Abstract:
The use of the
estímation-before-modellíng (EBM) two step identification
procedure for the determination of aircraft aerodynamic derivatives from flight
test data is
analysed and illustrated. In the first step of the identification
procedure the usual Extended Kalman Filter (EKF) associated with the Modified
Bryson-Frazíer (MBF) smoother is compared with a new alterative filtering and
smoothing process. The new smoother is simpler and less computationally
demanding than the MBF smoother. However, its main advantage is that it
enables simultaneous data
smoothing with state derivative estimation, thereby
avoiding the need for a separate differentiation algorithm. The new smoother differentiator
has an
important feature that is the determination of the noise
characteristics of the measurement
signal under analysis prior to the smoothing
process. This is done by variance matching between the theoretical and
measured autocorrelation of the innovation
process generated by a Kalman filter.
The new
technique is compared with the old one by determining the aerodynamic
models for a EMB-312 Tucano dutch roll manoeuvre. It is demonstrated that
the new smoother
may be used to replace the MBF. Otherwise the new
technique
is used in the
analysis of the Handley Page Jetstream-100 aircraft low speed
controls free
phugoid trying to identify the contribution of the power Variation
observed
during the phugoid to the stability of the oscillation. Finally the models
obtained from the
phugoid analysis are reprocessed using the Total Least Square
regression and the results are compared with those from the ordinary Least
Square formulation.
