CAJ | 학술논문

采空区地面塌陷的危险性判别受地质因素、采矿因素等多重因素的影响，各因素往往影响程度不同且部分因素之间又相互联系。为了能够较准确地对采空塌陷危险性进行评估，引入了T-S模糊神经网络模型。以北京西山地区采空塌陷为例，根据塌陷特点，分别选取了地质构造复杂程度、覆盖层类型、第四系覆盖层厚度、覆岩强度、煤层倾角、采深采厚比、采空区埋深、采空区空间叠置层数8项影响因素作为评价指标，并建立了分级标准。将单因素评价指标均匀线性插值作为训练样本，建立了T-S模糊神经网络判别模型。利用训练好的神经网络模型对选取的8处采空区进行评估，结果分别为：Ⅰ、Ⅱ、Ⅲ、Ⅱ、Ⅲ、Ⅱ、Ⅲ、Ⅱ，结果与实际情况吻合。研究表明，利用T-S模糊神经网络研究采空塌陷危险性是可行的。
채공구지면탑함적위험성판별수지질인소、채광인소등다중인소적영향，각인소왕왕영향정도불동차부분인소지간우상호련계。위료능구교준학지대채공탑함위험성진행평고，인입료T-S모호신경망락모형。이북경서산지구채공탑함위례，근거탑함특점，분별선취료지질구조복잡정도、복개층류형、제사계복개층후도、복암강도、매층경각、채심채후비、채공구매심、채공구공간첩치층수8항영향인소작위평개지표，병건립료분급표준。장단인소평개지표균균선성삽치작위훈련양본，건립료T-S모호신경망락판별모형。이용훈련호적신경망락모형대선취적8처채공구진행평고，결과분별위：Ⅰ、Ⅱ、Ⅲ、Ⅱ、Ⅲ、Ⅱ、Ⅲ、Ⅱ，결과여실제정황문합。연구표명，이용T-S모호신경망락연구채공탑함위험성시가행적。
The stability of underground goaf is affected by many factors,especially the conditions of mining and geology.These factors always have different influences,and some of them are interconnected.The above fea-tures bring great difficulty to evaluate the ground collapse risk quantitatively.In order to appropriately evaluate the stability of underground goaf,the T-S fuzzy neural network model was introduced in this paper.According to the ground collapse information of Xishan mining area of Beijing,eight factors influencing the stability of under-ground goaf were selected as the evaluation indexes at first,and then the grading standards were also built up. These factors include the complexity of geological structure,the type of overburden layer,thickness of quaterna-ry cover,the strength of overlying strata,the dip angle of coal seam,the ratio of mining depth and thickness, the depth of underground goaf and the number of underground goaf in space.Based on the training samples which were generated by means of linear interpolation algorithm,the T-S fuzzy neural network model was con-structed.Finally eight new samples of Xishan mining area in Beijing were evaluated by the trained T-S fuzzy neural network model.The results were Ⅰ,Ⅱ,Ⅲ,Ⅱ,Ⅲ,Ⅱ,Ⅲ and Ⅱ,respectively.The results co-incided with the actual situation.The study shows that it is feasible to evaluate the stability of underground goaf by using the T-S fuzzy neural network model.