00063 –51a85b9bcc00dad5aed8c89c2cc901

62

Hembree & Zimmer

and therefore the estimate tracks the random walk component of the process sequcnce quite well.

Adaptive Kalman Filtering

In the examples in the previous section, we assumed that we had perfect information on the various process model parameters. In practice this is rarely the case. For example, the most commonly found case of parameter mis-specification is the one where we assume the process data are uncorrelated when there is actually correlation present in the data. The effect of this type of parameter mis-specification on Controls charts has received widespread attention in the literaturÄ™.

An example of the effect of this type of parameter mis-specification on the Kalman filter is demonstrated in FigurÄ™ 5. In this example we input the randomly generated process data from the second example, from the previous

FigurÄ™ 4. Process Data and Kalman Estimate for Example 2.