改进SOFM网络算法及其在裂隙统计分析中的应用

A IMPROVED SOFM NEURAL NETWORK ALGORITHM AND APPLICATION

在分析 SOF M自组织特征映射神经网络原有算法的基础上 ,从提高算法收敛速度出发 ,提出了一种改进算法。该算法首先采用 C均值法从输入数据中寻找出 N 2个聚类中心 ,然后用一种启发式的方法把选取的 N2个数据点放置到一个 N× N 的空阵列中。利用这种算法 ,可以避免传统 SOF M算法中不断地用大量的数据去调整连接权的过程 ,从而快速地构造特征映射。应用这种算法 ,通过对某隧道工程围岩裂隙统计数据的快速分类、仿真判别 ,为围岩渗透性评价计算提供精确程度较高的量化依据 ,取得了较好的效果。

A new developed algorithm for Self-Organizing Feather Map is presented. the C-means algorithm is used to select N 2 cluster centers from a data set and a heuristic strategy is employed to organize the N 2 selected data points into an N × N neural array so as to form an initial feather map. By this method, a topologically ordered feather map would be formed very quickly instead of requiring a huge amount of iterations to fine-tune weights, which usually happened in the conventional SOFM algorithm. For the application of this approved arithmetic to GELESHAN tunnel rock mass fracture statistic, it was used to classify and simulate the fracture system and acquired a sound consequence.