将未知输入卡尔曼滤波器扩展到非线性系统,并将其应用于检测航天器飞轮出现的早期渐变型故障,以提高故障检测的及时性,避免重大事故的发生。通过构造一组结构化残差实现故障分离,采用Wald序列检测方法处理滤波残差,最后给出故障判断逻辑。仿真结果验证了结论的有效性。 The unknown input Kalman filter is extended to non-linear system and used to detect the early gradual failure of spacecraft flywheel to improve the timeliness of fault detection and avoid the occurrence of major accidents. By constructing a set of structured residuals to realize the fault separation, the Wald sequence detection method is used to process the filtering residuals, and finally the fault judgment logic is given. Simulation results verify the validity of the conclusion.