基于退化协变量的非连续工作产品寿命预测理论

Power battery is the key component of electric vehicle. The accuracy prediction of service life for the battery will directly affect the cost and performance of electric vehicle. There appear two distinct characteristics of power battery. (1)Discontinuous working, that is, the state of storage and use appears alternatively during the service period. (2)There exist multiple covariates, such as temperature, change of speed, etc. that have close relate with performance degradation and can be monitored directly. Moreover, the covariates change randomly during battery service process. This causes that the reliability theory and life prediction methods based on assumptions of continuous-single state and certain environmental profile cannot be used directly for the life prediction of discontinuous working products similar to power battery. Take electric automobile power battery as background, this project will do research works on service life prediction theory for discontinuous working products based on performance degradation covariates. Firstly, time-varying stochastic process model is established for environmental covariates which are continuous or shock according to monitoring condition data of similar products. Then, multiple stress accelerated degradation tests are designed according to the characteristics of environmental covariates. Degradation rate models with covariates are proposed for discontinuous working products based on multiple degradation mechanisms and proportional hazards assumptions. Based on the performance degradation theory, service life indexes of service life in changing environments for population, as well as indexes of residual life for individual online, are estimated. Finally, applicability and effectiveness of above theory are demonstrated through life prediction of such typical discontinuous working products as power battery, high performance capacitor, etc.

动力电池是电动汽车的关键部件，其服役寿命预测的精确性，将直接影响电动汽车成本和汽车使用性能的发挥。动力电池在应用中具有两个明显特征：(1)非连续工作。贮存和使用状态交替出现；(2)存在多个与退化相关可直接监测的协变量(温度、车速变化等)，协变量在电池服役期间随机变化。立足于产品单一状态假设及确定环境剖面的寿命预测方法，难以直接应用于动力电池等非连续工作产品。以建立基于退化协变量的非连续工作产品寿命预测理论为目标，首先根据相似产品状态监测数据建立环境协变量(连续、冲击型)时变随机模型；然后根据环境协变量特性设计多应力加速退化寿命测试试验；基于多机理相关和比例风险假设建立非连续工作产品含协变量退化率模型；再由性能退化理论，实现批产品变环境下服役寿命预测及在役个体产品剩余寿命预测；并针对动力电池、高性能电容等典型非连续工作产品开展实例研究，验证上述理论方法的适用性和有效性。

动力电池是电动汽车的关键部件，其服役寿命预测的精确性，将直接影响电动汽车成本和汽车使用性能的发挥。动力电池在应用中具有两个明显特征：（1）非连续工作。贮存和使用状态交替出现；（2）存在多个与退化相关可直接监测的协变量（温度、车速变化等），协变量在电池服役期间随机变量。这两个特点使得立足于产品单一状态假设及确定环境剖面寿命预测方法，难以直接应用于动力电池等非连续工作产品。以建立基于性能退化协变量的非连续工作产品服役寿命预测理论为目标，首先根据相似产品状态监测数据建立环境协变量（连续、冲击型）时变随机模型；然后根据环境协变量特性设计多应力加速退化试验；同时，基于多机理相关和比例风险假设建立非连续工作产品含协变量退化率模型；再利用性能退化理论，实现批产品变环境下服役寿命预测及在役个体剩余寿命预测；针对动力电池等典型非连续工作产品开展了实例研究，验证了上述理论方法的适用性和有效性。. 项目成果的应用前景主要体现在两个方面：. （1）针对电动汽车锂离子电池特有的非连续工作机制和失效机理，及关键性能参数（容量和内阻）难以在线监测的难题，提出了基于二元协变量的退化建模方法。两个协变量分别为：恒压充电时间和恒流充电时间，并利用NASA锂离子电池试验数据验证了方法的精度和有效性。. （2）通过动量轮轴承、柴油机主轴、伺服机构等一系列工程实例，验证了协变量退化建模的有效性。对于工程实践中大量存在的性能参数难以直接监测的退化失效产品，项目组提出的间接监测的协变量方法及在线更新规则，可实现复杂系统高精度的在线剩余寿命监测。