Aerodynamic Parameter Estimation from Flight Test Data

Abstract

vi Abstract Aerospace industry is becoming increasingly interested in unmanned aerial vehicles UAVs, due to increased computational power of micro-controllers, ease of development and flexibility of operation. The role of UAVs for surveillance and reconnaissance missions is well established, however civil applications like crop monitoring, remote sensing, crowd monitoring, etc. are also becoming popular. There is huge variation in size and complexity of UAVs, from hand-sized micro UAVs to UAVs weighing several thousand pounds. UAVs are tasked to autonomously navigate increasingly complicated missions. This autonomous navigation requires robust flight control system. Robustness of control system largely depends on accuracy of aerodynamic parameters used in flight dynamic model. Aerodynamic parameters constitute of stability and control derivatives, can be computed from empirical formulations like DATCOM, vortex lattice methods, VLAero, Computational fluid dynamics codes, ANSYS Fluent, Wind tunnel testing of scaled models and Flight test data. System identification and parameter estimation is the art of building mathematical model of dynamical system based on experimental data. Aerodynamic parameter estimation combines system identification with aeronautics to estimate aerodynamic parameters from flight test data. Aerodynamic parameter estimation is not only used to compute aerodynamic parameters of a new configuration but also to validate results of CFD computations and wind-tunnel testing. The present work starts with the application of parameter estimation techniques to a simple dynamical system, mass-spring system. Several important concepts i.e. effect of white and colored noise on confidence bounds for the estimates, model structure determination, hypothesis testing, parameter significance testing, etc. are investigated. This is followed by application of time domain least square and maximum likelihood techniques to flight simulation data. Simulation data is used instead of actual flight test data due to non-availability hardware and piloting skills. Estimated aerodynamic parameters are found to be within 95% confidence interval bounds of the true values. Model structure determination, design of inputs for informative outputs, data consistency checks are also investigated.