Abstract：As the progress of city modernization, there is an urgently demand for underground engineering. The existence of underground Karst is one of the main reasons of collapse and affects underground engineering construction. Underground karst distribution and collapses are controlled by many factors, so they are very difficulty to predict, especially covered by the sediments of Quaternary system. The distribution of underground karst and karst collapses are affected by groundwater, strata thickness of Quaternary system, dynamic head of groundwater, beside carbonate strata. By analyzing the mainly influence factors of karst distribution and collapses, the calculating steps of neural network method is introduced. As an example of Tangshan city, prediction model is constructed. Tangshan is one of the important industrial estates of Hebei province in China, with extensive distribution of carbonate strata and underground karst, so the disaster of karst collapse is very seriously. The distribution of underground karst in urban is mainly controlled by thickness of Quaternary system, altitude and depth of groundwater, so those factors are considered in prediction model. In the model, the component number of input layer is 3, and the component number of output layer is 1, plus coefficient of learning is 1.2, inertial coefficient is 0.5. There are two connotative layers, the component number of connotative layer 1 is 4, and the component number of connotative layer 2 is 2, the error of calculating is 9%. By iterative calculating of 65993 times, the distribution of karst in Tangshan city is predicted and some suggestions for city construction are given.
收稿日期: 2003-03-31
引用本文:
朱庆杰;苏幼坡;陈静. 基于神经网络模型的地下岩溶分布预测[J]. 岩石力学与工程学报, 2003, 22(S1): 2378-2381.
Zhu Qingjie, Su youpo, Chen Jing. PREDICTION OF UNDERGROUND KARST DISTRIBUTION BASED ON NEURAL NETWORK MODEL. , 2003, 22(S1): 2378-2381.
2003, Vol. 22 
