基于SOFM神经网络模型的岩爆烈度等级预测方法

Prediction method of rockburst intensity grade based on SOFM neural network model

为简化岩爆烈度等级预测指标体系、解决预测分级模糊问题、利于工程人员分析预测结果,建立基于SOFM神经网络的岩爆烈度等级预测模型,并根据竞争层拓扑结构的不同将预测模型拓展成3个模型。将硐壁最大切向应力、岩石单轴抗压强度、岩石单轴抗拉强度作为模型输入向量,将40组国内外岩爆工程数据作为数据集输入3个模型进行训练、测试,3个模型在测试集上岩爆烈度等级预测正确率均达到90%。比较3个模型的聚类、测试及训练效果,得到竞争层神经元个数为16的预测模型最优。将最优预测模型的预测结果与可拓理论、Russenes判据、基于模糊C-均值算法粗糙集理论云模型的预测结果对比,基于SOFM神经网络模型的岩爆烈度等级预测方法优于其他方法,表明该方法具有一定可行性和实用性,为岩爆预测提供了一种新的方法和手段。

In order to simplify the rockburst intensity grade prediction index system,solve the problem of fuzzy classification of prediction,and help engineers to analyze the prediction results,a rockburst intensity grade prediction model based on SOFM neural network was established.The prediction model was expanded into three models according to the different topology of the competition layer.Taking the maximum tangential stress,the uniaxial compressive strength and the uniaxial tensile strength of the rock as the model input vectors,40 sets of rockburst engineering data at home and abroad were used as data sets to input three models for training and testing.The results of the testing indicated that the prediction accuracy of the three models was up to 90%.Through comparing the clustering,testing and training effects of the three models,the prediction model with 16 neurons in the competition layer was the best.Compared the prediction results of best prediction model with that of extension theory,Russenes criteria and cloud model based on rough set of FCM algorithm,the prediction method of rockburst intensity grade based on SOFM neural network was superior to other methods.It shows that the method proposed in this paper is feasible and practical,which provides a new method and means for rockburst prediction.