Improving System Identification Results: Combining a Physics-Based Stitched Model with Transfer Function Models Obtained Through Inverse Simulation

Kurzfassung

It is state-of-the-art to model helicopter dynamics by low-order equivalent linear systems at certain speed-dependent operating points for the flight envelope of interest. These models are obtained from flight test data by system identification. Based on the linear models, it is desired to simulate maneuvering flight such as acceleration and deceleration between hover and high forward-speed. Additionally, the identified operating point model should consist of a high modeling accuracy. Thus, the two modeling and simulation goals are: generation of a full-envelope simulation by embedding all linear models and improvement of the modeling accuracy at each operating point. These two goals stimulate the application of two techniques: model stitching and inverse simulation. Model stitching allows to implement a full-envelope quasi-nonlinear helicopter simulation whereas inverse simulation allows to analyze and model additional non-physical transfer functions that improve the linear model accuracy at a certain operating point. These transfer functions describe the remnants that are originally not considered by the system identification results. This paper presents the combination of the two techniques model stitching and inverse simulation as well as their application. Finally, the paper assesses the modeling accuracy for ACT/FHS flight test data.