Loess tunnels are prone to collapse during construction. The risk assessment of loess tunnel collapse is therefore of great significance to prevent collapse accidents. An attribute recognition model of loess tunnel collapse risk assessment is established based on attribute mathematical theory. Several risk assessment indexes, including surrounding rock lithology, topography, excavation span, burial depth, groundwater, rainfall, and construction technology and management level, are selected according to the statistical analysis of 62 cases of loess tunnel collapse. The indexes are described using qualitative and quantitative methods, and the grading standards for each index are determined. The frequency statistic method is used to compute the evaluation index weights. Attribute measurement functions are established to calculate the single-index and synthetic attribute measures, and a confidence criterion is adopted to determine the loess tunnel collapse risk level. The constructed attribute recognition model is used to evaluate the collapse risk of four loess tunnel projects. The results are consistent with the actual construction conditions, which verifies the reliability of the evaluation method and provides a new method to evaluate the collapse risk of loess tunnels during construction.

黄土隧道在施工过程中容易发生坍塌。因此，黄土隧道塌方风险评估对于预防塌方事故具有重要意义。基于属性数学理论建立了黄土隧道坍塌风险评估的属性识别模型。根据对62例黄土隧道坍塌事故的统计分析，选取围岩岩性、地形地貌、开挖跨度、埋深、地下水、降雨量、施工技术和管理水平等多项风险评价指标。采用定性和定量的方法描述指标，并确定每个指标的分级标准。采用频数统计方法计算评价指标权重。建立属性测度函数计算单指标和综合属性测度，采用置信度准则确定黄土隧道坍塌风险等级。运用构建的属性识别模型对4个黄土隧道工程的坍塌风险进行评价。结果与实际施工条件一致，验证了评价方法的可靠性，为黄土隧道施工期间坍塌风险评价提供了一种新方法。