摘要
应用神经网络理论提出了水平面上矿体自动圈定的新方法。在该方法中建立了由输入层、中间处理层、输出层和训练层构成的，且具有很强映射、联想、推广和概括能力的多层式前向神经网络模型。从待估点附近控制点数据中收集与待估点有关信息，利用已知控制点数据反复对神经网络进行训练，以使神经网络对水平面上矿岩分布特征充分了解。最后利用该神经网络以较小点距逐点对水平面上矿岩分布进行估计，完成矿体自动圈定。
In this paper, the neural network theory is used to put forward a new method of automatic enclosing of orebody on horizontal plane. In this method, a multilayer feedforward neural network model with strong mapping, associating, generalizing and summarizing abilities which includes input layer, middle processing layers, output layer and training layer, is established. For accurate estimation of the information for unknown points,the following information, such as the space position of the point to be estimated, the data about ore and rock on given control points, geologic structure data,etc,is collected from the control points adjacent to the unknown point. The data of given control points are used to train the neural network repeatedly, making the network fully grasping the distribution characterstics of ore and rock. Finally the automatic enclosing of orebody is fulfilled based on the information collected from the estimation of a large number of the points with smaller interval distance.
出处
《矿冶》
CAS
1996年第2期1-9,共9页
Mining And Metallurgy
