对于急倾斜特厚煤层巷道，对其进行安全合理的支护是实现安全高效开采的技术基础。要得到合理的支护参数，确定合理可靠的支护方式，应该对巷道及煤层进行松动圈监测、钻孔窥视仪监测以及矿压和巷道变形监测。通过对监测数据的分析，得到松动圈范围、巷道壁裂隙发育情况和破碎程度以及巷道围岩矿压和变形规律，从而为巷道支护提供有效可靠的数据参数，为巷道支护设计提供合理依据。人工神经网络是研究影响条件复杂的非线性的巷道变形的有效方法,对煤矿安全生产具有重要意义。本文以神华新疆煤业公司乌鲁木齐矿区碱沟煤矿为背景，对巷道支护和巷道矿压和变形理论进行了研究，选取碱沟煤矿有代表性的巷道和煤层进行了松动圈监测和钻孔窥视监测，得到了巷道的松动圈范围和裂隙发育情况，对支护方式进行了评价并提出了改进意见，也为支护设计提供了重要的参数依据。同时，本文对518B6巷道设计了矿压观测方案并进行观测，在对观测数据处理与分析的基础上，综合各种因素对回采巷道变形的影响，总结了该矿急倾斜特厚煤层回采巷道变形的基本规律，根据这一变形规律对现行的巷道支护形式进行了分析与评价。BP神经网络是研究复杂的非线性的巷道变形的有效方法，本文还研究了BP神经网络的算法及其改进算法，针对碱沟煤矿518B6巷道的实际情况，建立了合适的BP神经网络模型，利用观测到的巷道变形数据对网络模型进行训练，并用训练好的网络对518B6巷道进行了变形预测，预测结果比较准确，从而为研究其它类似巷道的变形提供了新的有效方法。碱沟煤矿煤层赋存情况、巷道布置方式与开采方式与乌鲁木齐矿区其它煤矿类似，对于研究其它位于乌鲁木齐矿区且条件类似的巷道，本文的研究成果也具有借鉴意义。

To support the entry in steep-thick coal seam safely is the technical basis of safely and effectively mining. We must monitor and test the relaxation zone, fracture and the deformation of the surrounding rock using the instruments. By the analysis for monitoring and observation data, we can get the relaxation zone, fracture and the law of the deformation, then these data can be used for the support and they can be used as important and reasonable basis. Artificial Neural Network is a efficient method for the research on the complex and not linear deformation of the entry and safely mining production. This paper base on the Jian Gou coal company of the ShenHua Energy Co.,Ltd in Xinjiang and use relaxation zone testing instrument and sight testing instrument to get the relaxation zone, fracture area for some typical entry or coal seam, then analyze and evaluate the current supply method and offer the improve method. And this paper design the observation scheme for 518B6 entry and observe it, and considering multiple factors to get the deformation law of the entry. Then analyze the supporting method using the law. BP Neural Network is a efficient method for the research on the complex and not linear deformation of the entry, this paper research the BP Neural Network, and base on the Actual situation of the 518B6 entry. Design the appropriate model, then train the model using the deformation data, and forecast the future deformation of the entry. By the verification, forecast data is correct very much. So the BP Neural Network is a new method for the research on deformation of other similar entry. As JianGou coal company is similar with other coal company in Urumqi, the research result is very useful for them.