土木工程智能计算分析研究进展与应用

随着数字孪生等数字化转型理念的推广，土木工程传统计算分析方法逐渐无法满足高效仿真、实时交互等新时代需求。而人工智能技术凭借其强大的拟合能力、卓越的计算效率以及优异的开放拓展性，日益受到研究者的关注，成为传统方法极具前景的更替选择，并发展出了土木工程智能计算分析这一新的研究方向。为系统梳理并展示相关前沿进展，针对土木工程智能计算分析研究进行全面综述，根据处理的问题输入属于静态、动态或是复合特征对现有研究进行分类，归纳各类研究的主要技术路线，统计常用智能算法，并概述在材料、构件、体系三层次场景的典型应用，进而分析了现有研究的局限性。为克服这些局限性，重点介绍完全基于深度学习的土木工程端到端智能计算框架DeepSNA，展示其相较于传统计算方法的高准确性与高效性。并提出四个开放性问题，为土木工程智能计算分析的未来研究方向提供参考。

Research and application of intelligent computation in civil engineering

With the prevalence of digital transformation represented by the concept of digital twins, traditional computational methods in civil engineering fail to satisfy the new-era need, such as efficient simulation and real-time interaction, etc. Artificial intelligence technology, which features powerful expressiveness, remarkable computational efficiency, and excellent scalability, is attracting increasing researchers and becoming a promising alternative. Accordingly, intelligent computation is growing as a new research branch in civil engineering. In order to systematically sort out and display the relevant frontier progress, a comprehensive review of the research on intelligent computation in civil engineering was conducted, where the literature was categorized into static feature learning, dynamic feature learning, and composite feature learning as per the problem inputs. For studies in each category, the main technical route was abstracted and the commonly-used intelligent algorithms was counted statistically. The typical applications on the material, member, and system levels were briefly introduced, respectively. On this basis, the limitations of the present studies were analyzed. To overcome these limitations, DeepSNA, the first end-to-end intelligent computational framework in civil engineering entirely based on deep learning, was introduced, demonstrating its high accuracy and outstanding computational efficiency of the developed framework compared with the conventional numerical methods. Finally, four open problems for future research were proposed, which can provide references for the future research on intelligent computational analysis in civil engineering.