【摘要】 地下洞室开挖后,洞室将产生一定的变形,若变形速率过大,将影响洞室的稳定性,洞室施工期稳定性通常以隧洞变形总位移量表示的相对位移值为判据,受各种施工因素的影响,监测点并不能完全紧跟掌子面布设,致使洞室开挖后部分重要的前期位移释放值丢失。采用遗传进化算法搜索最优的Elman神经网络结构,大大提高了网络学习和预测能力;并利用Elman映射动态和适应时变特性的能力,以现场实测期数据作为训练和检测样本,建立仿真预测模型,用该模型前推丢失位移值;以前推位移值修正实测值,得到实际位移值,并应用于评价围岩的稳定性。实践表明,该方法不仅简单而且是可行的。更多还原

【Abstract】 The underground workshop stability during construction period was usually evaluated based on the relative variations expressed by the complete displacement of the tunnel deformation.However monitoring points could not always be mounted in time following excavating work face because of many kinds of factors influenced by practical construction which would lead to part of important prophase data lost after the some steps of excavation.In this paper,the optimization network structure of Elman was searched by using SGA with the advantages of its dynamic mapping features and suitable ability varied with time.In this way,the study and prediction abilities of neural network were greatly increased.The monitoring data was used as training and test samples,and a new imitating prediction model was established.With this model the lost displacement could be predicted and the monitoring displacement might be amended to obtain the complete deformation displacement,which was very useful for evaluating the surrounding rock stability.The application results in an underground workshop excavation showed that the prediction accuracy conforms to the need of engineering construction and this method was rational and feasible.更多还原