Accelerating the process of intelligent manufacturing and the demand for new industrial productivity,the operating conditions of machinery and equipment have become ever more severe.As an important link to ensure the ...Accelerating the process of intelligent manufacturing and the demand for new industrial productivity,the operating conditions of machinery and equipment have become ever more severe.As an important link to ensure the stable operation of the production process,the condition monitoring and fault diagnosis of equipment have become equally important.The fault diagnosis of equipment in actual production is often challenged by variable working conditions,large differences in data distribution,and lack of labeled samples,etc.Traditional fault diagnosis methods are often difficult to achieve ideal results in these complex environments.Transfer learning(TL)as an emerging technology can effectively utilize existing knowledge and data to improve the diagnostic performance.Firstly,this paper analyzes the trend of mechanical equipment fault diagnosis and explains the basic concept of TL.Then TL based on parameters,TL based on features,TL based on instances and domain adaptive(DA)methods are summarized and analyzed in terms of existing TL methods.Finally,the problems faced in the current TL research are summarized and the future development trend is pointed out.This review aims to help researchers in related fields understand the latest progress of TL and promote the application and development of TL in mechanical equipment diagnosis.展开更多
To diagnose the aeroengine faults accurately,the supervised Kohonen(S-Kohonen)network is proposed for fault diagnosis.Via adding the output layer behind competitive layer,the network was modified from the unsupervised...To diagnose the aeroengine faults accurately,the supervised Kohonen(S-Kohonen)network is proposed for fault diagnosis.Via adding the output layer behind competitive layer,the network was modified from the unsupervised structure to the supervised structure.Meanwhile,the hybrid particle swarm optimization(H-PSO)was used to optimize the connection weights,after using adaptive inheritance mode(AIM)based on the elite strategy,and adaptive detecting response mechanism(ADRM),HPSO could guide the particles adaptively jumping out of the local solution space,and ensure obtaining the global optimal solution with higher probability.So the optimized S-Kohonen network could overcome the problems of non-identifiability for recognizing the unknown samples,and the non-uniqueness for classification results existing in traditional Kohonen(T-Kohonen)network.The comparison study on the GE90 engine borescope image texture feature recognition is carried out,the research results show that:the optimized S-Kohonen network has a strong ability of practical application in the classification fault diagnosis;the classification accuracy is higher than the common neural network model.展开更多
The power infrastructure of the power system is massive in size and dispersed throughout the system.Therefore,how to protect the information security in the operation and maintenance of power equipment is a difficult ...The power infrastructure of the power system is massive in size and dispersed throughout the system.Therefore,how to protect the information security in the operation and maintenance of power equipment is a difficult problem.This paper proposes an improved time-stamped blockchain technology biometric fuzzy feature for electrical equipment maintenance.Compared with previous blockchain transactions,the time-stamped fuzzy biometric signature proposed in this paper overcomes the difficulty that the key is easy to be stolen by hackers and can protect the security of information during operation and maintenance.Finally,the effectiveness of the proposed method is verified by experiments.展开更多
针对多型传感器采样频率不统一,现有机器学习算法难以有效处理混频数据输入,无法充分挖掘混频信号中的设备故障特征的问题,首先提出一种混频数据输入下的长短时记忆网络(multi-frequency long and short term memory network,MF-LSTM)架...针对多型传感器采样频率不统一,现有机器学习算法难以有效处理混频数据输入,无法充分挖掘混频信号中的设备故障特征的问题,首先提出一种混频数据输入下的长短时记忆网络(multi-frequency long and short term memory network,MF-LSTM)架构;然后,对不同采样频率的状态数据分别进行特征提取并进行特征融合,实现混频数据输入下的电气设备的故障诊断任务;最后,利用凯斯西储大学轴承数据集对所提模型进行了算例验证,结果表明:相比于单频信号输入,混频输入平均提高故障诊断精度1.72%。该实验结果证明了所提出的基于MF-LSTM的故障诊断框架的有效性和混频数据输入的必要性。展开更多
基金National Natural Science Foundation of China(52065030)Key Scientific Research Projects of Yunnan Province(202202AC080008).
文摘Accelerating the process of intelligent manufacturing and the demand for new industrial productivity,the operating conditions of machinery and equipment have become ever more severe.As an important link to ensure the stable operation of the production process,the condition monitoring and fault diagnosis of equipment have become equally important.The fault diagnosis of equipment in actual production is often challenged by variable working conditions,large differences in data distribution,and lack of labeled samples,etc.Traditional fault diagnosis methods are often difficult to achieve ideal results in these complex environments.Transfer learning(TL)as an emerging technology can effectively utilize existing knowledge and data to improve the diagnostic performance.Firstly,this paper analyzes the trend of mechanical equipment fault diagnosis and explains the basic concept of TL.Then TL based on parameters,TL based on features,TL based on instances and domain adaptive(DA)methods are summarized and analyzed in terms of existing TL methods.Finally,the problems faced in the current TL research are summarized and the future development trend is pointed out.This review aims to help researchers in related fields understand the latest progress of TL and promote the application and development of TL in mechanical equipment diagnosis.
基金Joint Funds of the National Natural Science Foundation of China(NSAF)(No.U1330130)General Program of Civil Aviation Flight University of China(No.J2015-39)
文摘To diagnose the aeroengine faults accurately,the supervised Kohonen(S-Kohonen)network is proposed for fault diagnosis.Via adding the output layer behind competitive layer,the network was modified from the unsupervised structure to the supervised structure.Meanwhile,the hybrid particle swarm optimization(H-PSO)was used to optimize the connection weights,after using adaptive inheritance mode(AIM)based on the elite strategy,and adaptive detecting response mechanism(ADRM),HPSO could guide the particles adaptively jumping out of the local solution space,and ensure obtaining the global optimal solution with higher probability.So the optimized S-Kohonen network could overcome the problems of non-identifiability for recognizing the unknown samples,and the non-uniqueness for classification results existing in traditional Kohonen(T-Kohonen)network.The comparison study on the GE90 engine borescope image texture feature recognition is carried out,the research results show that:the optimized S-Kohonen network has a strong ability of practical application in the classification fault diagnosis;the classification accuracy is higher than the common neural network model.
基金This research was funded by science and technology project of State Grid JiangSu Electric Power Co.,Ltd.(Research on Key Technologies of power network security digital identity authentication and management and control based on blockchain,Grant No.is J2021021).
文摘The power infrastructure of the power system is massive in size and dispersed throughout the system.Therefore,how to protect the information security in the operation and maintenance of power equipment is a difficult problem.This paper proposes an improved time-stamped blockchain technology biometric fuzzy feature for electrical equipment maintenance.Compared with previous blockchain transactions,the time-stamped fuzzy biometric signature proposed in this paper overcomes the difficulty that the key is easy to be stolen by hackers and can protect the security of information during operation and maintenance.Finally,the effectiveness of the proposed method is verified by experiments.
文摘针对多型传感器采样频率不统一,现有机器学习算法难以有效处理混频数据输入,无法充分挖掘混频信号中的设备故障特征的问题,首先提出一种混频数据输入下的长短时记忆网络(multi-frequency long and short term memory network,MF-LSTM)架构;然后,对不同采样频率的状态数据分别进行特征提取并进行特征融合,实现混频数据输入下的电气设备的故障诊断任务;最后,利用凯斯西储大学轴承数据集对所提模型进行了算例验证,结果表明:相比于单频信号输入,混频输入平均提高故障诊断精度1.72%。该实验结果证明了所提出的基于MF-LSTM的故障诊断框架的有效性和混频数据输入的必要性。
