基于观测器的传感器故障检测与信号重构方法

针对含未知有界干扰输入的线性系统进行传感器故障检测研究.通过对输出信号进行非奇异线性变换，将传感器故障等效转化为执行机构故障，提出对干扰具有鲁棒性、对故障具有敏感性的残差生成方法，设计滑模观测器进行残差估计，实现对传感器故障的检测，得到对传感器故障信号的精确估计，并实现对传感器信号的重构.基于飞机纵向小扰动模型，对于传感器的缓变偏差、突变偏差、完全失效和反相等故障的信号重构仿真结果都表明了该方法的有效性.

Observer-based sensor fault detection and signal reconstruction method

The sensor-fault detection for linear system with bounded unknown input disturbance was studied. By non-singular linear transformation, the sensor-faults were converted equivalently into actuator failures. The residual generation was proposed which is robust to disturbance and sensitive to sensor faults. A sliding mode observer was further designed to give the estimation of residuals. The sensor faults can be detected and the sensor signals can be reconstructed by using the estimation of residuals. For typical sensor failure modes (slow bias fault and step bias fault, complete invalidation and reverse fault), the simulation results on the aircraft longitudinal model show the effectiveness of the proposed methods.