To identify a nonlinear reduced model accurately, and thus obtaining reliable results, the reduction basis must be constructed with representative sets of mode shapes containing the transverse and in-plane (membrane) displacements. Obtaining the first set is generally straightforward since the first modal modes of a model are generally transverse. However, detecting which modes are in-plane is proved to be difficult in the case of complex structures. In this thesis, a method using the modal derivatives to generate a set of membrane modes from the transverse one is investigated. This approach enables a more automated definition of the reduction basis while keeping the quality of the results at a same or higher level than existing methods.
Moreover, an investigation of optimisation methods to improve the computation time of the identification of the reduced order model coefficients is performed. Those methods are evaluated and their results compared to the original approach to estimate their usability.