The authors present a Dynamic Data-Driven Estimation Framework (DDDEF) to estimate aeroelastic responses of an aircraft with sensor observations. The SensorCraft, an experimental joined-wing aircraft, is considered as a case study for this framework. Due to the thin and long flexible wings of this aircraft, the system is susceptible to aeroelastic instabilities. Simulations of the aeroelastic effects of such flexible structures with acceptable accuracy are computationally expensive. Currently, these simulations can only be performed offline. Additionally, sensor measurements are typically made at spatially discrete locations. On their own, these measurements may not be sufficient to determine an aircraft's state and safe flight envelope. The main contribution of this work is in the construction of a general dynamic data-driven prediction framework. In this framework, the data from offline simulations of dynamic system responses are combined with sensor measurements at discrete locations to improve the accuracy of the state estimates of the aircraft response. This is achieved through several steps. First, the proper orthogonal decomposition is applied to reduce high-dimensional time series simulation data into low-dimensional data with unknown parameters. Next, Gaussian processes are constructed to approximate these parameters in order to obtain the full time series responses. Then, particle filtering is used to assimilate sensor data collected to further improve predictive performance. The effectiveness of the proposed approach is demonstrated on the SensorCraft's aeroelastic response state with one sensor's measurements as an illustrative example. Along with the state estimates being dynamically enhanced, a reduction of the uncertainty in the estimates is also shown.
展开▼