Modeling Stress-Strain Behavior of Unsaturated Soils Using A Genetic Algorithm-Based Neural Network

Modeling of unsaturated soils has been the subject of many research works in the past few decades. A number of constitutive models have been developed to describe the complex behavior of unsaturated soils. Despite the complexity of the existing constitutive theories, none of the existing constitutive models can predict various aspects of behavior of unsaturated soils. In this paper a new approach is presented, based on integration of a neural network and a genetic algorithm, as a unified approach to modeling of unsaturated soils. A computer program coded in visual basic was developed and used for prediction of the stress-strain behavior of unsaturated soils. In the proposed approach genetic algorithm is used to optimize the weights of the neural network. The topology of the neural network is determined by trial and error. The network has three layers with five input neurons, namely, initial gravimetric water content, initial dry density, axial strain, mean effective stress with respect to pore air pressure, soil suction and volumetric strain, five neurons in the hidden layer and three neuron in the output layer namely, deviatoric stress, suction and volumetric strain. The network was trained and tested using a database including results from a comprehensive set of triaxial tests on unsaturated soils from the literature. Neural network simulations were compared with the experimental results. Comparison of the results indicates the roposed approach is very effective and robust in modeling the stress-strain behavior of unsaturated soils.

کلیدواژه ها:

unsaturated soil ، neural network ، genetic algorithm ، stress-strain relationship