用户名 密 码 登 录 忘记密码 免费注册

网站会员中心 | 设为首页 | 加入收藏

 您好,欢迎光临中国轴承行业网         退出登陆

会员中心 | 设为首页 | 加入收藏

国外文献 > 1.轴承 > 5. 监测与诊断 > 寿命与维护


关键词搜索:


标 题: A Probabilistic Bayesian Transformer for Bearings Remaining Useful Life Prediction
用于轴承剩余使用寿命预测的概率贝叶斯变换器
作 者: Huachao Peng
Bin Jiang
Zehui Mao
关键词: Remaining useful life  [剩余使用寿命]
Bayesian deep learning  [ 贝叶斯深度学习]
Transformer  [ 变压器]
Limited data  [ 数据有限]
Prognostic uncertainty  [ 预后不确定性]
出版年: 2023
文集名: 2023 CAA Symposium on Fault Detection, Supervision and Safety for Technical Processes (SAFEPROCESS 2023)
2023年技术过程故障检测、监督与安全专题研讨会(SAFEPROCESS 2023)
会议名: 2023 CAA Symposium on Fault Detection, Supervision and Safety for Technical Processes (SAFEPROCESS 2023)
2023年技术过程故障检测、监督与安全专题研讨会(SAFEPROCESS 2023)
届 次:
摘 要: Deep-learning (DL) have been widely applied to fault prognostic based on sufficient condition monitoring data. However, when given limited training data, most existing methods with fixed, deterministic parameters fail to address prognostics uncertainty and lead to over-fitting and poor generalization. To overcome these problems, a probabilistic Bayesian Temporal Convolutional Transformer (BTCT)-based fault prognostic framework is developed to RUL prediction. The BTCT model first integrates temporal convolutional network (TCN) into Transformer to extract multiscale features directly from raw monitoring data. Then, the model is integrated into the Bayesian deep learning (BDL) to learn uncertainties and provide RUL prediction confidence intervals (CIs) at certain level. Finally, comparative experiments are carried out on an open-sourced bearings datasets to demonstrate the accuracy and superiority of the proposed approaches under limited data.

深度学习(DL)已被广泛应用于基于充分状态监测数据的故障预测。然而,当训练数据有限时,大多数采用固定、确定性参数的方法无法有效应对预测不确定性,容易导致过拟合和泛化能力差的问题。为克服这些问题,本文提出了一种基于概率贝叶斯时间卷积Transformer(BTCT)的故障预测框架,用于剩余使用寿命(RUL)预测。BTCT模型首先将时间卷积网络(TCN)集成到Transformer中,直接从原始监测数据中提取多尺度特征。随后,该模型与贝叶斯深度学习(BDL)结合,以学习不确定性,并在特定置信水平下提供RUL预测的置信区间(CI)。最后,通过一个开源轴承数据集进行对比实验,验证了所提方法在数据有限条件下的准确性和优越性。
中文内容为网络AI翻译,专业术语未必准确,仅供参考。

VIP网员

联系我们 | 登录注册 | 会员管理 | 网员服务 | 服务信息 | 广告服务 | 高薪聘请 | 法律声明 | 关于我们

京公网安备 11010202009788    京ICP备2021015424号