Exploitation of multi-models identification with decoupled states in twin shaft gas turbine variables for its diagnosis based on parity space approach

In practice, model-based fault diagnosis methods are essential to improve availability with reduced operating costs and good operational reliability of industrial systems. This is based firstly on the choice of the model identification method adapted to the system, depending on the complexity of the system and its interaction with its environment. Then, on the choice of an adequate diagnostic strategy for the generation of system failure indicators. In this work, the identification problem of the model variables of a double-shaft gas turbine is treated, to deal with the dynamics of model nonlinearities of this rotating machine. Hence, the equations which govern this turbine are carried out, using the local multi-models’ techniques with decoupled states, from the input/output measurements collected on the examined turbine. To best characterize their dynamic behavior in diverse operating areas. Subsequently, the resulting multi-model decoupled states are used to develop a fault diagnosis approach for this turbine. This makes it possible to generate symptoms of turbine failure from consistency tests between the measurements extracted on its real behavior, and the estimated signals which translate the reference behavior, given by the obtained multi-models. The obtained results in this work show the implementation efficiency of the proposed techniques of modeling and estimation of the examined decoupled turbine states, up to the phase of its implementation in the diagnostic strategy of the examined turbine based on the parity space approach.