Flight Simulation Model Validation Using the Concept of Coefficient Matching

Although flight-control laws would have been designed using the linear math models of the aircraft, it is necessary to validate the entire design using nonlinear flight-simulation procedures. If required the models used in the design procedures should be modified based on the feedback from the flight tests conducted on the aircraft and its versions [19]. Most of the techniques for validation of the simulation model depend on estimating the linear models of the aircraft from flight test data at several flight conditions. However, as seen in Section 6.2.1, the aerodynamic database is given in the coefficient form and direct comparison is difficult and this approach of validation does not give a scope of updating the database that is in ‘‘lookup table’’ forms. An alternative approach is to estimate the aerodynamic coefficients from the flight data and determine the differences between the estimated coefficients and the ones obtained from the tables by applying the application formulae. Then an appropriate method can be used to model any discrepancies. The procedure requires a thorough validation of these models based on several sets of flight data and many flight conditions for getting consistent results. The discrep­ancy can be modeled by using either feed forward neural network (FFNN) or polynomial models.

Nominal aerodynamic coefficients can be computed from the application for­mula, which uses the aero database. The aerodynamic coefficients from flight/flight – like responses are estimated using suitable parameter estimation/Kalman filtering method. Direct use of the flight responses to compute the coefficient would give inaccurate results due to noise in the measure data. The estimation technique would be more accurate due to filtering out of the noise processes. It is obvious from Section 6.2.1 that all the signals/variables used for computing the coefficients from the application formula and parameter estimation would not be the same. However, within the structure of the rule used for these computations consistent sets of the flight variables/data should be used. Primarily the comparison should be carried out for the same flight conditions. A procedure of using this method is outlined below:

1. Compute the (nominal) force and moment coefficients using the application rule and related data.

2. Compute the coefficients from the flight responses by using an estimation technique.

3. Obtain the differences between these coefficients.

4. Fit FFNN or a polynomial model to these differences as functions of selected/required flight parameters.

5. Incorporate these incremental models into the nominal database models; this is likely to bring the upgraded coefficients closer to the actual ones.

6. Incremental models are compact ways to incorporate any future data – dependent variations into the database/coefficient base, i. e., new flight data can be used to get an updated database.

7. Model structure selection methods would be required to be used for getting adequate fit to the difference time histories of the coefficients.

Figure 6.10 depicts the process of this method of validation and upgrading of the aero database. A typical result of the application of this process with simulated data is shown in Figure 6.11. Other important aspects to be kept in mind are the flight data should be in proper time-stamped form, accurate flow angle data should be used, and

there should be variation of mass and CG. The task of determining the terms in the aerodynamic model buildup that should be revised is not easy. It is an iterative process and requires considerable engineering judgment. The team should have extensive knowledge of the aerodynamic effects on stability, control, and perform­ance of the aircraft for which this exercise is being done.