Intelligent design of simply-supported planar truss based on reinforcement learning

Name: Zhao, Xianzhong

Organization : Tongji University. Department of Structural Engineering

Email : x.zhao@tongji.edu.cn

Name: Liu, Zhiyuan

Organization : Tongji University. Department of Structural Engineering

Email : 2211128@tongji.edu.cn

Name: Xiao, Weifang

Organization : Tongji University. Department of Structural Engineering

Email : weifangxiao@tongji.edu.cn

Name: Luo, Ruifeng

Organization : East China Architectural Design and Research Institute Co., Ltd

Email : ruifeng_luo@ecadi.com

keyword: Truss layout design

LCSH: Trusses, Concrete

; Markov decision process

LCSH: Intelligent design (Teleology)

LCSH: Reinforcement learning

LCSH: Monte Carlo method

Abstract: Intelligent design is a multidisciplinary research field involving artificial intelligence (AI), mathematical optimization, and human experience. The aim is to realize the automatic design of structures. As an important research topic of intelligent design in civil engineering, the truss design is characterized by high complexity, strict constraints and diversity of decision-making, which are the obstacles to the application of AI in designing the trusses. How to realize "brain-inspired design" by combining human knowledge with intelligent thinking is still a major challenge in the intelligent design of structures. This study develops an intelligent design algorithm for simply-supported planar truss based on reinforcement learning. The layout design problem of simply-supported planar truss is modeled as a Markov Decision Process (MDP). In order to solve the model, Monte Carlo Tree Search (MCTS), which performs well in sequential decision problems, is applied. Furthermore, MDP can include the engineering logic and knowledge that are extracted from the remarkable characteristics of simplysupported planar truss. This significantly condenses the initial solution space and optimizes the decision-making process. A contrast experiment is conducted to demonstrate the validity and efficiency of the proposed algorithm

King Mongkut's University of Technology North Bangkok. Central Library

Address: BANGKOK

Email: library@kmutnb.ac.th

Created: 2023

Modified: 2024-03-29

Issued: 2024-03-29

บทความ/Article

application/pdf

BibliograpyCitation : In Chulalongkorn University. 5th International Conference on Civil and Building Engineering Informatics (ICCBEI 2023) (pp.52-58). Bangkok : Chulalongkorn University

eng

©copyrights King Mongkut's University of Technology North Bangkok

ลำดับที่.	ชื่อแฟ้มข้อมูล	ขนาดแฟ้มข้อมูล	จำนวนเข้าถึง	วัน-เวลาเข้าถึงล่าสุด
1	ICCEBEI 2023pp.52-58.pdf	445.12 KB

ใช้เวลา

0.034038 วินาที