摘要:This paper deals with the problem of detection and identification of sequential leaks which are non-strong isolable. To solve the problem in the framework of model-based procedures, a nonlinear input-output model with five state is proposed which is equivalent to the fluid model with multiple faults in steady-state. This input-output model has the property to retain the data of the past leaks allowing only the identification on-line of the new event during a time window for any arbitrary number of leaks. Specifically, two sets of parameters characterize each event in the model: one set associated with the past events in the equivalent leak's model and the second one which contains the latest event parameters. Thus, an iterative Fault Detection and Isolation scheme is suggested which identifies each time a new leak is present its outflow an position, together with the equivalent model by using adaptive Extended Kalman Filters. The simulation for a sequential set of three leaks shows successful results.
关键词:Isolability of sequential faultsequivalent models with faultsholding data for leaks’ detection