FINITE ELEMENT QUANTITATIVE ANALYSIS AND DEEP LEARNING QUALITATIVE ESTIMATION IN STRUCTURAL ENGINEERING

Abstract

In the past two decades, finite element method (FEM) has been widely used to study mechanics of solids, fluid-structure interactions, and building construction strategies. FEM has been rapidly grown all over the world due to development of computer technology. Computer has much more powerful computing capability than humans. However, structural engineering education not only focused on teaching engineers to use FEM as computation tool, but also concentrated on cultivating engineers' capability of experience-based qualitative analysis. In addition, artificial intelligence techniques have been rapidly developed in recent years. It is demonstrated that human experience-based capabilities might also be replaced by deep learning methods in various game-playing areas. Thus, this study aims at exploring what role artificial intelligence techniques will play in the futural structural analysis area. In this paper, several finite element analyses are carried out for three representative boundary value problems, such as tightly stretched wires under loading, soil seepage, and plane stress. Corresponding deep neural networks are trained using FEM simulation data to quickly and accurately predict results of relevant problems. It is indicated that to some extent artificial intelligence technique might replace human experience-based qualitative analysis as a surrogate of FEM.