论文部分内容阅读
针对卫星姿控系统故障具有并发性和多发性的特点,敏感器和执行机构均可能发生故障,依据相关向量机(RVM)回归理论,采用基于模型辩识,残差评价的方法实现姿控系统多故障检测。通过对太阳敏感器、陀螺和反作用轮的历史输入输出数据建立RVM回归模型,考虑到建模精确度直接影响到检测精确度,对比分析最小二乘支持向量机回归(LSSVR)的回归模型,并给出了二者辩识精确度对比结果。对比结果表明,RVM较LSSVR具有较好的建模精确度。将RVM回归模型应用于太阳敏感器、陀螺和反作用轮的单一故障和多故障检测过程中,仿真结果表明,RVM回归能有效实现姿控系统的多故障检测。
In view of the concurrency and multiplicity of satellite attitude control system faults, the sensors and actuators may both fail. According to the regression vector theory (RVM), the attitude control system is implemented based on the model identification and residual evaluation Multiple fault detection. Based on the historical input and output data of sun sensor, gyroscope and reaction wheel, the RVM regression model is established. Considering that the accuracy of the model directly affects the detection accuracy, the regression model of least square support vector regression (LSSVR) is compared and analyzed. The result of the comparison between the two is given. The comparison results show that RVM has better modeling accuracy than LSSVR. The RVM regression model is applied to single fault and multiple fault detection of sun sensor, gyro and reaction wheel. Simulation results show that RVM regression can effectively detect multi-fault in attitude control system.