Optimization of Retaining Walls Using Genetic Algorithm and Neural Network Hybrid

In the present research, it has been tried to analyze the wall economic aspects based on the structural designing viewpoint using the optimization procedure i.e. the geneticalgorithm (GA), such that full control on stability features of retaining wall including overturning, sliding and loading capacity is obtained along with wall member highstrength against the imposed forces (bending, shearing, etc.). The Mononobe-EKB semistatic method has been applied to determine the wall imposed forces. A computerprogram has also been developed which is able to optimize the retaining walls withrespect to backfill material properties, the base backfill, the design seismic conditions and backfill overload; even with various limitations imposed on the selection of walldimensions, the program facilitates improving the volume of consumed concrete, and it determines the wall optimized dimensions so it is quite possible to control the wallstructure stability and strength. For engineers to use more conveniently from the proposeddimensions, an analysis is undertaken using various soils in the back of wall and the base deck with different seismic conditions and backfill shapes in variety of heights, and theanalysis results are presented in tabular and relational forms in the research through which it is easy to abstain the concrete optimal volume and dimensions without directreference to computer program. In the research also the retaining wall dynamic analysis isundertaken according to the nonlinear strain–stress analysis method with the use of Plaxis software and then the optimized dimensions are calculated with the neural network andgenetic algorithm hybrid for comparing both methods results and obtaining the optimized dimensions and calculated volumes differences between the statistic analysis (Mononobe- EKB method) and dynamic analysis.

Keywords:

Retaining wall , Genetic algorithm (GA) , Neural network , Optimization , MononobeInternational EKB Theory , Plaxis software