A platform for research: civil engineering, architecture and urbanism
A platform for research: civil engineering, architecture and urbanism
Two-Dimensional Truss Topology Design by Reinforcement Learning
In this study, a reinforcement learning algorithm for topology design of 2D trusses is proposed. When topology design of trusses is done by computers, optimization algorithms, especially metaheuristic optimization algorithms, are normally used. These optimization algorithms require large computational resources and time. In addition, each design problem has to be solved individually from scratch. This study takes a different approach to this type of problem. Instead of using optimization algorithms to search for optimal solutions of these problems, each of them from scratch, an intelligent agent is created and trained to solve them. To this end, topology design of simple 2D trusses is formulated as a game called “TrussGame.” An intelligent agent is then created for the game and is trained to play the game via the reinforcement learning process. After the training, the agent becomes an expert that can instantaneously design the topologies of simple 2D trusses.
Two-Dimensional Truss Topology Design by Reinforcement Learning
In this study, a reinforcement learning algorithm for topology design of 2D trusses is proposed. When topology design of trusses is done by computers, optimization algorithms, especially metaheuristic optimization algorithms, are normally used. These optimization algorithms require large computational resources and time. In addition, each design problem has to be solved individually from scratch. This study takes a different approach to this type of problem. Instead of using optimization algorithms to search for optimal solutions of these problems, each of them from scratch, an intelligent agent is created and trained to solve them. To this end, topology design of simple 2D trusses is formulated as a game called “TrussGame.” An intelligent agent is then created for the game and is trained to play the game via the reinforcement learning process. After the training, the agent becomes an expert that can instantaneously design the topologies of simple 2D trusses.
Two-Dimensional Truss Topology Design by Reinforcement Learning
Lecture Notes in Civil Engineering
Reddy, J. N. (editor) / Wang, Chien Ming (editor) / Luong, Van Hai (editor) / Le, Anh Tuan (editor) / Sahachaisaree, S. (author) / Sae-ma, P. (author) / Nanakorn, P. (author)
2020-07-28
9 pages
Article/Chapter (Book)
Electronic Resource
English
Machine learning , Reinforcement learning , Actor-critic algorithm , Truss design , Model-free reinforcement learning algorithm Engineering , Building Construction and Design , Sustainable Development , Structural Materials , Landscape/Regional and Urban Planning , Solid Mechanics , Construction Management
Two-Dimensional Truss Topology Design by Reinforcement Learning
| TIBKAT | 2020
| British Library Online Contents | 2013