Intelligent layout design of curvilinearly stiffened panels via deep learning-based method

Peng Hao,Dachuan Liu,Kunpeng Zhang,Ye Yuan,Bo Wang,Gang Li,Xi Zhang

DOI: https://doi.org/10.1016/j.matdes.2020.109180

2021-01-01

Abstract:<p>The structural efficiency of stiffened panels can be significantly improved by utilizing curvilinear stiffeners because of their outstanding design flexibility. However, the explosion of design variables poses a stiff challenge to the design of layouts of such structures. In this study, a novel layout optimization method is proposed for curvilinearly stiffened panels based on deep learning-based models, which enables their intelligent design. Unlike traditional methods, the image-based structural layout, which is characteristic of curvilinear stiffener paths, is employed as a design variable, and convolutional neural networks (CNNs) are used to extract the layout features from the curvilinear stiffeners and construct a surrogate model between layout features and structural performance. Subsequently, sub-optimization is performed using the constructed CNN to obtain new designs and correspondingly update the dataset. As the trained CNN does not require input data to exhibit the same number of stiffeners present in the training data, the proposed optimization framework can be used to address layout designs of stiffeners with variable numbers. Numerical examples demonstrate that the proposed intelligent optimization framework significantly improves the optimization efficiency compared to traditional models. It also indicates the extraordinary promise of deep learning-based methods in the field of engineering optimization.</p>

materials science, multidisciplinary