论文部分内容阅读
利用Fisher判别分析(FDA)和支持向量机(SVMs)等来识别地下矿山矿柱稳定性,从多种煤矿和石材矿山中提取一些指标和力学参数作为识别因子,包括矿柱宽度、高度、矿柱的高宽比、岩石单轴抗压强度和矿柱应力。包括取样、训练、建模和评估4个主要步骤。在建模阶段,基于统计学习理论,建立两类矿柱稳定性预测的FDA和SVMs模型,以40组世界不同矿山的实测数据进行模型的训练和测试,并将其模型应用于其他6组待测样本来验证建立模型的有效性,将SVMs模型预测结果与FDA模型及实际情况进行对比,采用指标回代估计法和交叉验证法来考察模型的识别能力。研究表明,SVMs和FDA模型都能较好地预测矿柱的稳定性,但SVMs的优势更明显,有望成为一种可靠、实用的地下矿山矿柱稳定性的评价工具。
Fisher discriminant analysis (FDA) and support vector machines (SVMs) are used to identify the stability of underground mine pillars. Some indexes and mechanical parameters are extracted from various coal mines and stone mines as identifying factors, including pillar width, height, Aspect ratio, uniaxial compressive strength of rock and pillar stress. Including sampling, training, modeling and evaluation of four major steps. In the modeling stage, based on the statistical learning theory, two types of FDA and SVMs models for predicting the stability of pillars were established. The models were trained and tested with the measured data from 40 different world mines. The model was applied to the other 6 groups The samples were used to verify the validity of the model. The prediction results of SVMs model were compared with the FDA model and the actual situation. The index back-estimation method and cross-validation method were used to examine the recognition ability of the model. The results show that both SVMs and FDA models can predict the stability of pillars well, but the advantages of SVMs are more obvious. It is expected to become a reliable and practical tool to evaluate the stability of pillars in underground mines.