Due to the influence of the rock mass structure, ground stress, groundwater conditions and construction process, the distribution of the strength and stress of surrounding rock in the soft rock tunnel is nonuniform. The supporting structure may undergo nonuniform deformation and local damage, which has a considerable impact on the safe construction of the tunnel. In this paper, two reference indexes, basic deformation grade and deformation nonuniformity grade, are defined to classify the basic deformation and deformation nonuniformity of an excavation section. The influencing factors of the nonuniform deformation are reduced using the Fuzzy Delphi- Rough Set and then used as the input parameters of a back-propagation neural network (BPNN). Taking the average relative deformation and deformation nonuniformity coefficient as the output parameters, the BPNN model for the nonuniform deformation of the soft rock tunnel is established and verified by actual engineering data. In this study, the influencing factor weights of the nonuniform deformation of the soft rock tunnel are quantified by combining the subjective and objective weight calculation methods. The prediction results of the BPNN after the factor reduction are consistent with the actual results. According to the prediction grade of the basic deformation and deformation nonuniformity, the excavation method and basic support strength, and the abnormal deformation support strength of the tunnel can be optimized, respectively; this approach can provide targeted guidance for planning the safe construction of soft rock tunnels.

受岩体结构，地应力，地下水条件和施工过程的影响，软岩隧道围岩强度和应力分布不均匀。支撑结构可能会发生不均匀的变形和局部破坏，这对隧道的安全施工产生了相当大的影响。本文定义了两个参考指标，基本变形等级和变形不均匀等级，以对开挖断面的基本变形和变形不均匀进行分类。使用Fuzzy Delphi-Rough Set减少不均匀变形的影响因素，然后将其用作反向传播神经网络（BPNN）的输入参数。以平均相对变形和变形不均匀系数为输出参数，建立了软岩隧道非均匀变形的BPNN模型，并通过实际工程数据进行了验证。本文结合主客观权重计算方法，对软岩隧道非均匀变形的影响因素权重进行了量化。因子减少后BPNN的预测结果与实际结果一致。根据基本变形和变形不均匀性的预测等级，可以分别对隧道的开挖方法和基本支撑强度以及异常变形支撑强度进行优化。这种方法可以为规划软岩隧道的安全施工提供有针对性的指导。