This paper proposes the novel multi-physical nonlinear dynamic biaxial buckling and reliability analysis for sandwich graphene platelets reinforced porous plates with magneto-electro-elastic sheets (SGPLRP-MEE) advanced by the adaptive virtual model, accounting for physically inspired multi-dimensional uncertainties. To circumvent cumbersome multi-physical simulations while minimizing computational costs, an adaptive virtual modelling technique, namely the sequential design of experiment enhanced adaptive Kriging (SDOE-AKriging), is newly developed. The adaptive virtual model can accurately depict the sophisticated relationship between intrinsic uncertainties and concerned multi-physical responses using the fewest possible training samples, significantly reducing the computational expenditure in a dual fashion. Consequently, comprehensive quantified statistical information, structural reliability, and parameter sensitivity regarding multi-physical buckling failure behaviour are obtained, which are essential for smart composite safety assessment and serviceability limit state design. The superior performance, desirable accuracy, and advantageous efficiency of the SDOE-AKriging are validated through benchmark tests and extensive numerical experiments against existing machine learning algorithms. Additionally, the model's flexible modularity and rapid information update merits facilitate the establishment of a continuous failure diagnosis-prognosis loop, enabling swift decision-making and timely failure prevention in rapidly changing conditions. Parametric studies further reveal that focusing solely on deterministic responses can yield misleading conclusions in practical design scenarios.

中文翻译：

---

自适应虚拟建模增强了 SGPLRP-MEE 板的动力学和可靠性分析


该文提出了一种新颖的多物理非线性动态双轴屈曲与磁电弹性片 （SGPLRP-MEE） 的夹层石墨烯片增强多孔板的可靠性分析方法，该文考虑了物理启发的多维不确定性。为了规避繁琐的多物理场仿真，同时最大限度地降低计算成本，新开发了一种自适应虚拟建模技术，即实验增强自适应克里金的顺序设计 （SDOE-AKriging）。自适应虚拟模型可以使用尽可能少的训练样本准确描述内在不确定性和相关多物理响应之间的复杂关系，以双重方式显着减少计算支出。因此，获得了有关多物理屈曲失效行为的全面量化统计信息、结构可靠性和参数敏感性，这对于智能复合材料安全评估和正常使用极限状态设计至关重要。SDOE-AKriging 的卓越性能、理想的准确性和有利的效率通过对现有机器学习算法的基准测试和广泛的数值实验得到了验证。此外，该模型的灵活模块化和快速信息更新的优点有助于建立连续的故障诊断-预测回路，从而在快速变化的条件下实现快速决策和及时预防故障。参数研究进一步表明，在实际设计场景中，仅关注确定性响应可能会产生误导性的结论。