| About this Abstract |
| Meeting |
MS&T23: Materials Science & Technology
|
| Symposium
|
Materials Informatics for Images and Multi-dimensional Datasets
|
| Presentation Title |
Efficient Void Shape Optimization Using Deep Generative Convolutional Neural Networks |
| Author(s) |
Zihan Wang, Anindya Bhaduri, Sandipp Krishnan Ravi, Piyush Pandita, Changjie Sun, Liping Wang |
| On-Site Speaker (Planned) |
Anindya Bhaduri |
| Abstract Scope |
Structure-to-property (SP) linkage is a key component to the development and design of advanced materials. Physics-based simulation models like finite element method (FEM) are often used to predict physical quantities of interest, such as, deformation, stress and strain, as a function of material microstructure in material and structural systems. Multiple runs of such models for design evaluations may be computationally expensive and time intensive when performing structural optimization, especially when dealing with the high dimensionality of the input microstructure. Deep convolutional neural network (CNN) based surrogate models have been typically found to be very useful in handling such high-dimensional problems. In this work, the system under study is a single ellipsoidal void structure under uniaxial tensile load represented by a linear elastic FEM model. We deploy an integrated deep generative model based design framework to optimize the void shape that yields the minimum stress conditions. |