Gas Turbine Model-Based Robust Fault Detection Using a Forward – Backward Test

Abstract

The problem of robust model based fault detection of dynamic systems using interval observers has been mainly addressed checking if the measured output is inside the interval of possible estimated outputs obtained considering uncertainty on model parameters. This task can be computationally expensive because the interval observers can be affected by the wrapping effect. In this paper, a mixed approach consisting in determining a computationally cheaper inner approximation of the estimated output interval, based only on simulating vertices of parameter uncertainty region (forward test), is combined with a backward consistency check when the real measured output falls outside this inner solution (backward check). The backward check is implemented using interval constraint satisfaction algorithms which can perform efficiently in deciding if the measured output is consistent with the interval model. The classical alternative to this backward check will force to solve a global optimisation problem, or equivalently, a global consistency problem. Finally, this approach will be tested on a gas turbine nozzle servosystem.