Improvements to Wind-Tunnel Flutter Prediction with Application to the Active Aeroelastic Aircraft Testbed Wind-Tunnel Model

The paper presents a study of aeroelastic system identification and flutter prediction in a wind tunnel test. Multi-output autoregressive moving-average and multi-output autoregressive models are fitted to acceleration and strain data and their correlation functions, respectively, to generate models that account for the dependencies between the different measured data. The modal frequencies and damping values are extracted from these models and used to assess the flutter margin at different airspeeds and predict the flutter onset speed. The method is applied in a flutter prediction wind-tunnel study of a full-scale half-span aeroelastic demonstrator aircraft, using multiple acceleration data and strain data from fiber-optic strain sensors. Both models provide a smooth variation of the modal parameters (frequencies and damping values) from which it was easy to identify the critical aeroelastic modes that coalesce in flutter. Based on the different data sets, all models predict flutter onset at approximately the same airspeed. An advantage of the proposed system identification approach is that it concurrently accounts for the data from multiple sensors, avoiding the need to average the different modal dynamics estimated from the different sensors. The multi-output autoregressive modeling is more compact and computationally efficient than the multi-output autoregressive moving-average model, but also yields less smooth parameter variation. It can be used to compute an approximate flutter margin. The greatest challenge of the multi-output autoregressive, which is based on the correlation functions of the filtered data, is mode tracking due to a large number of roots at close proximity. This problem is aggravated with larger data sets. Therefore, despite the method’s accurate system identification and flutter prediction, other system identification methods that are more suitable for large data sets have to be examined for flutter testing.