A novel affine qLPV model derivation method for fault diagnosis H∞ performance improvement

Abstract

In this paper, a methodology for an affine quasilinear parameter varying (qLPV) model derivation is proposed. The nonlinear model of the system is converted into a qLPV model by hiding the nonlinearities in the scheduling parameters. In order to select the most suitable model among all the possible models, an algorithm is introduced and proposed to generate affine qLPV models for enhancing the fault diagnosis performance as measured in terms of the fault estimation accuracy. The fault diagnosis is accomplished by an HA novel affine qLPV model derivation method for fault diagnosis H∞ performance improvement filter in a linear fractional transformation structure that is designed in the LMI framework. To assess the performance of the proposed approach, our scheme is applied to a gas turbine model. A number of different model structures are considered to design the fault diagnosis filter and performance comparisons conducted show the advantages of the proposed model derivation scheme. 2015 American Automatic Control Council.