摘要
影响矿井工作面顶板稳定性的主要因素为岩石介质条件、环境条件以及工程因素3个方面,分类方案中尽可能考虑到上述3个因素,同时,要具有超前性和实用性。本分类方案选用了5个指标,即岩石单轴抗压强度sc、岩石质量指标RQD、煤体抗压强度scm、地下水状况SW和工作面月推进速度。前3个为煤、岩石介质条件,第4个为环境条件,第5个为工程因素。运用MATLAB软件包中的神经网络工具箱,构造5种网络拓朴结构,即5∶5∶1,5∶8∶1,5∶11∶1,5∶15∶1,5∶20∶1,对同一组样本进行学习。通过比较,最优的网络拓朴结构为5∶20∶1。运用选定的最优网络结构,进行了顶板稳定性的动态分类,同时,对影响因素进行了帕累托表意义上的因素分析,按影响程度大小排序,依次为岩石质量指标(0.438 2)、地下水状况(0.198 8)、岩石单轴抗压强度 (0.147 2)、煤体抗压强度(0.122 5)及工作面推进速度(0.093 2)。
The major influence factors on stope roof stability are rock property condition, environmental condition and engineering factor, which should be considered as possible as you can in the scheme of classification of stope roof stability. Five influence factors, such as uniaxial compressive strength of rock, RQD, compressive strength of coal mass, hydrological condition of roof, and advance velocity of stope, are adopted in the classification. The first three factors belong to rock property condition, the fourth belongs to environmental condition, and the fifth belongs to engineering factor. Five network topology structures, such as 5:5:1, 5:8:1, 5:11:1, 5:15:1, 5:20:1, are adopted with back-propagation artificial neural network in MATLAB language. Based on the errors of training samples, the topology of 5: 20:1 is the best one. The stability of roof is classified dynamically with the BP network topology structure of 5:20:1, in which there are five input neural units, twenty hidden neural units, and one output neural unit. The influence factors are also analyzed with Parato list. According to the magnitude of influence degree on roof stability, there are RQD (0.438 2), hydrological condition of roof (0.198 8), uniaxial compressive strength of rock (0.147 2), compressive strength of coal mass (0.122 5), and advance velocity of stope (0.093 2).
出处
《岩石力学与工程学报》
EI
CAS
CSCD
北大核心
2003年第9期1474-1477,共4页
Chinese Journal of Rock Mechanics and Engineering
基金
国家自然科学基金(40172059)
国家杰出青年基金(50025413)
和华北科技学院科研基金资助项目。
关键词
采矿工程
人工神经网络
采场项板
岩石介质
煤体
抗压强度
Compressive strength
Mines
Mining engineering
Neural networks
Rock mechanics
Stability
