• 主题
高级搜索  >
历史搜索 清空历史
首页> 外文期刊>Mechanical systems and signal processing >Remaining lifetime modeling using State-of-Health estimation
【24h】

Remaining lifetime modeling using State-of-Health estimation

机译：使用健康状况估算的剩余寿命建模

获取原文
获取原文并翻译 | 示例
           
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

Technical systems and system's components undergo gradual degradation over time. Continuous degradation occurred in system is reflected in decreased system's reliability and unavoidably lead to a system failure. Therefore, continuous evaluation of State-of-Health (SoH) is inevitable to provide at least predefined lifetime of the system defined by manufacturer, or even better, to extend the lifetime given by manufacturer. However, precondition for lifetime extension is accurate estimation of SoH as well as the estimation and prediction of Remaining Useful Lifetime (RUL). For this purpose, lifetime models describing the relation between system/component degradation and consumed lifetime have to be established. In this contribution modeling and selection of suitable lifetime models from database based on current SoH conditions are discussed. Main contribution of this paper is the development of new modeling strategies capable to describe complex relations between measurable system variables, related system degradation, and RUL. Two approaches with accompanying advantages and disadvantages are introduced and compared. Both approaches are capable to model stochastic aging processes of a system by simultaneous adaption of RUL models to current SoH. The first approach requires a priori knowledge about aging processes in the system and accurate estimation of SoH. An estimation of SoH here is conditioned by tracking actual accumulated damage into the system, so that particular model parameters are defined according to a priori known assumptions about system's aging. Prediction accuracy in this case is highly dependent on accurate estimation of SoH but includes high number of degrees of freedom. The second approach in this contribution does not require a priori knowledge about system's aging as particular model parameters are defined in accordance to multi-objective optimization procedure. Prediction accuracy of this model does not highly depend on estimated SoH. This model has lower degrees of freedom. Both approaches rely on previously developed lifetime models each of them corresponding to predefined SoH. Concerning first approach, model selection is aided by state-machine-based algorithm. In the second approach, model selection conditioned by tracking an exceedance of predefined thresholds is concerned. The approach is applied to data generated from tribological systems. By calculating Root Squared Error (RSE), Mean Squared Error {MSE), and Absolute Error (ABE) the accuracy of proposed models/approaches is discussed along with related advantages and disadvantages. Verification of the approach is done using cross-fold validation, exchanging training and test data. It can be stated that the newly introduced approach based on data (denoted as data-based or data-driven) parametric models can be easily established providing detailed information about remaining useful/consumed lifetime valid for systems with constant load but stochastically occurred damage.
机译：随着时间的推移，技术系统和系统组件会逐渐退化。系统中发生的持续降级反映在系统可靠性下降中，不可避免地导致系统故障。因此，不可避免地要对健康状态（SoH）进行连续评估，以至少提供制造商定义的系统的预定义寿命，甚至更好地延长制造商给出的寿命。但是，延长寿命的前提是准确估算SoH以及估算和预测剩余可用寿命（RUL）。为此，必须建立描述系统/组件降级与消耗的寿命之间关系的寿命模型。在这一贡献中，讨论了基于当前SoH条件从数据库中建模和选择合适的寿命模型。本文的主要贡献是开发了新的建模策略，该策略能够描述可测系统变量，相关系统降级和RUL之间的复杂关系。介绍并比较了两种具有优点和缺点的方法。通过将RUL模型同时适应当前的SoH，这两种方法都能够对系统的随机老化过程进行建模。第一种方法需要有关系统中老化过程的先验知识以及对SoH的准确估算。在这里，SoH的估算是通过跟踪系统中实际累积的损坏来确定的，以便根据有关系统老化的先验已知假设来定义特定的模型参数。在这种情况下，预测准确度高度依赖于SoH的准确估计，但包括许多自由度。此贡献中的第二种方法不需要有关系统老化的先验知识，因为根据多目标优化过程定义了特定的模型参数。该模型的预测准确性并不高度依赖于估算的SoH。此模型的自由度较低。两种方法都依赖于先前开发的寿命模型，它们各自对应于预定义的SoH。关于第一种方法，模型选择是基于状态机的算法来辅助的。在第二种方法中，关注通过跟踪预定义阈值的超出来进行模型选择。该方法适用于从摩擦学系统生成的数据。通过计算均方根误差（RSE），均方根误差（MSE）和绝对误差（ABE），讨论了所提出模型/方法的准确性以及相关的优缺点。使用交叉验证，交换培训和测试数据来验证方法。可以说，可以轻松建立新的基于数据（表示为基于数据或数据驱动的）参数模型的方法，提供有关对于具有恒定负载但随机发生损坏的系统有效的剩余可用/消耗寿命的详细信息。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱：kefu@zhangqiaokeyan.com

京公网安备：11010802029741号 ICP备案号：京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有

  • 客服微信

  • 服务号