基于深度强化学习的高层剪力墙结构智能设计方法

在传统的高层剪力墙结构设计中，需结构工程师依据自身经验反复地对剪力墙的布置方案进行调整，致使工作效率低、重复性劳动多、设计周期长，且设计结果依赖于工程师的个人经验，很难得到最优结构方案。为此，提出了高层剪力墙结构的智能设计方法，包括智能建模和智能优化两个环节。基于连通区域分析、图结构算法、普氏分析、边缘检测算法提出了高层剪力墙结构智能建模方法，包括房间检测、建筑分割与识别、剪力墙自动布置等，智能建模结果可以满足实际工程要求；基于深度强化学习给出了高层剪力墙结构智能优化方法，优化方法收敛性好。通过实际工程算例对所提出的智能设计方法进行了验证，结果表明，所提出的高层剪力墙结构智能设计方法可行，与传统的设计方法相比，设计周期可缩短90%以上，剪力墙材料用量可节省20%以上。

Intelligent design method of high-rise shear wall structures based on deep reinforcement learning

In previous design schemes of high-rise shear wall structures, shear wall structural drawings are repeatedly modified and checked by designers based on their own experience, which causes low-efficiency, much repetitive work, long design period and high-subjectivity. Besides, optimal structural design is hard to be manually obtained. Therefore, an intelligent design approach to structural design of high-rise shear wall structures was proposed, which included the intelligent parametric modelling and intelligent optimization. Based on the connected component analysis (CCA), graph structure (GS), Procrustes analysis (PA) and edge detection algorithm (EDA), a novel intelligent method for generating structural models of high-rise shear wall structures was proposed where the detection of rooms, segmentation and recognition of buildings, and automated layout formation of shear walls were included, and generated structural models meeted the requirements of practical application. A novel intelligent method for optimization was developed based on the deep reinforcement learning (DRL) with good convergence. The proposed approach was demonstrated to be feasible and validated by the design of a real-world shear wall structure where design period was reduced by above 90% and material cost of shear walls was decreased by above 20%.