Evolutionary Identification of Deformation Dynamics of Geotechnical Structures

摘要

Affected by rheologic property of geo-materials and many other engineering factors, deformation of geotechnical structures is commonly characterized with complex nonlinear dynamic behavior. It is very difficult to model the highly dynamic nature using physical-based approach. Based on the time series analysis theory, combining genetic programming and genetic algorithms, this paper develops a hybrid evolutionary identification scheme to model the deformation dynamics. In this method, genetic programming technique is firstly employed to perform non-parametric evolution of model structures using function tree representation. For each model structure generated, an interpreter is designed to construct the mathematical expression and introduce necessary model parameters. Then, genetic algorithm is used to perform parametric evolution of the internal model parameters. The evolution process is repeated by using genetic operators and the principle of 'survival of the fittest' until find the satisfied model. Applications to the deformation prediction of the high slope relating to the Three Gorges Project permanent shiplock and Xintan landslide are used to demonstrate that the present method can efficiently evolve mathematical models for predicting the dynamic behavior of deformation of geotechnical structures.