An Approach to Tracking Filter Design using Natural Selection Methods

When an observer is based upon an imperfect model with constant disturbances, non-zero errors will exist between the observer\´s outputs and the plant\´s. Based upon these errors, certain parameters in the observer can be adjusted to improve its estimate of other, nonmeasured parameters. The tracking filter design problem deals with determining which outputs should be used to adjust which parameters in the model to optimize the model\´s estimate of the nonmeasured parameter. An algorithm for designing a tracking filter based upon a natural selection approach to optimization is presented in this paper. It is shown that if only one parameter is to be adjusted, this approach will produce the optimal result. To illustrate this design procedure, the selection of three sensors and adjustments for a modern jet engine is considered. Here, it is shown that over 80% of the permutations can be eliminated while retaining 90% of the "best" tracking filters.