Comprehensive System Identification of Ducted Fan UAVs - Cal Poly

access to hardware inside of test vehicles, manufacturer data must be applied for errorand noise modeling. These tools and techniques combine to represent the comprehensiveidentification of these vehicles.2.2 CIFERCIFER provides an environment and set of programs that perform the varioussteps of the system identification process. Nonparametric modeling, in which no modelstructure or order is assumed; in the form of frequency responses represented as Bodeplots are first extracted with CIFER. This then allows for the parametric modeling.Transfer functions, low order equivalent (LOE) systems, or state-space models withstability and control derivative representation 3 are all used. The identification process canbe summarizes as 4 :1. Nonparametric frequency response calculation from time history datao Use of Chirp-Z Fast Fourier Transforms (FFT) and complex functions to generatethe frequency responses over multiple windows and samples2. Multi-input frequency response conditioningo Off axis control inputs’ contribution to on axis response is removed3. Multi-window averaging of frequency responseso Combination of different window sampling sizes4. Parametric models fit to frequency responseso Transfer function models fit to single input single output (SISO) systemso State-space models fit to all controls and states for parameter extraction5. Time domain verification of parametric modelsWhen complete, this procedure yields accurate models to be applied for a varietyof tasks. CIFER does require flight test time histories in which the vehicle’s modes havebeen excited by frequency rich inputs. It is not limited to vehicle dynamics either. Thistool can be used anywhere frequency domain analysis is needed. CIFER is a powerfultool that incorporates all of the tools to needed to model in the frequency domain.- 12 -