Abstract:When the system is working under nonperiodic nonsteady conditions, the traditional straight line limit is hard to detect the faults. For these problems above, this paper presented a fault detection model based on the dynamic limit. According to the need for realtime fault detection, the process data is used to segment handling. And the paper proposes the dynamic limit to substitute the straight limit line. Then slid on the test data transformation, result in subsliding line. And then achieve fault detection by determining the subtraverse line and the dynamic limit. At last, the feasibility analysis is given to the model. The experiment results verify that the model is applicable to the fault detection under the nonperiodic nonsteady conditions even if the system’s multivariate outputs are mutated and the amplitude of Hotelling T2 changes greatly. Compared with traditional methods, the model greatly reduces the system’s leakage alarm and false alarm.

Key words:T2-chart,dynamic limit,non-steady conditions,fault detection