Pub Date : 2023-06-05DOI: 10.1109/ICPHM57936.2023.10194052
G. Murtas, Henrique Cabral, E. Tsiporkova
The formation of ice on the blades of wind turbines can severely affect their power production, lead to a degradation of the assets and even cause safety hazards. Predicting blade icing allows mitigating or preventing altogether its impact on the turbines and their performance by activating blade heating mechanisms. A novel data-driven approach is proposed which estimates a turbine-specific icing risk between 0 and 1 using only meteorological historical and forecasted data. The method is based on the creation of a repository of meteorological profiles characteristics of icing, to which all other profiles are compared in order to compute a similarity score, then converted into an icing risk. The approach is robust against icing sample imbalance in the dataset and thus performant even in locations where icing incidence is extremely low. The icing risk provides wind farm operators with a meaningful indicator, allowing for more flexibility, a better view of the onset of ice formation, and a measure of the severity of an upcoming icing event. The validation is performed on a dataset of 7 turbines belonging to the same wind farm.
{"title":"Data-driven estimation of blade icing risk in wind turbines","authors":"G. Murtas, Henrique Cabral, E. Tsiporkova","doi":"10.1109/ICPHM57936.2023.10194052","DOIUrl":"https://doi.org/10.1109/ICPHM57936.2023.10194052","url":null,"abstract":"The formation of ice on the blades of wind turbines can severely affect their power production, lead to a degradation of the assets and even cause safety hazards. Predicting blade icing allows mitigating or preventing altogether its impact on the turbines and their performance by activating blade heating mechanisms. A novel data-driven approach is proposed which estimates a turbine-specific icing risk between 0 and 1 using only meteorological historical and forecasted data. The method is based on the creation of a repository of meteorological profiles characteristics of icing, to which all other profiles are compared in order to compute a similarity score, then converted into an icing risk. The approach is robust against icing sample imbalance in the dataset and thus performant even in locations where icing incidence is extremely low. The icing risk provides wind farm operators with a meaningful indicator, allowing for more flexibility, a better view of the onset of ice formation, and a measure of the severity of an upcoming icing event. The validation is performed on a dataset of 7 turbines belonging to the same wind farm.","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":"122575562","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/ICPHM57936.2023.10194125
Jie Meng, Jiji Cai, Liang Chang
In satellite operations, one of the essential tasks is to monitor the health status of the systems, which involves forecasting telemetry data that reflects the state of health. The application of data-driven approaches in system monitoring has led to significant improvements in health monitoring and anomaly detection. However, existing methods fail to fully leverage the complex inter-sensor relationships present in satellites. They do not explicitly exploit the structure of these relationships to predict the expected behavior of telemetry time series either. To address these limitations, this paper introduces a novel health monitoring framework for artificial satellites that combines causal graphs and deep learning. In the causality learning phase, we propose a method that integrates mRMR (Maximum Relevance Minimum Redundancy) and PCMCI (Peter-Clark Momentary Conditional Independence) to construct an efficient and accurate causal discovery approach for learning causal graphs for high-dimensional telemetry data. Subsequently, we design a graph attention-based neural network that incorporates these causal graphs into a deep network for prediction. Experimental evaluation on two datasets from satellite attitude control systems and power systems demonstrates the superior performance of our proposed method in accurately predicting health status compared to baseline approaches. Furthermore, the experiments highlight the interpretability-enhancing role of causal graphs, which is beneficial for health monitoring and anomaly detection.
{"title":"A causal graph-based framework for satellite health monitoring","authors":"Jie Meng, Jiji Cai, Liang Chang","doi":"10.1109/ICPHM57936.2023.10194125","DOIUrl":"https://doi.org/10.1109/ICPHM57936.2023.10194125","url":null,"abstract":"In satellite operations, one of the essential tasks is to monitor the health status of the systems, which involves forecasting telemetry data that reflects the state of health. The application of data-driven approaches in system monitoring has led to significant improvements in health monitoring and anomaly detection. However, existing methods fail to fully leverage the complex inter-sensor relationships present in satellites. They do not explicitly exploit the structure of these relationships to predict the expected behavior of telemetry time series either. To address these limitations, this paper introduces a novel health monitoring framework for artificial satellites that combines causal graphs and deep learning. In the causality learning phase, we propose a method that integrates mRMR (Maximum Relevance Minimum Redundancy) and PCMCI (Peter-Clark Momentary Conditional Independence) to construct an efficient and accurate causal discovery approach for learning causal graphs for high-dimensional telemetry data. Subsequently, we design a graph attention-based neural network that incorporates these causal graphs into a deep network for prediction. Experimental evaluation on two datasets from satellite attitude control systems and power systems demonstrates the superior performance of our proposed method in accurately predicting health status compared to baseline approaches. Furthermore, the experiments highlight the interpretability-enhancing role of causal graphs, which is beneficial for health monitoring and anomaly detection.","PeriodicalId":169274,"journal":{"name":"2023 IEEE International Conference on Prognostics and Health Management (ICPHM)","volume":"121 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":"115551252","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/ICPHM57936.2023.10194131
G. Nuzzo, H. Lewitschnig, M. Tuellmann, S. Rzepka, A. Otto
The electric vehicle of the future requires smarter semiconductor power devices to fulfill higher reliability requirements. Several electro-thermal parameters on the chip level can be used to assess the health condition of power electronics systems and to predict the remaining useful life. This paper analyses promising indicators to monitor the degradation level in the chip solder layer of SiC power switches. Active power cycling tests accelerate the aging of a population of SiC power modules for traction inverters. On-state voltage and junction temperature are monitored until the end of life of the devices. The collected data are input to a predictive regression model to estimate the state of health in the power switches. Moreover, a prognostic concept on the system level is introduced. Measurements at operating temperature during the vehicle idle times serve as input to a product-related predictive model. The processor determines the condition of the SiC power switches to issue a maintenance alert and avoid the possible occurrence of unexpected failures. This work provides investigations in data-driven predictive models for wide-bandgap technologies such as SiC power modules and defines an innovative prognostic method on the edge device.
{"title":"A Data-driven Condition Monitoring method to predict the Remaining Useful Life of SiC Power Modules for Traction Inverters","authors":"G. Nuzzo, H. Lewitschnig, M. Tuellmann, S. Rzepka, A. Otto","doi":"10.1109/ICPHM57936.2023.10194131","DOIUrl":"https://doi.org/10.1109/ICPHM57936.2023.10194131","url":null,"abstract":"The electric vehicle of the future requires smarter semiconductor power devices to fulfill higher reliability requirements. Several electro-thermal parameters on the chip level can be used to assess the health condition of power electronics systems and to predict the remaining useful life. This paper analyses promising indicators to monitor the degradation level in the chip solder layer of SiC power switches. Active power cycling tests accelerate the aging of a population of SiC power modules for traction inverters. On-state voltage and junction temperature are monitored until the end of life of the devices. The collected data are input to a predictive regression model to estimate the state of health in the power switches. Moreover, a prognostic concept on the system level is introduced. Measurements at operating temperature during the vehicle idle times serve as input to a product-related predictive model. The processor determines the condition of the SiC power switches to issue a maintenance alert and avoid the possible occurrence of unexpected failures. This work provides investigations in data-driven predictive models for wide-bandgap technologies such as SiC power modules and defines an innovative prognostic method on the edge device.","PeriodicalId":169274,"journal":{"name":"2023 IEEE International Conference on Prognostics and Health Management (ICPHM)","volume":"7 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":"129326381","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/ICPHM57936.2023.10194162
Zhuang Ye, Jianbo Yu, Pu Yang, Shang Yue, Ruixu Zhou, Mingyan Ma
State of health (SOH) estimation is significant to predict the capacity of battery in the battery management systems. The most existing methods require sufficient labeled data to obtain the precise results. However, in the industrial application, it is difficult and costly to collect sufficient battery aging data. Thus, this paper proposed a generative model to tackle the data augmentation and SOH estimation of battery. Firstly, a conditional generative adversarial network is developed for data augmentation. Secondly, a hybrid feature generator, i.e., convolutional long short-term memory (CLSTM) is employed to reconstruct the real signals. Thirdly, a LSTM-based SOH estimator is employed to learn the degradation trance of the original and the artificially generated signals. Finally, a SOH estimation of battery testing is performed to verify the effectiveness of the proposed method. The experimental results indicate that the model can effectively implement data augmentation and SOH estimation of battery.
{"title":"Generative Adversarial Network for State of Health Estimation of Lithium-ion Batteries","authors":"Zhuang Ye, Jianbo Yu, Pu Yang, Shang Yue, Ruixu Zhou, Mingyan Ma","doi":"10.1109/ICPHM57936.2023.10194162","DOIUrl":"https://doi.org/10.1109/ICPHM57936.2023.10194162","url":null,"abstract":"State of health (SOH) estimation is significant to predict the capacity of battery in the battery management systems. The most existing methods require sufficient labeled data to obtain the precise results. However, in the industrial application, it is difficult and costly to collect sufficient battery aging data. Thus, this paper proposed a generative model to tackle the data augmentation and SOH estimation of battery. Firstly, a conditional generative adversarial network is developed for data augmentation. Secondly, a hybrid feature generator, i.e., convolutional long short-term memory (CLSTM) is employed to reconstruct the real signals. Thirdly, a LSTM-based SOH estimator is employed to learn the degradation trance of the original and the artificially generated signals. Finally, a SOH estimation of battery testing is performed to verify the effectiveness of the proposed method. The experimental results indicate that the model can effectively implement data augmentation and SOH estimation of battery.","PeriodicalId":169274,"journal":{"name":"2023 IEEE International Conference on Prognostics and Health Management (ICPHM)","volume":"22 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":"129156852","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/ICPHM57936.2023.10193966
Ao Ding, Yong Qin, Biao Wang, L. Jia
Continual learning is promising in intelligent motor fault diagnosis because it enables networks to increase diagnosable fault classes without time-consuming retraining during new fault happening. However, the traditional continual learning based on knowledge distillation keeps the absolute positions of samples in representation spaces to prevent catastrophic forgetting, which limits new fault samples to embedding into representation spaces flexibly. To address this issue, a continual learning method based on a novel knowledge distillation strategy is proposed for motor fault diagnosis. At incremental stages of continual learning, new and old diagnosis networks are first regarded as the teacher and student networks. Then, the improved distillation strategy is designed to guide knowledge transfer from teacher networks to student networks, meanwhile, student networks learn from the new fault samples. Finally, new diagnosis networks are obtained which can diagnose incremental fault classes. For the improved knowledge distillation strategy, knowledge is inherited by maintaining the proximity behavior of samples in the representation spaces, thereby networks can learn to map samples into representation spaces more flexibly. Through a study case of class-added fault diagnosis of motors, it is proved that the proposed method can improve diagnostic accuracy during continual learning.
{"title":"A Class-Added Continual Learning Method for Motor Fault Diagnosis Based on Knowledge Distillation of Representation Proximity Behavior","authors":"Ao Ding, Yong Qin, Biao Wang, L. Jia","doi":"10.1109/ICPHM57936.2023.10193966","DOIUrl":"https://doi.org/10.1109/ICPHM57936.2023.10193966","url":null,"abstract":"Continual learning is promising in intelligent motor fault diagnosis because it enables networks to increase diagnosable fault classes without time-consuming retraining during new fault happening. However, the traditional continual learning based on knowledge distillation keeps the absolute positions of samples in representation spaces to prevent catastrophic forgetting, which limits new fault samples to embedding into representation spaces flexibly. To address this issue, a continual learning method based on a novel knowledge distillation strategy is proposed for motor fault diagnosis. At incremental stages of continual learning, new and old diagnosis networks are first regarded as the teacher and student networks. Then, the improved distillation strategy is designed to guide knowledge transfer from teacher networks to student networks, meanwhile, student networks learn from the new fault samples. Finally, new diagnosis networks are obtained which can diagnose incremental fault classes. For the improved knowledge distillation strategy, knowledge is inherited by maintaining the proximity behavior of samples in the representation spaces, thereby networks can learn to map samples into representation spaces more flexibly. Through a study case of class-added fault diagnosis of motors, it is proved that the proposed method can improve diagnostic accuracy during continual learning.","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":"126420270","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/ICPHM57936.2023.10194028
Zhonghai Lu, R. Shi, Chao Guo, Mingrui Liu
Like other deep learning problems, critical features are critical to enable effective estimation of Remaining Useful Lifetime (RUL) for power electronic devices using Neural Networks (NNs). However, these critical features are often indirectly obtained after data pre-processing, complicated either in form (high dimension) or in computation (computation-intensive pre-processing). In the paper, we suggest adding a simple direct feature, age, into the NN based RUL estimation technique. The rationale for incorporating this feature is that the device lifetime is a sum of past time (age) plus RUL. Thus it has a strong correlation to RUL. In our experiments using accelerated aging tests, we show that the new age feature enhanced Recurrent Neural Network (RNN) model can significantly improve estimation accuracy and shorten training convergence time. It also outperforms a state-of-the-art RNN model using derived time-domain statistical features.
{"title":"Age Feature Enhanced Neural Network for RUL Estimation of Power Electronic Devices","authors":"Zhonghai Lu, R. Shi, Chao Guo, Mingrui Liu","doi":"10.1109/ICPHM57936.2023.10194028","DOIUrl":"https://doi.org/10.1109/ICPHM57936.2023.10194028","url":null,"abstract":"Like other deep learning problems, critical features are critical to enable effective estimation of Remaining Useful Lifetime (RUL) for power electronic devices using Neural Networks (NNs). However, these critical features are often indirectly obtained after data pre-processing, complicated either in form (high dimension) or in computation (computation-intensive pre-processing). In the paper, we suggest adding a simple direct feature, age, into the NN based RUL estimation technique. The rationale for incorporating this feature is that the device lifetime is a sum of past time (age) plus RUL. Thus it has a strong correlation to RUL. In our experiments using accelerated aging tests, we show that the new age feature enhanced Recurrent Neural Network (RNN) model can significantly improve estimation accuracy and shorten training convergence time. It also outperforms a state-of-the-art RNN model using derived time-domain statistical features.","PeriodicalId":169274,"journal":{"name":"2023 IEEE International Conference on Prognostics and Health Management (ICPHM)","volume":"63 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":"131010443","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/ICPHM57936.2023.10194102
Daniel O. Williams, Z. Li, A. Ghanavati
With the current alarming exponential increase in global energy demand chiefly due to population growth, electrification, and the issues associated with fossil generation, utilities are reinvesting their returns in alternative ways of clean power generation. Although, finding alternative ways to provide clean energy and to advance the power grid are of the main interest globally, many countries face power theft as a frequent problem. in Ghana, power losses in the distribution system cost the nation over a billion Ghana Cedis in the country's total annual revenue, of which power theft plays a predominant role. This paper presents an electricity theft mitigation technique through a programmable smart energy meter. The proposed method is such that interruptions are added to the smart energy meters in order to detect input signals from an added current sensor placed at the terminal point of the service line, from where in-between the sensor and the meter, illegal connections are made. The proposed Advanced Metering Infrastructure (AMI) system will provide smart services, including calculating consumed energy in kWh and generating a bill sent to the utility station. After which, the AMI system will disconnect the power supply from the meter.
{"title":"Mitigating Electrical Losses Through a Programmable Smart Energy Advanced Metering Infrastructure System","authors":"Daniel O. Williams, Z. Li, A. Ghanavati","doi":"10.1109/ICPHM57936.2023.10194102","DOIUrl":"https://doi.org/10.1109/ICPHM57936.2023.10194102","url":null,"abstract":"With the current alarming exponential increase in global energy demand chiefly due to population growth, electrification, and the issues associated with fossil generation, utilities are reinvesting their returns in alternative ways of clean power generation. Although, finding alternative ways to provide clean energy and to advance the power grid are of the main interest globally, many countries face power theft as a frequent problem. in Ghana, power losses in the distribution system cost the nation over a billion Ghana Cedis in the country's total annual revenue, of which power theft plays a predominant role. This paper presents an electricity theft mitigation technique through a programmable smart energy meter. The proposed method is such that interruptions are added to the smart energy meters in order to detect input signals from an added current sensor placed at the terminal point of the service line, from where in-between the sensor and the meter, illegal connections are made. The proposed Advanced Metering Infrastructure (AMI) system will provide smart services, including calculating consumed energy in kWh and generating a bill sent to the utility station. After which, the AMI system will disconnect the power supply from the meter.","PeriodicalId":169274,"journal":{"name":"2023 IEEE International Conference on Prognostics and Health Management (ICPHM)","volume":"105 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":"131338143","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/ICPHM57936.2023.10193968
H. Rasay, Fariba Azizi, Mehrnaz Salmani, F. Naderkhani
This paper focuses on development of joint optimal maintenance and production policy for a specific type of production system that allows for adjustable production rates. The rate of deterioration of the system is directly related to the production rate, with higher production rates resulting in greater expected deterioration. The system's deterioration can be controlled through two main actions: (1) scheduling and conducting maintenance actions referred to as maintenance policy; and (2) adjusting the production rate referred to as production policy. To determine the optimal actions given the system's state, a Markov decision process (MDP) is developed and a reinforcement learning algorithm, specifically a Q-learning algorithm, is utilized. The algorithm's hyper parameters are tuned using a value-iteration algorithm of dynamic programming. The goal is to minimize expected costs for the system over a finite planning horizon.
{"title":"A Reinforcement Learning Algorithm for Optimal Dynamic Policies of Joint Condition-based Maintenance and Condition-based Production","authors":"H. Rasay, Fariba Azizi, Mehrnaz Salmani, F. Naderkhani","doi":"10.1109/ICPHM57936.2023.10193968","DOIUrl":"https://doi.org/10.1109/ICPHM57936.2023.10193968","url":null,"abstract":"This paper focuses on development of joint optimal maintenance and production policy for a specific type of production system that allows for adjustable production rates. The rate of deterioration of the system is directly related to the production rate, with higher production rates resulting in greater expected deterioration. The system's deterioration can be controlled through two main actions: (1) scheduling and conducting maintenance actions referred to as maintenance policy; and (2) adjusting the production rate referred to as production policy. To determine the optimal actions given the system's state, a Markov decision process (MDP) is developed and a reinforcement learning algorithm, specifically a Q-learning algorithm, is utilized. The algorithm's hyper parameters are tuned using a value-iteration algorithm of dynamic programming. The goal is to minimize expected costs for the system over a finite planning horizon.","PeriodicalId":169274,"journal":{"name":"2023 IEEE International Conference on Prognostics and Health Management (ICPHM)","volume":"45 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":"123872917","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/ICPHM57936.2023.10194092
Jianqun Zhang, Qing Zhang, X. Qin, Yuantao Sun
In recent years, the deep learning-based fault diagnosis method has made remarkable achievements, but it is still challenging in the small sample problem. The image texture features of the vibration signal can effectively represent different gearbox states, which is expected to alleviate the dependence on the number of training samples. Therefore, a new time-frequency diagram characterization method based on multi-symplectic geometric modal decomposition (MSGMD) is proposed. Based on the characterization analysis of multi-component simulation signals, it is proved that the MSGMD time-frequency diagram is feasible to characterize signals, and its advantages over other signal decomposition methods. On this basis, a gearbox fault diagnosis method based on MSGMD and convolutional neural network (CNN) is proposed and applied to solve the small sample problem. The experiment results show that the method can achieve more than 95% recognition accuracy even in dealing with small samples (the average number of training samples for each gearbox state is only 22). Compared with other intelligent diagnosis methods, it gets higher recognition accuracy. The above analysis shows that the proposed method is expected to be used in practical engineering gearbox fault diagnosis.
{"title":"2D Characterization Based on MSGMD And Its Application in Gearbox Fault Diagnosis","authors":"Jianqun Zhang, Qing Zhang, X. Qin, Yuantao Sun","doi":"10.1109/ICPHM57936.2023.10194092","DOIUrl":"https://doi.org/10.1109/ICPHM57936.2023.10194092","url":null,"abstract":"In recent years, the deep learning-based fault diagnosis method has made remarkable achievements, but it is still challenging in the small sample problem. The image texture features of the vibration signal can effectively represent different gearbox states, which is expected to alleviate the dependence on the number of training samples. Therefore, a new time-frequency diagram characterization method based on multi-symplectic geometric modal decomposition (MSGMD) is proposed. Based on the characterization analysis of multi-component simulation signals, it is proved that the MSGMD time-frequency diagram is feasible to characterize signals, and its advantages over other signal decomposition methods. On this basis, a gearbox fault diagnosis method based on MSGMD and convolutional neural network (CNN) is proposed and applied to solve the small sample problem. The experiment results show that the method can achieve more than 95% recognition accuracy even in dealing with small samples (the average number of training samples for each gearbox state is only 22). Compared with other intelligent diagnosis methods, it gets higher recognition accuracy. The above analysis shows that the proposed method is expected to be used in practical engineering gearbox fault diagnosis.","PeriodicalId":169274,"journal":{"name":"2023 IEEE International Conference on Prognostics and Health Management (ICPHM)","volume":"20 3 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":"129749075","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 : 2023-05-05DOI: 10.1109/ICPHM57936.2023.10194112
Xian Yeow Lee, Aman Kumar, L. Vidyaratne, Aniruddha Rajendra Rao, Ahmed K. Farahat, Chetan R. Gupta
This paper focuses on solving a fault detection problem using multivariate time series of vibration signals collected from planetary gearboxes in a test rig. Various traditional machine learning and deep learning methods have been proposed for multivariate time-series classification, including distance-based, functional data-oriented, feature-driven, and convolution kernel-based methods. Recent studies have shown using convolution kernel-based methods like ROCKET, and 1D convolutional neural networks with ResNet and FCN, have robust performance for multivariate time-series data classification. We propose an ensemble of three convolution kernel-based methods and show its efficacy on this fault detection problem by outperforming other approaches and achieving an accuracy of more than 98.8%.
{"title":"An ensemble of convolution-based methods for fault detection using vibration signals","authors":"Xian Yeow Lee, Aman Kumar, L. Vidyaratne, Aniruddha Rajendra Rao, Ahmed K. Farahat, Chetan R. Gupta","doi":"10.1109/ICPHM57936.2023.10194112","DOIUrl":"https://doi.org/10.1109/ICPHM57936.2023.10194112","url":null,"abstract":"This paper focuses on solving a fault detection problem using multivariate time series of vibration signals collected from planetary gearboxes in a test rig. Various traditional machine learning and deep learning methods have been proposed for multivariate time-series classification, including distance-based, functional data-oriented, feature-driven, and convolution kernel-based methods. Recent studies have shown using convolution kernel-based methods like ROCKET, and 1D convolutional neural networks with ResNet and FCN, have robust performance for multivariate time-series data classification. We propose an ensemble of three convolution kernel-based methods and show its efficacy on this fault detection problem by outperforming other approaches and achieving an accuracy of more than 98.8%.","PeriodicalId":169274,"journal":{"name":"2023 IEEE International Conference on Prognostics and Health Management (ICPHM)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-05-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131167356","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：