Prof. Zhang Jingzhong, a well-known Chinese mathematician, is now serving as the honorary director of the Chengdu Institute of Computer Application under the CAS and as an instructor for doctorate students.His outstan...Prof. Zhang Jingzhong, a well-known Chinese mathematician, is now serving as the honorary director of the Chengdu Institute of Computer Application under the CAS and as an instructor for doctorate students.His outstanding attainments in the field of mathematics and computer science include his pioneering work of enriching theories and methods in mechanized demonstration,the concept of "educational mathematics" and his meritorious service as a prolific popular science writer. On December I.1936, Prof.Zhang was born in a teacher’s family in Runan County,Henan Province.After successfully passing an entrance examination in 1954,展开更多
剩余寿命(Remaining Useful Life, RUL)预测对于维护工业设备的可靠性和安全性至关重要,但现有的RUL预测方法在处理高维传感器数据以及捕捉时间退化模式方面仍然面临诸多挑战。为了解决上述问题,提出一种退化趋势平滑约束下基于双向长...剩余寿命(Remaining Useful Life, RUL)预测对于维护工业设备的可靠性和安全性至关重要,但现有的RUL预测方法在处理高维传感器数据以及捕捉时间退化模式方面仍然面临诸多挑战。为了解决上述问题,提出一种退化趋势平滑约束下基于双向长短时记忆网络-变分自编码器(Bidirectional Long Short Term-Memory-Variational Auto Encoder, BLSTM-VAE)的RUL预测方法。该方法首先进行数据预处理,包括数据降噪、滑动窗口分段和标签修正等步骤。然后设计基于BLSTM的VAE型特征提取器,以有效提取时间序列数据中的非线性关系和长距离依赖关系。最后提出一种基于流形学习的退化趋势平滑约束模块,通过局部不变性假设来增强模型的稳健性和泛化能力。通过航空发动机数据集数据集进行验证,结果表明所提出的RUL预测方法在数据集上的表现优于现有的多种RUL预测方法,具有更低的预测误差和更高的稳定性。展开更多
文摘Prof. Zhang Jingzhong, a well-known Chinese mathematician, is now serving as the honorary director of the Chengdu Institute of Computer Application under the CAS and as an instructor for doctorate students.His outstanding attainments in the field of mathematics and computer science include his pioneering work of enriching theories and methods in mechanized demonstration,the concept of "educational mathematics" and his meritorious service as a prolific popular science writer. On December I.1936, Prof.Zhang was born in a teacher’s family in Runan County,Henan Province.After successfully passing an entrance examination in 1954,
文摘剩余寿命(Remaining Useful Life, RUL)预测对于维护工业设备的可靠性和安全性至关重要,但现有的RUL预测方法在处理高维传感器数据以及捕捉时间退化模式方面仍然面临诸多挑战。为了解决上述问题,提出一种退化趋势平滑约束下基于双向长短时记忆网络-变分自编码器(Bidirectional Long Short Term-Memory-Variational Auto Encoder, BLSTM-VAE)的RUL预测方法。该方法首先进行数据预处理,包括数据降噪、滑动窗口分段和标签修正等步骤。然后设计基于BLSTM的VAE型特征提取器,以有效提取时间序列数据中的非线性关系和长距离依赖关系。最后提出一种基于流形学习的退化趋势平滑约束模块,通过局部不变性假设来增强模型的稳健性和泛化能力。通过航空发动机数据集数据集进行验证,结果表明所提出的RUL预测方法在数据集上的表现优于现有的多种RUL预测方法,具有更低的预测误差和更高的稳定性。
