On-line forgetting factor adaptation for parameter estimation based diagnosis

摘要

This paper presents a fault detection method based on a classical transfer function parameter estimation algorithm in the discrete time domain. Non persistently exciting inputs plant an important problem for the convergence of the estimator. Here, the forgetting factor is adapted on-line in order to improve the convergence. Redundant discrete time transfer functions are used to improve the isolation capacity and obtain a signature table. The fault detection and isolation (FDI) is achieved by the exploitation of this table, with a distance computation.