SOME NOVEL METHODS BASED ON RECURSIVE OPTIMAL ESTIMATION - APPLICATIONS TO ANALYTICAL-CHEMISTRY

Two novel methods based on recursive optimal estimation were approache d: fading Kalman filtering (FKF) and networked Kalman filtering (NFK). Fading Kalman filtering was employed to enhance overlapped spectrosco pic resolution. Based on the nature of the Kalman filter, that the res idual sequence is uncorrelated when the optimal gain is obtained, a ne w fading optimal adaptive algorithm is proposed and utilized. By on-li ne adjustment of the fading factor, the convergency and optimality of the Kalman filter were improved using measured outputs or estimated re sults, even in the presence of model errors and/or the effects of unme asurable external disturbances. The FKF algorithm developed was applie d to overlapped peak resolutions and gave good results. The networked Kalman filtering (NKF) of multiple models was found to have some advan tages over the ordinary Kalman filtering and was used to detect concen trations of minor impurities of less than 1% of that of the major anal yte.