Deep Q-learning agent for building earthquake resisting wall
The construction of dry-joint brick masonry that is subject to seismic loading can be modelled as a strategy problem wherein bricks must be laid to form a load resisting structure. To learn how to mount bricks while considering relevant engineering criteria, this paper trains an artificial intelligence using a reinforcement learning approach. Here, we created a simulation environment wherein an agent assembles bricks to receive rewards based on the measured force capacity of the dry-joint brick masonry and the geometry of the constructed wall in comparison with the target volume of the wall. The horizontal force capacity was assessed using kinematic analysis based on a rigid block modelling. This work shows that agents can learn construction strategies by finding optimum placements for the larger bricks to improve the earthquake resistance of masonry structures.
2022_Wang_et_al_3ECEES_Deep_Q.pdf
n/a
restricted
n/a
997.15 KB
Adobe PDF
679ad8b44938bf08f1689eb2e2f7ef44