基于不确定油液光谱数据的综合传动装置剩余寿命预测

原子发射光谱分析得到的磨损微粒元素浓度是综合传动装置性能劣化评估和剩余寿命预测的重要监测指标。由于系统随机劣化过程和光谱测量误差的影响，油液光谱数据中不可避免包含系统劣化随机性和光谱测量不确定性。然而，现有基于油液光谱数据的剩余寿命预测研究中，没有考虑劣化过程的随机性和测量的不确定性对剩余寿命预测的影响。因此，针对综合传动装置劣化随机性和油液光谱数据测量不确定性对寿命预测的影响，提出一种考虑系统随机劣化和数据不确定测量的综合传动装置劣化过程建模方法。基于随机过程首中时间的概念，定义了综合传动装置的剩余寿命；基于Wiener随机过程，建立了考虑系统随机劣化和不确定测量数据的综合传动装置劣化模型，利用极大似然估计方法，估计了劣化过程模型的参数；利用卡尔曼滤波技术，实现了综合传动装置劣化状态的实时估计与更新，进一步得到了考虑系统劣化随机性和光谱数据测量不确定性的剩余寿命分布。研究结果表明，提出的劣化建模方法能够准确估计装置的运行状态，避免了采用条件维护时间对装置进行维护与保养的局限性；综合传动装置的维护时间预测值比条件维护时间延长了193 Mh（113.5%）；考虑光谱数据测量不确定性的剩余寿命预测方法优于不考虑测量不确定性的方法。

Remaining Useful Life Prediction of Power-Shift Steering Transmission Based on Uncertain Oil Spectral Data

The oil spectral data are introduced to indicate the performance degradation and the remaining useful life(RUL) prediction in the reliability evaluation of Power shift steering transmission(PSST). Because of the PSST’s stochastic degradation and spectral measurement error, the measured data inevitably contain the stochasticity of the degradation and the uncertainty of spectrum measurement. However, in current studies of RUL prediction based on oil spectral data, no one has been reported to consider the effect of degradation stochasticity and measurement uncertainty on the prediction of RUL. Thus, aimed at reducing the adverse impact of measurement error of oil spectrum data on the remaining useful life(RUL) prediction of Power shift steering transmission(PSST), a degradation modeling method considering degradation stochasticity and measurement uncertainty is proposed. The concept of RUL of PSST is defined based on the concept of first hit time(FHT) of stochastic process. The parameters of the degradation model are estimated using the maximum likelihood method. The degradation state of the PSST is estimated and updated in real time using Kalman filtering technique, and the RUL distributions considering the system degradation stochasticity and spectral data measurement uncertainty are obtained. The experimental results show that the degradation modeling method proposed in this paper can accurately estimate the running state of the device and avoid the limitation of using the condition maintenance time to maintain the equipment. The time interval of condition-based maintenance has extended as 193 Mh (113.5%), and the RUL prediction method considering uncertain measurements is superior to the method without considering.