COOKIES: By using this website you agree that we can place Google Analytics Cookies on your device for performance monitoring. |
University of Cambridge > Talks.cam > Engineering - Dynamics and Vibration Tea Time Talks > Friction damping in structural dynamics: from fundamental understanding to physics-based machine learning identification
Friction damping in structural dynamics: from fundamental understanding to physics-based machine learning identificationAdd to your list(s) Download to your calendar using vCal
If you have a question about this talk, please contact div-c. Structural joints characterised by moving surfaces are present in most engineering structures. Idenfying the nonlinear forces generated by friconal interfaces, as well as understanding how their damping effects can be exploited to achieve vibraon reducon and energy dissipaon, are among the present crical challenges in structural dynamics. Moreover, the unpredicted nonlinear behaviour induced by fricon can lead to undesirable effects, such as stuck contacts, sck-slip, excessive oscillaons and even unexpected failures. Predicng the nonlinear response of fricon damped systems is therefore essenal to develop robust designs and efficient monitoring strategies. Nonetheless, the analysis and idenficaon of these systems is complicated by the nonlinear and nonsmooth nature of the friconal forces, which is responsible for the generang sharp variaons and mulple moon regimes in their dynamic response. This contribuon deals with three major challenges: (i) the derivaon of analycal soluons for the response of discrete mechanical systems with Coulomb fricon, and the predicon of their moon regimes; (ii) the development of an experimental framework for reproducing and validang the proposed mathemacal soluons; (iii) the formulaon of a physics-based machine learning approach for the idenficaon of friconal systems. Mathemacal soluons are obtained for the steady-state response of single- and mul-degree-offreedom (DOF) systems with a fricon contact to harmonic excitaon. The derived formulaons also enable the invesgaon of the dynamic behaviour of systems with oscillang contacng parts and/or different forms of damping. The results include the analycal predicon of the moon regimes (connuous, sck-slip, permanent scking) and the displacement transmissibility curves resulng from varying system parameters. The experimental invesgaon is performed on shear-frame setup with a brass-to-steel contact, able to reproduce the behaviour of single- and 2-DOF systems in contact with a fixed or oscillang wall. Tests run at different excing frequencies and fricon force amplitudes were able to closely reproduce the theorecal results, showing that Coulomb model can be used, in most condions, for describing the dynamic behaviour of structures with a metal-to-metal contact. The proposed idenficaon method combines a parally-known physics-based model of the system with noisy measurements of its response in a switching Gaussian process (GP) latent force model, where mulple GPs are used to model the nonlinear force across different moon regimes and a reseng model to generate disconnuies. Regime transions and disconnuies are inferred in a Bayesian manner, along with the nonlinear force, and can be used to implement forward models able to make reliable response predicons. This talk is part of the Engineering - Dynamics and Vibration Tea Time Talks series. This talk is included in these lists:
Note that ex-directory lists are not shown. |
Other listsThe Faraday Institute for Science and Religion Women in AI Food 4 Thought WebinarsOther talksAnimal movement with self-interaction via multiple memories Out-of-Equilibrium Transport Measurements Through Nanopores Reveals Velocity Fluctuations PCF-GAN and beyond: High-Fidelity Generative Models for Synthetic Time Series Generation Operator Algebras Spontaneous Transitions in Fish Schools CANCELLED - LMB Seminar: RNA Meets DNA: Dangerous Liaisons in the Genome |