[home] [Personal Program] [Help]

---

Session: Aeroelastic testing 5
Session starts: Thursday 20 June, 13:30
Presentation starts: 15:00
Room: Room 1.6

---

Francisco P. Reis (ISAE-SUPAERO)
Bilal Sharqi (University of Michigan)
Leandro Lustosa (ISAE-SUPAERO)
Charles Poussot-Vassal (ONERA)
Carlos E. S. Cesnik (University of Michigan)

Abstract:
As aircraft become more flexible as a consequence of design choices to improve fuel efficiency and operating costs, they experience larger deflections during normal operation. The ability to dynamically estimate or measure shape is useful for such flexible structures for path planning, advanced control and handling qualities. This work proposes an investigation of three model-free, real-time shape estimation methods for control applications focused on very flexible wings. Model-free techniques do not rely on precise dynamic aeroelastic models, resulting in estimators that do not require high computing power or in-depth system knowledge. Consequently, these methods are more easily adaptable to new applications. The methods rely on, respectively, fiber optic strain sensing (FOSS), a combination of inertial measurement units (IMUs) with sighting devices, and a combination of IMUs with magnetometer measurements. This paper compares the performance of the aforementioned methods regarding shape estimation and analyzes their data handling and processing requirements. Data from various static and dynamic tests conducted on the enhanced aeroservoelastic (EASE) model, a very flexible wing developed at the University of Michigan, are used to recover the shapes of the structure. The model consists of a very flexible wing attached to a rigid fuselage and tail, and is equipped with IMUs containing magnetometers along its wing. In addition, fiber optic cables were installed on the wing to acquire continuous strain measurements, and visual markers were attached for use with motion capture devices. The shape estimation methods include displacement recovery from the FOSS, complementary filters based on magnetometer and inertial data, and extended Kalman filters based on inertial and visual data to minimize error covariance and increase the methods’ precision. Results from the suite of tests performed on the EASE model were chosen to highlight differences between the estimators’ performances. Expected differences include deviations caused by the sensors’ bandwidth and precision, and estimator tuning. Since the control application dictates this tuning, these differences will be further discussed in the final paper.