垃圾填埋场稳定影响因素敏感性神经网络分析

垃圾填埋场稳定影响因素的选取和研究是稳定分析的重要组成部分，而确定主次要因素的方法就是进行影响因素敏感性分析。利用BP神经网络结合正交试验设计方法，对填埋场沿衬里滑移破坏的影响因素进行敏感性分析。通过对工程实例的计算，该方法与传统法有着相同的因素敏感性排序，但决定排序的极差值有所不同。由于该法能考虑多因素同时变化，故其极差值的大小更能反映因素敏感程度的差异。BP神经网络结合正交试验设计法与传统法相比能节省计算工作量。研究结果显示，影响填埋场滑移稳定性的最主要参数为复合衬垫层间最小摩擦角和垃圾的堆填坡角。因素敏感性排序的确定为填埋场的优化设计、施工、试验、运营管理提供了理论依据。

Sensibility analysis of factors aeffecting stability of waste landfill based on neural network

The selection and research of stability influence factors are very important for stability analysis of solid waste landfills. The sensibility analysis is a method to identify the importance of the factors. In the traditional method of sensibility analysis, only the results caused by changing single affecting factor are considered. Therefore, it is difficult to reveal the real picture of the multi factors simultaneously changed in real case. A method of BP artificial neural network combined with the orthogonal design is presented to study the sensibility of influence factors on translational failure of landfill along the liner system. The comparison between the new and traditional methods is conducted by analyzing a real project. It shows that the sequences of factor sensibility are the same but the maximum difference of safety factors for an influence factor used to determine the sequences are different. The maximum difference of safety factors in the new method are more accurate to reflect the differences of sensibility of the factors because several factors simultaneously changed are considered. In addition, the computing and working time for using the new method is much less than for using the traditional method. The results of sensibility analysis indicate that the major factors to impact landfill stability against translational failure include the smallest interface friction angle in the composite liner and the waste filling slope angle. The results of sensibility analysis can be used to optimize the design, construction, monitoring, and operation management for solid waste landfills.