村镇砌体建筑群信息智能获取与震害预测

我国村镇地区砌体结构分布广、数量大、抗震性能普遍较差，且具有结构形式多样、无数据和随机性强等特点，获取其建筑信息是实现震害预测的前提和基础，但传统人工检测技术效率低、不经济。为此，提出了一种图像识别与模糊推理相结合的村镇砌体建筑群信息智能获取方法。建立了基于卷积神经网络的砌体建筑图像识别模型，结合图像测量技术获取建筑几何信息；提出了基于模糊推理的砌体结构隐蔽信息确定方法，可有效提取村镇砌体结构墙体属性、构造措施、材料强度等隐蔽属性信息；依据获取的建筑信息与场地信息，分别建立单体结构力学模型与空间地震动场模型，通过结构地震动力响应分析实现村镇区域砌体建筑群的震害预测。最后，对北方地区某村落进行群体建筑信息获取和震害预测，结果表明,基于影像数据的图像识别和图像测量技术可以有效获取村镇建筑结构类型以及外观尺寸信息,提出的模糊推理方法可以高效、准确地识别砌体建筑群的隐蔽信息，进一步通过对震害场景的数值模拟，实现村镇区域群体建筑震害的有效预测。

Intelligent information acquisition methods and seismic damage prediction of rural masonry building groups

Masonry structures (MS) are numerous and widely distributed in rural areas of China, and these structures have features such as generally poor seismic performance, variety of structural forms, insufficient data on mechanics analysis and randomness. Building information is the precondition and basis for seismic damage predictions (SDP), while traditional methods by manual field survey are inefficient and uneconomical. In this paper, an intelligent method combining fuzzy inference and image recognition was proposed to get building information. An image recognition model for MS based on convolutional neural network was established, and image measurement technology was used to obtain geometric information. The hidden information identification method for MS was given based on fuzzy inference model, which can effectively obtain the hidden information such as wall thickness, seismic detailing and material strength. The mechanics model of building and spatially variable ground motion model were established based on building information and site information, respectively. The SDP was then conducted through structural seismic response analysis. Finally, the accuracy and efficiency of the proposed method were verified by a numerical example of MS groups in a village located in north China. The results show that the image identification and measurement techniques based on photos can effectively obtain the type of building structure and geometric information. The proposed fuzzy inference method can efficiently and accurately identify the hidden information of the masonry buildings, and seismic damage of the rural buildings can be effectively predicted through the numerical analysis.