Intelligent Optimization of Reinforcement Design Using Evolutionary Artificial Neural Network for the Muzishu Landslide Based on GIS

Abstract

Reinforcement measures are inevitably taken for most landslides in the reservoir area of Three Gorges, China. The reinforcement designs are often based on numerical analysis and land-use planners’ experience, which will make them conservative in most cases. Obtaining a global optimum design which can guarantee both the landslide stability and lower costs of remedial works is a significant problem. This chapter presents a new method for the optimization of reinforcement design for landslides that is an integration of evolutionary artificial neural network algorithm, numerical analysis and GIS. Artificial neural network is used to build the nonlinear relationship between the parameters of reinforcement measures, factor of safety (FOS), and engineering cost; and network structure is optimized by genetic algorithm. Numerical analysis is used to create training and learning examples, and GIS technique plays the role of result visualization and data provision for numerical analysis. During the optimization procedure, the engineering cost is taken as a fitness function of genetic algorithm and FOS is regarded as a constraint condition. Accordingly, the optimized parameters of reinforcement measures, such as pile space, length, and the geometry and size of slide-prevention piles, will be obtained. Based on these optimized parameters and the intelligent prediction model, the reinforcement design will directly lead to more economical engineering costs. Finally, the reinforcement design will be visualized in a three-dimensional strata model of a landslide, which can contribute to land-use planners’ synthetic decision-making. In addition, some integrating geological sections and local strata model can be generated in the GIS platform which will be used for numerical analysis for landslide stability. The proposed method is applied to the Muzishu landslide in the reservoir area of Three Gorges, China. The results provide a satisfactory optimum design which makes it possible to significantly reduce engineering costs.