Model-based fault detection and identification with online aerodynamic model structure selection

Abstract

This publication describes a recursive algorithm for the approximation of time-varying nonlinear aerodynamic models by means of a joint adaptive selection of the model structure and parameter estimation. This procedure is called adaptive recursive orthogonal least squares (AROLS) and is an extension and modi¦cation of the previously developed ROLS procedure. This algorithm is particularly useful for model-based fault detection and identi¦cation (FDI) of aerospace systems. After the failure, a completely new aerodynamic model can be elaborated recursively with respect to structure as well as parameter values. The performance of the identi¦cation algorithm is demonstrated on a simulation data set.