【摘要】 随着预测与健康管理技术需求日益迫切，基于设备润滑油信息的剩余有效寿命预测(RUL)成为制定预防性维护策略的关键技术。然而，受限于油液监测数据稀疏性和多指标综合表征复杂性，现有油液剩余有效寿命预测模型精度难以满足实时应用需求。为了提升油液剩余有效寿命预测准确性，提出了一种基于D-S证据理论的多指标融合的油液剩余有效寿命预测模型。首先，考虑到油液状态监测数据多样性，采用模糊概率化描述润滑油多指标，加权获取隶属于同一特征下的多属性信息，并通过D-S证据理论融合多指标信息，获得油液状态综合的量化描述。其次，为了获取润滑油多指标表征下的剩余有效寿命，建立维纳随机过程描述润滑油的衰退过程，基于实时监测数据采用期望最大算法(EM)算法更新和校准模型参数，求解出油液寿命的概率密度函数。这些策略旨在最大限度地减少多属性数据可能带来的冲突，同时在资源与时间受限条件下，有效提升设备和系统的可靠性。最终，通过实际工程应用案例验证了该模型的有效性和实用性，并对模型相关评价指标进行了深入分析。更多还原

【Abstract】 With the increasingly urgent demand for predictive and health management technologies, remaining effective life(RUL) prediction based on equipment lubricant information has become a key technology for developing preventive maintenance strategies. However, due to the sparseness of lubricant monitoring data and the complexity of comprehensive characterization of multiple indicators, the accuracy of the existing RUL prediction models is difficult to meet the real-time application requirements. In order to improve the accuracy of the remaining effective life prediction of fluids, a multi-indicator fusion of fluid remaining effective life prediction model based on the Dempster-Shafer(D-S) evidence theory is proposed. First, considering the diversity of fluid condition monitoring data, fuzzy probabilistic description of lubricant multi-indicators is adopted to weight the multi-attribute information belonging to the same feature, and the multi-indicator information is fused by the D-S evidence theory in order to obtain a comprehensive quantitative description of the fluid condition. Second, in order to obtain the remaining effective life under the lubricant multicriteria characterization, a Wiener stochastic process is established to describe the lubricant’s decline process, and the probability density function of the oil life is solved based on the real-time monitoring data by updating and calibrating the model parameters using the expectation maximization(EM) algorithm. These strategies aim to minimize the conflicts that may be caused by multi-attribute data, and at the same time effectively improve equipment and system reliability under resource and time constraints. Ultimately, the model validity and practicality are verified through the actual engineering application cases, and the model-related evaluation indexes are analyzed in depth.更多还原