摘要
土遗址充分反映了古代的营造技术与文化属性,是研究古代生产生活方式的重要载体。暴露于室外的土遗址常年遭受降雨、携沙风、温差等因素的影响,裂隙病害显著发育,严重威胁了土遗址的整体稳定性。本文基于有限的现场调查数据,充分考虑土遗址赋存环境的影响,采用k-means聚类及高斯过程多分类方法对土遗址的裂隙病害发育等级进行概率预测,为进一步采取对应防治措施提供一定理论依据。研究结果表明:基于本文所提出的数据模拟方法可有效增加样本数量,且模拟数据与现场调查数据具有高度相似性;100次随机试验统计结果对应的测试集的查准率、查全率、准确率与F1度量的中值均在0.92左右,说明基于模拟数据所建立的模型具有较好的泛化性能,可实现土遗址裂隙病害发育等级的合理预测。
Earthen archaeological sites are a fundamental embodiment of ancient construction methods and provide invaluable sources for studying ancient lifestyles and material culture.However,open air earthen sites are constantly affected by various climactic factors such as rain,wind carrying sand,and temperature differences that change throughout the year,which leads to visible cracking that severely threatens the overall stability of earthen structures.Based on limited data from in-situ investigations,and in full consideration of the influence of the environment in which earthen sites are located,this paper proposes that a Gaussian process classification method combined with a k-means algorithm would be an effective means of evaluating the level of development of cracks that have appeared at earthen sites,so as to provide a theoretical basis for enacting appropriate prevention and control measures.The results show that the data simulation and augmentation method proposed in this paper are a suitable way of addressing issues caused by imbalanced data for cracks in different stages of development,and that the simulated data is very similar to that recorded during in-situ investigation.The median value of four crucial quantitative metrics,namely the precision ratio,recall ratio,accuracy and F1 score of the validation dataset corresponding to one hundred randomized experiments,are all approximately or greater than 0.92,which indicates that the proposed method has an adequate generalization score,and is thus suitable for predicting the development levels of cracks at earthen sites.
作者
赵腾远
宋超
谌文武
郭志谦
许领
ZHAO Tengyuan;SONG Chao;CHEN Wenwu;GUO Zhiqian;XU Ling(School of Human Settlements and Civil Engineering,Xi’an Jiaotong University,Xi’an 710049,China;College of Civil Engineering and Mechanics,Lanzhou University,Lanzhou 730000,China;School of History and Culture,Lanzhou University,Lanzhou 730000,China)
出处
《石窟与土遗址保护研究》
2023年第4期75-86,共12页
Research On The Conservation Of Cave Temples And Earthen Sites
基金
国家重点研发计划项目(2020YFC1522201)
关键词
遗址
裂隙病害
数据增广
数据驱动
earthen sites
cracking
data simulation and augmentation
data-driven method
作者简介
第一作者:赵腾远(1988—),男,河北省邢台市人,西安交通大学教授,主要从事空间数据分析与建模、岩土工程风险分析、地质灾害风险分析与评估等方面的研究,E-mail:tyzhao@xjtu.edu.cn;通讯作者:许领(1982—),男,安徽省淮南市人,西安交通大学教授,主要从事黄土灾变机理、监测、预警与防控理论、人工智能技术在岩土与地质灾害防治领域的应用等方面的研究,E-mail:xuling82@xjtu.edu.cn
