Modelling of uncertainties for robust fault diagnosis

Modeling uncertainty can be described as an additional term in the dynamic equation of a system that has a certain structure. Based on this structure, a robust fault diagnosis scheme can be designed. The authors present new methods for determining the structure of modeling uncertainties based on deconvolution of unknown inputs. Several theoretical conditions are discussed, and a jet engine system is used to illustrate the method. Simulation results show the power of this method for modeling uncertainties for the purpose of robust fault diagnosis.<<ETX>>