Strategy for Global Parametric Optimization in Nonlinear Vibration: Application on a Gantry Crane - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2023

Strategy for Global Parametric Optimization in Nonlinear Vibration: Application on a Gantry Crane

Résumé

With standards in permanent evolution (e.g., after the 2011 Japan earthquake), the modification of existing industrial structures is sometimes required. Parametric optimization is a useful tool to ensure the retrofit of these structures, allowing to find the most effective solutions. This study aims at the optimization of a gantry crane subjected to a base excitation to comply with newer seismic standards. To get a more precise model, contact phenomena are considered at joints, which bring nonlinearities. However, the dynamical simulation of nonlinear models is computationally expensive. The fulfilment of global optimization with classic methods is then not a viable solution because a very large number of computations can be required. Therefore, a Bayesian Optimization strategy based on (1) an efficient nonlinear vibration solver and (2) a Gaussian Process surrogate model is proposed to make the parametric optimization of the gantry crane computationally acceptable. Taking advantage of the fact that only a few interface degrees of freedom are nonlinear, the first step consists in performing a Craig Bampton condensation on the linear substructures. This results in a nonlinear reduced dynamical system in which the interface degrees of freedom are conserved. The solution of this system is then computed using the Harmonic Balance Method with an Alternating Frequency/Time approach. It allows to find periodic solutions with reasonable accuracy avoiding the costly computation of possibly long transients. Numerical path continuation is also applied using a predictor-corrector method to follow the evolution of the solution with respect to the frequency and overcome potential turning points. The computational cost is still quite expensive, and the direct use of classical global optimization algorithms combined with this solver is not practicable. Therefore, a Bayesian Optimization based on Gaussian Process and an acquisition function is performed to achieve the global parametric optimization with a limited number of calls to the mechanical solver. The strategy is applied on a simplified model of a gantry crane. In this study, the seismic load is replaced by a harmonic displacement applied on the basis within the typical frequency range of seismic action (0-35 Hz). The contact between a wheel and an anti-tipping device is studied in order to limit the acceleration of the payload and avoid the overturning of the gantry. References [1] Sarrouy E., Sinou J. (2011) "Non-Linear Periodic and Quasi-Periodic Vibrations in Mechanical Systems - On the use of the Harmonic Balance Methods". Adv. Acoust. Vib.. IntechOpen. Chap. 21:419-434. [2] Frazier P. I. (2018) "A Tutorial on Bayesian Optimization". arXiv. 1807.02811 [3] Rasmussen C. E., Williams C. K. I. (2006) Gaussian Processes for Machine Learning. The MIT Press.
Fichier non déposé

Dates et versions

hal-04077038 , version 1 (21-04-2023)

Identifiants

  • HAL Id : hal-04077038 , version 1

Citer

Quentin Ragueneau, Luc Laurent, Antoine Legay, Thomas Larroque, Romain Crambuer. Strategy for Global Parametric Optimization in Nonlinear Vibration: Application on a Gantry Crane. 9th International Conference on Computational Methods in Structural Dynamics and Earthquake Engineering, COMPDYN 2023, ECCOMAS, Jun 2023, Athens, Greece. ⟨hal-04077038⟩
24 Consultations
0 Téléchargements

Partager

Gmail Facebook X LinkedIn More