IEEE International Conference on Prognostics and Health Management (ICPHM)","volume":"14 1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-06-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128867014","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-06-05DOI: 10.1109/ICPHM57936.2023.10194165
Yamini Kotriwar, Obaid Elshafiey, Lei Peng, Zi Li, Vijay Srinivasan, Eric Davis, Y. Deng
The structural and material aging of the energy and transportation infrastructure requires the development of faster, better, and more efficient non-destructive evaluation (NDE) techniques to assess remaining life and structural health for prognostics and structural health management. Composite materials such as carbon fiber reinforced polymer (CFRP) have beneficial properties such as corrosion resistance, durability, and lightweight, which reduce maintenance requirements and extend service life. Their an-isotropic dielectric and mechanical properties make it challenging for NDE techniques to detect and locate material defects. A miniaturized capacitive imaging system was developed to detect multiple types of defects in CFRP materials. However, algorithms to convert the raw imaging data into defect detection, classification, sizing, and location is not currently available. This paper presents a defect localization algorithm using a gradient response feature-based method to reduce the noise in the imaging data. The algorithm calculates the co-occurrence matrix of the image. From this matrix, the local features such as contrast, homogeneity, energy, and correlation are extracted. A combination of these features is selected to define a defect area. The features extracted from the image processing are classified using a support vector machine (SVM) algorithm. The location of the defects identified through the algorithm is compared with the ground truth to achieve a probability of detection of 82%.
{"title":"Gradient feature-based method for Defect Detection of Carbon Fiber Reinforced Polymer Materials","authors":"Yamini Kotriwar, Obaid Elshafiey, Lei Peng, Zi Li, Vijay Srinivasan, Eric Davis, Y. Deng","doi":"10.1109/ICPHM57936.2023.10194165","DOIUrl":"https://doi.org/10.1109/ICPHM57936.2023.10194165","url":null,"abstract":"The structural and material aging of the energy and transportation infrastructure requires the development of faster, better, and more efficient non-destructive evaluation (NDE) techniques to assess remaining life and structural health for prognostics and structural health management. Composite materials such as carbon fiber reinforced polymer (CFRP) have beneficial properties such as corrosion resistance, durability, and lightweight, which reduce maintenance requirements and extend service life. Their an-isotropic dielectric and mechanical properties make it challenging for NDE techniques to detect and locate material defects. A miniaturized capacitive imaging system was developed to detect multiple types of defects in CFRP materials. However, algorithms to convert the raw imaging data into defect detection, classification, sizing, and location is not currently available. This paper presents a defect localization algorithm using a gradient response feature-based method to reduce the noise in the imaging data. The algorithm calculates the co-occurrence matrix of the image. From this matrix, the local features such as contrast, homogeneity, energy, and correlation are extracted. A combination of these features is selected to define a defect area. The features extracted from the image processing are classified using a support vector machine (SVM) algorithm. The location of the defects identified through the algorithm is compared with the ground truth to achieve a probability of detection of 82%.","PeriodicalId":169274,"journal":{"name":"2023 IEEE International Conference on Prognostics and Health Management (ICPHM)","volume":"5 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-06-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126255759","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-06-05DOI: 10.1109/ICPHM57936.2023.10194014
Zhenning Li, Hongkai Jiang, Shaowei Liu, Ruixin Wang
The reliable operation of rolling bearings is related to machinery safety. However, fault signals encountered in practical engineering applications are often characterized by high-dimensionality, complexity, and volume, which restricts the application of deep neural networks in fault diagnosis. Additionally, conventional diagnostic methods are limited by their reliance on manual feature extraction and a significant quantity of labeled samples, which can be time-consuming and resource-intensive. To address these limitations and improve the performance of fault diagnosis in the absence of labeled samples, an intelligent diagnostic agent (TERL-Agent) that combines transfer learning, ensemble learning and reinforcement learning is proposed. Firstly, an intelligent diagnostic agent is constructed by ensemble learning, which combines multiple reinforcement learning agents based on the Deep Q Network structure and has interactive learning capability to learn and classify fault data in the source domain environment. Secondly, transfer learning is used to transfer the feature extraction ability of the source domain intelligent diagnostic agent to the target intelligent diagnostic agent. Finally, the obtained target intelligent diagnostic agent is evaluated on fault data in the target domain and compared with other methods. The results indicate that the proposed method exhibits remarkable advantages and has great potential for practical application in fault diagnosis.
{"title":"Fault diagnosis of rolling bearing using a transfer ensemble deep reinforcement learning method","authors":"Zhenning Li, Hongkai Jiang, Shaowei Liu, Ruixin Wang","doi":"10.1109/ICPHM57936.2023.10194014","DOIUrl":"https://doi.org/10.1109/ICPHM57936.2023.10194014","url":null,"abstract":"The reliable operation of rolling bearings is related to machinery safety. However, fault signals encountered in practical engineering applications are often characterized by high-dimensionality, complexity, and volume, which restricts the application of deep neural networks in fault diagnosis. Additionally, conventional diagnostic methods are limited by their reliance on manual feature extraction and a significant quantity of labeled samples, which can be time-consuming and resource-intensive. To address these limitations and improve the performance of fault diagnosis in the absence of labeled samples, an intelligent diagnostic agent (TERL-Agent) that combines transfer learning, ensemble learning and reinforcement learning is proposed. Firstly, an intelligent diagnostic agent is constructed by ensemble learning, which combines multiple reinforcement learning agents based on the Deep Q Network structure and has interactive learning capability to learn and classify fault data in the source domain environment. Secondly, transfer learning is used to transfer the feature extraction ability of the source domain intelligent diagnostic agent to the target intelligent diagnostic agent. Finally, the obtained target intelligent diagnostic agent is evaluated on fault data in the target domain and compared with other methods. The results indicate that the proposed method exhibits remarkable advantages and has great potential for practical application in fault diagnosis.","PeriodicalId":169274,"journal":{"name":"2023 IEEE International Conference on Prognostics and Health Management (ICPHM)","volume":"34 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-06-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129206576","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-06-05DOI: 10.1109/ICPHM57936.2023.10193932
Rajesh Murthy
With the confluence of Artificial Intelligence (AI), Big Data Analytics, smart sensors supplemented by internet of things (IOT) technologies, 5G and fog/edge intelligence for secure and reliable enterprise solutions, software systems are increasing in complexity. With growing cyber threats and complexity, there is a need for a secure and resilience framework to ensure continuous and reliable system operations. In this paper, we explore the concepts of resilience, security, and associated capabilities using a minimalist potential holistic PHM end to end architecture with machine learning and data management operations (DataOps). The paper concludes with potential directions and impacts of emerging technologies in system resilience and security. Data and signal are used interchangeably in this paper.
{"title":"Exploring the Use of PHM for Software System Security and Resilience","authors":"Rajesh Murthy","doi":"10.1109/ICPHM57936.2023.10193932","DOIUrl":"https://doi.org/10.1109/ICPHM57936.2023.10193932","url":null,"abstract":"With the confluence of Artificial Intelligence (AI), Big Data Analytics, smart sensors supplemented by internet of things (IOT) technologies, 5G and fog/edge intelligence for secure and reliable enterprise solutions, software systems are increasing in complexity. With growing cyber threats and complexity, there is a need for a secure and resilience framework to ensure continuous and reliable system operations. In this paper, we explore the concepts of resilience, security, and associated capabilities using a minimalist potential holistic PHM end to end architecture with machine learning and data management operations (DataOps). The paper concludes with potential directions and impacts of emerging technologies in system resilience and security. Data and signal are used interchangeably in this paper.","PeriodicalId":169274,"journal":{"name":"2023 IEEE International Conference on Prognostics and Health Management (ICPHM)","volume":"98 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-06-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125460567","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-06-05DOI: 10.1109/ICPHM57936.2023.10193864
Ethan Wescoat, Vinita Jansari, L. Mears
Prognostics Health Management (PHM) in manu-facturing seeks to reduce the amount of unexpected downtime that inhibits manufacturing competitiveness. However, a common challenge for the manufacturing industry is the lack of known failure data to train a predictive classifier. This work optimizes the necessary quantity of required failure training data and healthy data for three different exemplar datasets by assessing classifier performance. Particle swarm optimization with penalty factors associated with the training data amount were used to identify the required training data amount for fault classification. Two separate analysis cases are considered: a binary classification and multi-class classification case termed the progressive case. In both analysis cases, the optimal training data depended on how separable the bearing data were between the different baseline and defect stages. In those instances where the differences in the data classes were apparent, the bearing data optimal training data amount was lower than in those instances where the data class differences were not present. Future work focuses on the investigation of these overlap cases to determine the best means for classifying progressive damage for remaining useful life calculations.
{"title":"Optimizing Data Training Quantity for Bearing Condition Monitoring","authors":"Ethan Wescoat, Vinita Jansari, L. Mears","doi":"10.1109/ICPHM57936.2023.10193864","DOIUrl":"https://doi.org/10.1109/ICPHM57936.2023.10193864","url":null,"abstract":"Prognostics Health Management (PHM) in manu-facturing seeks to reduce the amount of unexpected downtime that inhibits manufacturing competitiveness. However, a common challenge for the manufacturing industry is the lack of known failure data to train a predictive classifier. This work optimizes the necessary quantity of required failure training data and healthy data for three different exemplar datasets by assessing classifier performance. Particle swarm optimization with penalty factors associated with the training data amount were used to identify the required training data amount for fault classification. Two separate analysis cases are considered: a binary classification and multi-class classification case termed the progressive case. In both analysis cases, the optimal training data depended on how separable the bearing data were between the different baseline and defect stages. In those instances where the differences in the data classes were apparent, the bearing data optimal training data amount was lower than in those instances where the data class differences were not present. Future work focuses on the investigation of these overlap cases to determine the best means for classifying progressive damage for remaining useful life calculations.","PeriodicalId":169274,"journal":{"name":"2023 IEEE International Conference on Prognostics and Health Management (ICPHM)","volume":"10 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-06-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127034300","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-06-05DOI: 10.1109/ICPHM57936.2023.10194065
Masuda A. Tonima, Austin DeHart, Deniz Tabakci, Piramon Tisapramotkul, Andrew Munro-West, Aarushi Mehra, T. Shoa
While Li-ion batteries have proven long lifetimes, an accurate assessment of the battery ca-pacity and its remaining life cannot yet be made using current Battery Management Systems (BMS) devices. Battery sensors used in BMS typically mon-itor voltage, current and temperature of the battery, in order to predict the state of health (SoH) of the battery. SoH is a measure that indicates the remaining capacity that had been affected by degradation. Information obtained by monitoring voltage, current and temperature are often not sufficient to predict SoH. In this study we captured extra information from interfacial layers of the battery through applying Electrochemical Impedance Spectroscopy (EIS) and employed a XGBoost-based machine learning approach to train our models. The results show that SoH of batteries can be predicted with 90% accuracy, 95% confidence and 82% reliability. Additionally, it was shown that accuracy could be maintained with little to no change even when the number of features was dramatically reduced and the sample size was minimal, thus making this method very practical for embedded EIS/AI based solutions.
{"title":"Electrochemical Impedance Spectroscopy (EIS) and Machine Learning based Battery State of Health (SoH) Estimation","authors":"Masuda A. Tonima, Austin DeHart, Deniz Tabakci, Piramon Tisapramotkul, Andrew Munro-West, Aarushi Mehra, T. Shoa","doi":"10.1109/ICPHM57936.2023.10194065","DOIUrl":"https://doi.org/10.1109/ICPHM57936.2023.10194065","url":null,"abstract":"While Li-ion batteries have proven long lifetimes, an accurate assessment of the battery ca-pacity and its remaining life cannot yet be made using current Battery Management Systems (BMS) devices. Battery sensors used in BMS typically mon-itor voltage, current and temperature of the battery, in order to predict the state of health (SoH) of the battery. SoH is a measure that indicates the remaining capacity that had been affected by degradation. Information obtained by monitoring voltage, current and temperature are often not sufficient to predict SoH. In this study we captured extra information from interfacial layers of the battery through applying Electrochemical Impedance Spectroscopy (EIS) and employed a XGBoost-based machine learning approach to train our models. The results show that SoH of batteries can be predicted with 90% accuracy, 95% confidence and 82% reliability. Additionally, it was shown that accuracy could be maintained with little to no change even when the number of features was dramatically reduced and the sample size was minimal, thus making this method very practical for embedded EIS/AI based solutions.","PeriodicalId":169274,"journal":{"name":"2023 IEEE International Conference on Prognostics and Health Management (ICPHM)","volume":"17 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-06-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131268963","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-06-05DOI: 10.1109/ICPHM57936.2023.10194050
Kaiji Sun, S. Magnússon, O. Steinert, Tony Lindgren
The availability of data gathered from industrial sensors has increased expeditiously in recent years. These data are valuable assets in delivering exceptional services for manufacturing enterprises. We see growing interests and expectations from manufacturers in deploying artificial intelligence for predictive maintenance. The paper has adopted and transferred a state-of-the-art method from few-shot learning to failure prognostics using the Siamese neural network based contractive learning. The method has three main characteristics on top of the highest performance - a sensitivity of 98.4% for Scania truck's air pressure system failure capture, compared to the methods proposed by the previous related research: prediction stability, deployment flexibility, and the robust multi-shot diagnosis based on selected historical reference samples.
{"title":"Robust Contrastive Learning and Multi-shot Voting for High-dimensional Multivariate Data-driven Prognostics","authors":"Kaiji Sun, S. Magnússon, O. Steinert, Tony Lindgren","doi":"10.1109/ICPHM57936.2023.10194050","DOIUrl":"https://doi.org/10.1109/ICPHM57936.2023.10194050","url":null,"abstract":"The availability of data gathered from industrial sensors has increased expeditiously in recent years. These data are valuable assets in delivering exceptional services for manufacturing enterprises. We see growing interests and expectations from manufacturers in deploying artificial intelligence for predictive maintenance. The paper has adopted and transferred a state-of-the-art method from few-shot learning to failure prognostics using the Siamese neural network based contractive learning. The method has three main characteristics on top of the highest performance - a sensitivity of 98.4% for Scania truck's air pressure system failure capture, compared to the methods proposed by the previous related research: prediction stability, deployment flexibility, and the robust multi-shot diagnosis based on selected historical reference samples.","PeriodicalId":169274,"journal":{"name":"2023 IEEE International Conference on Prognostics and Health Management (ICPHM)","volume":"47 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-06-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132901620","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-06-05DOI: 10.1109/ICPHM57936.2023.10194226
Ryosuke Takayama, Masanao Natsumeda, T. Yairi
Accurate remaining useful life (RUL) prediction is an essential for efficient maintenance. In recent years, with the rapid development of industrial big data, many data-driven methods for RUL prediction have made significant progress, especially using deep learning. However, most of the proposed deep learning models only utilize labeled data and require a large amount of labeled data. In practice, the component of equipment is often replaced with a new one before it fails by preventive maintenance, resulting in a small number of failure histories and a large number of suspension histories. In other words, we have a small amount of labeled data and a large amount of unlabeled data. This paper proposes a new semi-supervised RUL prediction method using pseudo labels with flexibility in model architecture and low computational cost. For each suspension history, optimal pseudo labels are estimated using a likelihood-based method that takes into account important constraints, which enables more effective use of the information in both failure and suspension histories. The experiments on the C-MAPSS dataset validate the prediction accuracy of the proposed approach and provide several insights.
{"title":"A semi-supervised RUL prediction with likelihood-based pseudo labeling for suspension histories","authors":"Ryosuke Takayama, Masanao Natsumeda, T. Yairi","doi":"10.1109/ICPHM57936.2023.10194226","DOIUrl":"https://doi.org/10.1109/ICPHM57936.2023.10194226","url":null,"abstract":"Accurate remaining useful life (RUL) prediction is an essential for efficient maintenance. In recent years, with the rapid development of industrial big data, many data-driven methods for RUL prediction have made significant progress, especially using deep learning. However, most of the proposed deep learning models only utilize labeled data and require a large amount of labeled data. In practice, the component of equipment is often replaced with a new one before it fails by preventive maintenance, resulting in a small number of failure histories and a large number of suspension histories. In other words, we have a small amount of labeled data and a large amount of unlabeled data. This paper proposes a new semi-supervised RUL prediction method using pseudo labels with flexibility in model architecture and low computational cost. For each suspension history, optimal pseudo labels are estimated using a likelihood-based method that takes into account important constraints, which enables more effective use of the information in both failure and suspension histories. The experiments on the C-MAPSS dataset validate the prediction accuracy of the proposed approach and provide several insights.","PeriodicalId":169274,"journal":{"name":"2023 IEEE International Conference on Prognostics and Health Management (ICPHM)","volume":"31 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-06-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130770407","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：