This paper presents intelligent models for solving problems related to retaining walls in geotechnics. To do this, safety factors of 2800 retaining walls were modeled and recorded considering different effective parameters of retaining walls (RWs), i.e., height of the wall, wall thickness, friction angle, density of the soil, and density of the rock. Two intelligent methodologies including a pre-developed artificial neural network (ANN) and a combination of artificial bee colony (ABC) and ANN were selectively developed to approximate safety factors of RWs. In the new network, ABC was used to optimize weight and biases of ANN to receive higher level of accuracy and performance prediction. Many ANN and ABC–ANN models were built considering the most influential parameters of them and their performances were evaluated using coefficient of determination (R2) and root mean square error (RMSE) performance indices. After developing the mentioned models, it was found that the new hybrid model is able to increase network performance capacity significantly. For instance, R2 values of 0.982 and 0.985 for training and testing of ABC–ANN model, respectively, compared to these values of 0.920 and 0.924 for ANN model showed that the new hybrid model can be introduced as a capable enough technique in the field of this study for estimating safety factors of RWs.

A combination of artificial bee colony and neural network for approximating the safety factor of retaining walls

A series of physical model tests was conducted for the active case of a rigid retaining wall subjected to horizontal translation. The behaviour of a granular retained soil was investigated experimentally using a set of precise miniature pressure cells and particle image velocimetry. The good agreement between the experimental results and the arch-action-based theories for lateral active earth pressure could confirm the arching effect behind the retaining walls in the active translation mode. The distribution of shear strain confirms that, for active wall movement, the failure zone is distinguished from the stationary zone by a shear band behind the wall.

Experimental analysis of earth pressure against rigid retaining walls under translation mode

This paper reports on the magnitudes and distribution of static at‐rest stresses behind a rigid wall as a function of soil densification and on static active stresses mobilized behind a rigid wall rotating about its base. The experiments were conducted on the retaining wall system which is permanently affixed to the University of Washington shaking table which is described. Based on these experiments on Ottawa sand, the following conclusions are drawn: (1) The stress distribution behind a non‐yielding rigid wall is hydrostatic; (2) the well known Jaky equation applies only when the backfill is deposited at its loosest state; and (3) when the backfill behind the wall is either compacted or vibrated to increase its density, the magnitude of the at‐rest stresses increases due to densification and the total at‐rest stress exerted on the wall will then be the sum of the stresses due to gravity effects and the locked‐in horizontal stresses due to densification. For a rigid wall rotating about the base: (1) The ...

Ka and ko behind rotating and non-yielding walls - closure

The stability of retaining walls against overturning is analyzed in this study using artificial intelligence methods. Five input parameters including wall height, wall thickness, soil friction angle, soil density, and stone cement mixture density were varied and 2000 cases were considered in developing the predictive models. Using the artificial neural network (ANN) method, eight prediction models were developed and evaluated based on the coefficient of determination (R2) and the root mean square error. R2 values of 0.9740 and 0.9824 for training and testing datasets, respectively (for the best model), indicate the level of ANN capability in predicting safety factor (SF) of retaining walls. After developing the ANN model, the ant colony optimization (ACO) algorithm was used to maximize the safety factor of the wall by varying the input parameters. In fact, the best ANN model was selected to be used as a modeling function in ACO algorithm. The SF result from optimization section was obtained as 3.057 which show a significant difference from the mean SF values used in the modeling. It can be concluded that ACO may be used as a powerful optimization algorithm in optimizing SF results of retaining walls.

The use of new intelligent techniques in designing retaining walls

An important task of geotechnical engineering is a suitable design of safety factor (SF) of retaining wall under both static and dynamic conditions. This paper presents the advantages of both prediction and optimization of retaining wall SF through artificial neural network (ANN) and artificial bee colony (ABC), respectively. These techniques were selected because of their capability in predicting and optimizing science and engineering problems. To gain purpose of this research, a comprehensive database consisted of 2880 datasets of wall height, wall width, wall mass, soil mass and internal angle of friction as input parameters and SF of retaining wall as output was prepared. In fact, SF is considered as a function of the mentioned parameters. At the first step of modeling, several ANN models were constructed and the best one among them was selected. The coefficient of determination (R2) value of 0.998 for both training and testing datasets was obtained for the best ANN model which indicates an excellent accuracy level in predicting SF values. In the next step of modeling, the results of selected ANN model were used as an input for the optimization technique of ABC. In general, 11 models of ABC optimization with different strategies were built. As a result, by decreasing wall height value from 10 m to 8 m and 5.628 m and using almost constant values for the other input parameters, SF values were obtained as 2.142 and 5.628, respectively. Results of (8.003, 0.794, 0.667, 1800 and 2800) and (5.628, 0.763, 0.660, 1735 and 2679) were obtained for wall height, wall width, internal friction angle, soil mass and wall mass of the best models with 2.142 and 5.628 SF values, respectively.

Estimating and optimizing safety factors of retaining wall through neural network and bee colony techniques

Classical earth pressure theories are valid strictly for retaining walls subject to uniform free translation. Practically all retaining walls rotate, and movements of the wall could be restricted, particularly under working conditions. The lateral earth pressure on the wall often deviates from the fully active Coulomb value. There is a need for predicting the lateral earth pressure at any wall displacement behind a rotating wall. The finite element method (FEM) is capable of providing valid solutions of lateral pressures for different wall movements, but a simple alternative method has its practical value. This paper presents a simple analysis method using a modified Coulomb's solution of active pressure. The deformation pattern and the associated mobilization of shearing resistance in the soil as affected by the wall movement are considered in a simplified manner. Comparisons of calculated results with solutions from FEM and observations from model tests show that the method can provide a good prediction of lateral pressures for walls rotating about the base when proper distributions of mobilized shearing resistance and wall friction are used. For walls rotating about the top, the prediction is fair due to arching and the difference between assumed and observed rupture mechanisms.

Ming-Fang Chang

Papers

Lateral earth pressures behind rotating walls