Model Order Reduction of Aerodynamics for Flight Dynamic Simulations

Abstract

In the context of model order reduction and data compression, a parametric implementation of the proper orthogonal decomposition is presented. A linear vortex lattice method is adapted for unmodelled effects, creating large amounts of variables per parametric space and the basis formed via the method of snapshots. Based on the distribution of the sytems "energy" over the decomposition's modes, only a few can be retained resulting in a reduced order model. The behaviour of the general coefficients in the parametric space is examined and a cubic spline curve fitting procedure implemented. Due to the low number of interpolated modes, real time capabilities emerge. The reduced model is then combined with the 6 degree of freedom nonlinear equations of motion to simulate disturbance responses and examine the stability modes. Results were validated against a reference aerodynamic data set and showed a good agreement for both, aircraft totals in the extended flight envelope and state variable time responses. The effect of reducing the retained POD modes further was investigated and the resulting differences highlighted. The decrease in accuaracy variations with increasing modes, underlined the uneven "energy" distribution. The stability modes of the linearized system for both aerodynamic models correlated well with some larger differences for the spiral mode. Computational power and memory storage issues were addressed and a number of solutions and strategies proposed.