Sensor Incipient Fault Detection and Isolation Method of Hybrid System

DONG Xiao-liang，PAZILAI Mahemuti
（College of Electrical Engineering，Xinjiang University，Urumqi 830049，China）
Abstract：Taking into account the impact of modeling errors and measurement noise were unavoidable in the actual production process，the paper proposed the sensor incipient fault detection and isolation method combined with the global analytical redundancy relation，exponentially weighted moving average filter technology and adaptive threshold evaluation method.The paper analyzed the structure and characteristics of three tank hybrid system and introduced its incipient faults which may occur.In order to realize the sensor incipient fault detection and isolation，the global analytical redundancy relation method of hybrid bond graph model was utilized to generate residuals，and evaluated it by EWMA filter technology and adaptive threshold method.Finally 20?鄄sim and matlab were used to simulate and verify the feasibility of this method in sensor incipient fault detection and isolation.
Key words：three?鄄tank；sensor incipient fault；hybrid bond graph；global analytical redundancy relation；exponentionally weighted moving average（EWMA）