QUALITY SELF-MONITORING OF INTELLIGENT ANALYZERS AND SENSORS BASED ONAN EXTENDED KALMAN FILTER - AN APPLICATION TO GRAPHITE-FURNACE ATOMIC-ABSORPTION SPECTROSCOPY

A method for on-line quality self-monitoring for automatical operating but drifting analytical sensors is presented. The method is based on an on-line state estimation by the Kalman filter extended by quality c ontrol (QC) sampling as known from process monitoring. A linear calibr ation model with linear drift parameters has been chosen. Compared to conventional approaches, the advantage of the proposed method is that it performs simultaneously calibration and recalibration, detection an d correction of drift, and forecasting the expected drift situation, a s well as outlier detection and repair. Compared to the existing Kalma n filter algorithm, the presented one requires a minimal number of QC samples for updating its parameters. Thus, less recalibrations are nec essary in variable time distances adapted to the actual situation in d rift, analytical precision, and accuracy. The new procedure has been v alidated pseudo-on-line in a CF-AAS experiment with artifically enhanc ed drift. Approximately 1000 samples were analyzed using a continuousl y (45 h) running and independent working computer driven graphite furn ace AAS/autosampler setup.