A novel penalty-based reduced order modelling method for dynamic analysis of joint structures

Yuan, Jie and Salles, Loic and Wong, Chian and Patsias, Sophoclis; Fehr, Jörg and Haasdonk, Bernard, eds. (2019) A novel penalty-based reduced order modelling method for dynamic analysis of joint structures. In: IUTAM Symposium on Model Order Reduction of Coupled Systems. IUTAM Bookseries . Springer, Cham, Switzerland, pp. 165-176. ISBN 9783030210137 (https://doi.org/10.1007/978-3-030-21013-7_12)

[thumbnail of Yuan-etal-Springer-2020-A-novel-penalty-based-reduced-order-modelling-method]
Text. Filename: Yuan_etal_Springer_2020_A_novel_penalty_based_reduced_order_modelling_method.pdf
Accepted Author Manuscript

Download (653kB)| Preview


This work proposes a new reduced order modelling method to improve the computational efficiency for the dynamic simulation of a jointed structures with localized contact friction non-linearities. We reformulate the traditional equation of motion for a joint structure by linearising the non-linear system on the contact interface and augmenting the linearised system by introducing an internal non-linear penalty variable. The internal variable is used to compensate the possible non-linear effects from the contact interface. Three types of reduced basis are selected for the Galerkin projection, namely, the vibration modes (VMs) of the linearised system, static modes (SMs) and also the trial vector derivatives (TVDs) vectors. Using these reduced basis, it would allow the size of the internal variable to change correspondingly with the number of active non-linear DOFs. The size of the new reduced order model therefore can be automatically updated depending on the contact condition during the simulations. This would reduce significantly the model size when most of the contact nodes are in a stuck condition, which is actually often the case when a jointed structure vibrates. A case study using a 2D joint beam model is carried out to demonstrate the concept of the proposed method. The initial results from this case study is then compared to the state of the art reduced order modeling.