The Application Of BP Neural Network On Analysis Of The Observed Data Of Dam

A back-propagation network of predicting dam displacement is presented by applying the BP neural network theory to analyzing the dam prototype observation data in this dissertation. First, the determination of independent variable relativity according to partial relative coefficient are proposed so that input factors of BP network can be confirmed. Second, an improved algorithm which can speed computation and improve the resolution is put forward on the basis of modeling idea of neural network. Third, a hybrid BP algorithm which combines genetic algorithm and BP model is put forward. The comparison of computation results between the above-mentioned new methods and the traditional regression method is conducted 妙 displacement analysis of an earth dam and a concrete dam. It is found that the former is better than the latter in displacement fit as well as in displacement prediction. It states that the improved BP algorithm and the hybrid BP algorithm are adapted well to treating the prototype observation data of a dam.