电动汽车增程器耦合故障机理及诊断方法研究

As we all know, extended-range in electric vehicle is the most viable electric vehicle products in development and marketing. However, the extended-range is a coupling device of engine and generator and lacks of supporting and effective fault diagnosis techniques. It has a lot of coupling faults with some single failure in engine or generator. The dynamic feature of extended-range is much complicated with coupling fault and difficult to study with the single failure model. Aiming at the multi-component and multi-fault coupling phenomenon in extended-range, the project will establish the flexible multi-body dynamics model and analyze the dynamic characteristics and failure mechanism with coupling fault. Meanwhile, to solve the nonlinear and coupling problem in the extended-range failure diagnosis, through introducing the framework of support vector machines, the idea of alternating projection, the algorithm of online learning, it proposed a failure prediction model for the diesel engine based on online support tensor machine. At last, it analysis the prediction process and verifies the model accuracy with simulation and experiment. The research results of the project will provide some suggestions for the electric vehicle design, control and maintenance. Moreover, it will provide some theoretical and experimental reference for the OBD development in electric vehicles.

作为目前研发和市场最为可行电动汽车产品，柴油机/发电机型增程式电动汽车的增程器缺乏配套的故障诊断技术。由于增程器是柴油机和发电机的耦合装置，其故障模式除了柴油机和发电机各自的单一故障外，还存在大量的耦合故障。而耦合故障下的增程器动力学特性比单一故障复杂许多，难以用单一故障转子系统建模及诊断方法进行研究。本项目针对增程器中的多部件多故障耦合现象，通过建立增程器的柔性多体动力学模型，分析增程器在耦合故障状态下的动力特性和故障机理；同时，针对耦合故障诊断中的非线性和复杂耦合问题，借鉴支持向量机的学习框架和交替投影的思想，在引入在线学习算法基础上，提出基于在线支持张量机的故障诊断方法，通过仿真和实验验证方法的有效性。项目的研究成果对该类电动汽车的设计、控制及维护具有指导作用，为开发增程式电动汽车OBD系统提供理论依据和实验参考。

作为目前研发和市场最为可行电动汽车产品，柴油机/发电机型增程式电动汽车的增程器缺乏配套的故障诊断技术。由于增程器是柴油机和发电机的耦合装置，其故障模式除了柴油机和发电机各自的单一故障外，还存在大量的耦合故障。而耦合故障下的增程器动力学特性比单一故障复杂许多，难以用单一故障转子系统建模及诊断方法进行研究。本项目针对增程器中的多部件多故障耦合现象，通过建立增程器的柔性多体动力学模型，剖析了增程器在柴油机失火故障、发电机匝间短路故障和退磁故障、轴系不对中故障单一故障及耦合故障的动力学特性、故障机理和演化规律；针对耦合故障诊断中的非线性和复杂耦合问题，结合线性支持高阶张量机的学习框架和在线随机梯度下降法的思想，设计带核函数的快速在线支持张量机算法；搭建了增程式电动汽车故障诊断和健康管理实验平台，以“测试精度”“学习时间”“存储空间”作为评价指标，通过仿真和实验验证了诊断方法的有效性。项目的研究成果对该类电动汽车的设计、控制及维护具有指导作用，为开发增程式电动汽车OBD系统提供理论依据和实验参考。