With a tunnel in Shandong Province as an example, fuzzy neural network based on T-S model was adopted together with relevant influencing factors to forecast over-excavation of underground openings. Based on the practical engineering situation, 199 groups of data were chosen for this forecast model, among which 179 were used as training samples for training network and 20 as testing samples for validating the forecasting results of the model. It is found based on the calculation that the correlation coefficient, root mean square error and average relative error of the fuzzy neural network of over-excavation forecast based on T-S model were 0.9628, 0.449 and 6.33%, respectively, showing more precise than the forecasting results obtained by BP neural network and regression model, indicating its significance in controlling over-excavation.%以山东某隧道为例,运用基于T-S模型的模糊神经网络,结合相关影响因素对地下硐室超挖进行了预测.预测模型根据工程实际情况选用了199组数据,其中179组数据作为训练样本训练网络,20组数据作为测试样本验证模型的预测结果.通过计算,基于T-S模型模糊神经网络超挖预测的相关系数为0.9628,均方差为0.449,平均相对误差为6.33%.与BP神经网络和回归模型的预测结果进行了比较分析,结果表明基于T-S模型的模糊神经网络预测效果最好,能精确预测地下硐室爆破超挖量,对控制超挖量具有重要意义.
展开▼
