Ludovic Colo, Gabriel Broux, Eric Garrigues
DOI Number: N/A
Conference number: IFASD-2019-072
Flutter is a dynamic phenomenon occurring on aircraft structures for particular flight points when structural and aerodynamic modes couple in an unstable way. This phenomenon is computed during design phases to predict the aeroelastic behaviour of the aircraft and prevent any instability in the flight domain. To solve this problem, aircraft manufacturers currently use algorithms pertaining to a family of methods labelled as the “p-k method”. These algorithms have been optimized for the flutter problem and provide quite accurate results but in the same time exhibit some flaws which can be prejudicial when one wants to fully automate the flutter analysis. New techniques to solve the non-linear eigenvalue problems such as the flutter problem have been developed recently. Based on the invariant pairs theory, they lead to more robust solutions in terms of unicity. A bloc Newton algorithm applying these techniques has been implemented, validated on standard flutter cases (both civilian and military aircraft) and tested on cases displaying convergence and unicity issues with the p-k method. The outputs present the expected behaviour despite a longer computational time. In order to keep computational times within industrial timeframe, a hybrid method, switching from p-k to block Newton “when required” has been developed. Thanks to this hybrid method implemented in Dassault Aviation in-house solver Elfini, standard flutter computations at Dassault Aviation benefit from the advantages of the invariant pairs theory.