基于退化数据的产品剩余寿命预测方法研究

发布时间：2018-06-07 14:13

本文选题：剩余寿命预测 + 不确定性　；参考：《西安理工大学》2017年硕士论文

【摘要】：预测与健康管理技术(Prognostics and Health Management,PHM)是一种新型的装备综合保障技术,可以大大提高系统设备的可靠性和安全性,还能降低复杂系统设备的维护费用。预测与健康管理技术包括故障预测(Prognostics)和健康管理(Management)两个部分。其中,所谓故障预测就是根据系统现在或者历史的性能状态预测部件未来的健康状态,比如,确定设备的剩余寿命或者正常工作的时间长度。因此,剩余寿命(Remaining Useful Life, RUL)预测技术是预测与健康管理技术中一项核心问题。准确预测随机退化产品的剩余寿命是进行预测与健康管理的基础,也是工程实践中的重难点问题。对于此类复杂产品,随机多变的退化演变规律一般难以机理建模,而传统的基于寿命数据的方法对于小样本、高成本的设备则难以实施。因此,基于状态监测数据进行退化建模和剩余寿命预测,进而实现管理决策的技术成为了当前可靠性工程领域的研究前沿。本文主要基于退化数据的可靠性建模对剩余寿命预测方法进行了以下两个方面的研究:1、针对线性模型,本文在首达时间的概念下,提出一种同时考虑参数不确定性和测量不确定性的Wiener过程退化模型,并推导了考虑了含参数噪声和测量误差的wiener退化设备剩余寿命概率密度函数的解析解。同时实现了在线剩余寿命预测。并用蒙特卡洛仿真验证了本文方法的有效性,最后激光管的实验结果表明本文提出的方法能显著提高剩余寿命预测的精度。2、针对非线性模型,本文基于马里兰大学锂离子电池循环寿命退化数据,对锂离子电池的寿命退化过程进行分析并选择经验退化模型,提出一种基于EKF/KF算法的离子电池剩余寿命预测方法。使用EKF算法对历史数据进行参数估计,然后利用估计的参数基于KF算法来对锂离子电池剩余寿命进行估计,利用马里兰大学的锂离子电池数据验证算法的有效性,并用MAE指标对算法进行评价。
[Abstract]:Prognostics and Health Management is a new integrated equipment support technology, which can greatly improve the reliability and security of system equipment and reduce the maintenance cost of complex system equipment. Prediction and health management technology include two parts: fault prediction (Prognostics) and health management (management). The so-called fault prediction is to predict the future health state of the components according to the current or historical performance of the system, for example, to determine the remaining life of the equipment or the length of the normal working time. Therefore, residual life Useful Life, RUL) prediction technology is a core problem in prediction and health management technology. Accurate prediction of residual life of randomly degraded products is the basis of prediction and health management, and is also a difficult problem in engineering practice. For this kind of complex products, it is difficult to model the mechanism of random and changeable degradation evolution law, but the traditional method based on life data is difficult to implement for small sample and high cost equipment. Therefore, the technology of modeling degradation and predicting residual life based on state monitoring data has become the research frontier in the field of reliability engineering. In this paper, based on the reliability modeling of degenerate data, the residual life prediction method is studied in the following two aspects: 1. For the linear model, this paper is based on the concept of first arrival time. A degenerate model of Wiener process considering both parameter uncertainty and measurement uncertainty is proposed, and the analytical solution of residual life probability density function of wiener degenerate equipment with parameter noise and measurement error is derived. At the same time, the online residual life prediction is realized. The effectiveness of the proposed method is verified by Monte Carlo simulation. The experimental results of the laser tube show that the proposed method can significantly improve the accuracy of residual life prediction. Based on the cyclic life degradation data of Li-ion batteries at the University of Maryland, this paper analyzes the degradation process of Li-ion batteries and selects an empirical degradation model. A method for predicting the residual life of Li-ion batteries based on EKF/KF algorithm is proposed. The EKF algorithm is used to estimate the parameters of the historical data, then the estimated parameters are used to estimate the residual life of the lithium ion battery based on KF algorithm. The validity of the algorithm is verified by the lithium ion battery data of the University of Maryland. The algorithm is evaluated with MAE index.
【学位授予单位】：西安理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TM912;O213.2

【参考文献】