A nested non-invasive stochastic isogeometric method for the deformation of porous functionally graded material plates with high-dimensional material uncertainties
Junli Guo,Yahui Zhang
DOI: https://doi.org/10.1016/j.cma.2024.116898
IF: 6.588
2024-03-07
Computer Methods in Applied Mechanics and Engineering
Abstract:The stochastic isogeometric analysis (SIGA) based on random geometric methods plays a crucial role in structural random analysis. However, existing intrusive SIGA methods suffer from limitations of low applicability or efficiency. On the other hand, current material uncertainties are mostly limited to low dimensions and often ignore spatial correlations. Therefore, we propose an efficient and adaptive nested non-intrusive SIGA method based on Radial Basis Function Neural Network (RBFNN). This method combines quasi-Monte Carlo simulation (QMCS) to achieve a general solution for high-dimensional material uncertainty in porous functionally graded materials (PFGM) plates, including statistical quantities and probability density functions (PDFs) of random responses. Firstly, within the IGA framework, we establish a geometrically nonlinear model for PFGM plates using HSDT and Von Kármán strain-displacement relations, effectively mitigating shear locking effects. Secondly, by considering spatial correlations, we employ the Karhunen-Loève expansion to establish a high-order random field for material uncertainty. Subsequently, we propose a nested and non-intrusive response surface analysis method using RBFNN, significantly enhancing the accuracy and efficiency of SIGA in addressing high-dimensional uncertainty problems. Finally, by leveraging the efficiency of SIGA, we overcome the low convergence speed drawback of QMCS, achieving a comprehensive solution for statistical quantities and PDFs of high-dimensional uncertainty problems. The advantages of efficiency and accuracy are validated through random response analysis of high-dimensional uncertain PFGM plates. In randomly structured geometrically nonlinear PFGM plates, porosity significantly influences the response, while the gradient exponent and Young's modulus have a weaker but non-negligible impact on the response.
mechanics,engineering, multidisciplinary,mathematics, interdisciplinary applications