Rock slope deformation is a big problem challenge to engineer and scientist for the design and construction of rock engineering especially in the case of permanent shiplock engineering Three Gorges Project. The authors developed the rsfdANNet with the theory and method of artificial neural networks for the prediction of rock slope deformation. The software and some techniques for learning and global optimization in networks are given in the paper. The strategy for predicting the displacement of high rock slope in practical rock engineering is illustrated. The case study for Three Gorges project shows that it is reasonable and believable both in theoretical and practical.
Rock slope engineering is a comprehensive practical problem to be analyzed. To meet the needs of requirement of some engineering function, for example the reasonable gate operation and prevention of seepage in shiplock, rock slope deformation is a big problem challenge to the engineer and scientist for the design and construction of rock engineering in the case of permanent shiplock engineering TGP. Too many factors may affect the rock slope behavior. It depends on the geoenvironmental situation not only in historical but also in recent construction period. The construction techniques, such as excavation procedure & reinforcement measure are also the major effective influence control the evolution of rock slopes engineering. It is obviously very difficult and even impossible to control the entire factor, which would affect the behavior of rock slope engineering systematically. The traditional design methods are based on mathematical model. Due to the parameters of slope properties are known with difficulty, a problem becomes very complex when we are asked to forecast its evolution. A new method, which is called feedback progressive design or informative construction, appeared recently in rock engineering. Mechanical models are identified thanks to practical experience.