Abstract:

Considering both unknown inputs and modeling errors of system models, a design method for robust fault detection observers is presented. Assuming that there are no modeling errors and in order to reduce the design conservativeness, the H- index and the H∞ norm, which measure the sensitivity of the residual to the fault and the robustness of the effect of the unknown inputs on the residual respectively, as well as a new iterative linear matrix inequality(LMI) algorithm are used to deal with the H∞/H- optimization problem. The reference residual model and the adaptive threshold generators are also designed. Then the robust fault detection observer design as a standard H∞ model matching problem is presented via an LMI formulation to decrease the rate of false alarming and miss probability. Simulation shows that the designed fault detection observer has high sensitivity to faults and strong robustness to uncertain factors.