A robust strategy for FE model updating of composite panels using a combined ANN-SMA surrogate-assisted optimization framework

一种基于人工神经网络-形状记忆阵列代理辅助优化框架的复合材料面板有限元模型更新的稳健策略

阅读:1

Abstract

This study presents a novel hybrid surrogate-assisted optimization framework for finite element (FE) model updating of composite panels. Accurately updated FE models offer a reliable basis for design and analysis of modern structural systems and are widely employed in various industrial and engineering applications. However, model updating of composite structures remains a significant challenge due to their inherent anisotropy and complex mechanical behavior, which often limit the effectiveness of conventional updating techniques. To address these limitations, a hybrid framework is introduced for the first time, combining surrogate modeling with the Slime Mould Algorithm (SMA), a recently developed metaheuristic optimizer. For the proposed model updating strategy, a baseline FE model was initially developed, and an objective function was defined to quantify discrepancies between numerical predictions and experimental measurements. The baseline model generated training data for surrogate models constructed using Response Surface Methodology (RSM), Kriging (KRG), and Artificial Neural Networks (ANN). Among these, the ANN-based surrogates achieved superior predictive accuracy and were selected for the optimization phase. Sensitivity analysis identified the most influential parameters for model updating. Optimization was then carried out to minimize the objective function using the SMA. The resulting ANN–SMA framework substantially reduced discrepancies between experimental and numerical modal responses, yielding a high-fidelity FE model. Comparative assessments against established metaheuristic algorithms, including Particle Swarm Optimization (PSO) and Genetic Algorithm (GA), both integrated with ANN surrogates, revealed that the ANN-SMA framework exhibited enhanced convergence stability and computational efficiency, with reductions in runtime of approximately 7% and 30% relative to ANN-PSO and ANN-GA, respectively. These findings confirm that the proposed ANN–SMA hybrid framework provides a robust, efficient, and reliable tool for precise FE model updating of composite structures, with significant potential for broader applications in structural dynamics and structural health monitoring (SHM).

特别声明

1、本页面内容包含部分的内容是基于公开信息的合理引用;引用内容仅为补充信息,不代表本站立场。

2、若认为本页面引用内容涉及侵权,请及时与本站联系,我们将第一时间处理。

3、其他媒体/个人如需使用本页面原创内容,需注明“来源:[生知库]”并获得授权;使用引用内容的,需自行联系原作者获得许可。

4、投稿及合作请联系:info@biocloudy.com。