Session: 36-03 MDO & Multi-Fidelity

Submission Number: 177555

CNN-Assisted Robust Design of an Intentionally Mistuned Frictional Damper for Bladed-Disk Systems 

Bladed-disk assemblies are prone to high-cycle fatigue due to dynamic loads. Friction damping offers a low-cost, effective means of reducing resonant vibrations and improving durability. This study investigates the design of novel, intentionally mistuned ring dampers to enhance the robustness of friction damping. The motivation stems from the observation that mistuning can facilitate the coupling between different nodal diameters, potentially increasing the relative displacement at the damping interface and thereby improving energy dissipation. We also develop a comprehensive framework to optimise their nonlinear damping performance under both nominal and uncertain conditions assisted by convolutional neural network. Nonlinear backbone curves are computed using damped nonlinear modal analysis based on the Extended Periodic Motion Concept. A convolutional neural-network surrogate is trained to predict amplitude-dependent damping behaviour across variations in ring mass, mistuning extent, mistuning pattern, and contact parameters. The two metrics, the peak damping ratio and logarithmic width, are used to assess dissipation strength and effective amplitude range for the friction damping for robust design optimisation, where the uncertainties from contact parameters and manufacturing variations are taken into account. Overall, the robust solutions favour designs that suppress sensitivity rather than maximise nominal performance, resulting in damping responses that remain consistent despite parameter uncertainties.

Presenting Author: Peiyu Wang Universtiy of Southampton

Presenting Author Biography: Peiyu Wang is a PhD student at the University of Southampton, United Kingdom, under the supervision of Dr. Jie Yuan. His research focuses on nonlinear dynamics and frictional contact modelling in turbomachinery structures, with an emphasis on surrogate-assisted and data-driven approaches for efficient prediction and optimisation.

Authors:

Peiyu Wang Universtiy of Southampton
David Toal University of Southampton
Jie Yuan University of Southampton