DEEP LEARNING ASSISTED STOCHASTIC FREE VIBRATION ANALYSIS OF POROUS STRUCTURES

Porous structures are widely utilized across various fields, including biomedical engineering, environmental remediation, energy storage devices, and construction and building materials, owing to their lightweight nature, high surface areas, exceptional sound and energy absorption capacities, as well as their thermal insulation properties and customizable designs. Practical applications of porous structures encompass tissue scaffolds, air and water filtration systems, fuel cells, lightweight concrete, and acoustic panels. Investigating the free vibration behavior of porous structures is essential, as unfavorable behavior can jeopardize structural safety and serviceability. Therefore, this study examines the stochastic free vibration performance of porous structures, considering relatively high-dimensional uncertainties with disparate distributions. A deep learning strategy, i.e., the cutting-edge Gated Additive Tree Ensemble, is incorporated to delineate the complex relationship between various uncertainties and the vibration characteristics of porous structures. Extensive statistical information on structural behavior, involving the mean, standard deviation, coefficient of variation, probability density function, and cumulative distribution function, is offered to facilitate structural safety assessments and reliability-based material optimization. Moreover, numerical studies demonstrate quantitatively the advantages of the presented deep learning technique over existing surrogates in terms of estimating efficiency and accuracy.