地下工程围岩分类的神经网络可视化评价

围岩分类是地下工程技术基础研究的重要课题之一。影响围岩类别的因素有很多，并且具有高度的随机性和模糊性。神经网络方法具有自组织、自学习和高度非线性映射的能力，并且既能考虑定量因素又能考虑定性因素，因此，在围岩分类的应用方面，神经网络具有广泛的前景。文章基于改进的BP神经网络的一般原理，依据有关地下工程围岩的分类标准，选取岩体结构、岩石饱和单轴抗压强度、岩体结构面、岩石纵波速度作为围岩分类的评价指标，利用MATLAB语言构建了可视化的围岩分类神经网络模型，并收集了大量的工程资料对网络进行训练和检验。结果表明，网络的预测结果与实际结果比较一致。证明神经网络是能够在工程岩体分类方面得到推广应用的。

Visual evaluation of neural network on classification of surrounding rocks for underground engineering

The classification of surrounding rocks is one of the important research subjects for the underground engineering technique. The factors which affect rock type are random and fuzzy. As neural network method not only possesses abilities of self-organized, self-taught and high non-linear mapping, but also can consider both quantitative and qualitative factors, it has the extensive foreground in the aspects of classification of surrounding rocks. According to the concerning standard of classification of surrounding rock for underground engineering, rock structure, structural plane, saturated uniaxial compression strength, P wave velocity through the rock are chosen as the indexes of criterion for classification of surrounding rocks. Based on the general principle of improved BP network, a visual model is established by MATLAB and engineering data are collected to train and examine the model. The prediced results of ANN keep coincidence wall with the actual ones, and it is proved that neural network may be used widely in classification of engineering rock.