Neural Nets for Structure Design
Project Description

In the design of mechanical structures such as airplanes and bridges, a crucial step is the prediction of mechanical stress of the structure produced by the applied loading. The stress value determines the safety of the structure. For large-scale structures, the conventional approach for stress prediction is too computationally intensive/costly to be applied in the design process. In this project, we will use deep learning methods to build accurate and efficient mechanical models for stress prediction. In particular, we will construct an 3D convolutional neural network to output the stress field of the structure for a given structure. Since the neural network is analytical, the computational time will be several orders of magnitude less than the time needed in the conventional methods. Hence it can be applied in the design of large-scale structures.

Supervisor
ZHANG Nevin Lianwen
Co-Supervisor
YE Wenjing
ZHANG, Nevin Lianwen
Quota
2
Course type
UROP1100
UROP2100
UROP3100
UROP4100
Applicant's Roles

Students will be involved in constructing CNNs. Training data will be provided. The students should have some experiences with CNNs.

Applicant's Learning Objectives

Students will learn (1) how to construct neural networks for applications in mechanical field and (2) some basic knowledge about the design of mechanical structures.

Complexity of the project
Moderate