Statistical and Probabilistic Insights into the Internal Stability of an MSE Wall

Abstract

Mechanically stabilized earth (MSE) walls are widely used to stabilize steep slopes, especially when the crest is subjected to static or dynamic loading. These structures are frequently used in highway and railway structures where they are subjected to dynamic loading and seasonal variation. With rapid urbanization and a growing population, especially in developing countries, the transportation infrastructure is growing at a tremendous speed. Compared with conventional concrete retaining walls, MSE walls emit less carbon dioxide (CO2) and require less area. In addition, the cost of their construction is low, and less earthwork is required. The other advantage of MSE walls in transportation structures is their better performance under dynamic (or seismic) loads and shorter construction time. This paper discusses the pullout behavior of geogrid, which is an important parameter for determining the internal stability of an MSE wall. The internal stability of MSE walls has two major failure types: (1) pullout; and (2) rupture. In this paper, different data-driven algorithms were used to predict the outcome of the geogrid pullout test and to understand the influence of various physical parameters. This paper utilizes statistical models and machine learning (ML) algorithms, for example, correlation analysis, linear regression analysis with one and multiple variables, tree-based ensemble algorithms, artificial neural network (ANN), support vector machine (SVM), and adaptive neuro fuzzy inference systems (ANFIS) for predictive modeling. In total, 201 data sets from geogrid pullout tests were collected from the literature, which considered 12 features. The performance of the developed model was examined using the coefficient of determination (R2) and later compared using a Taylor diagram, Willmott’s index of agreement, the root mean square error standard deviation ratio (RSR), and mean absolute percentage error (MAPE). The performance of the ANFIS was superior to any other model. Then, the probabilistic approach was used to perform parametric analysis and evaluate the internal stability performance of a reinforced earth wall using Monte Carlo simulations.