Application of Artificial Neural Networks (ANNs) to Predict the Behavior of Reinforced Linear Pavements with Geosynthetics -Comparison with Finite Element Method-

Abstract

Basic infrastructure, such as roads, engineering structures and airport runways, is considered to be the backbone of any country's transport system and a key factor in economic stability, given the services it provides to meet the needs of the population. Algeria has a very extensive network of roads and runways, most of which are subject to repetitive loads from heavy goods vehicle and aircraft traffic, with dynamic effects. As a result, deterioration and crack propagation can be observed in the wearing courses, leading to a loss of load-bearing capacity in the pavement bodies of these runways, requiring repairs and/or reinforcement as part of routine or ongoing maintenance. The solution of inserting geosynthetic layers has shown very convincing advantages during its application (over the last thirty years). However, there is still little (if any) monitoring of the behaviour of any runway or road reinforced with this type of material. This is why experts and researchers in the field of geotechnics are thinking about finding solutions and predicting the behaviour over time of these infrastructures. This master's thesis proposes a contribution to the understanding of the prediction method known as "Artificial Neural Networks / ANNs", by collecting databases from the practical experience of companies, laboratories and BETs, as well as the results of researchers in this field. The results obtained from the application of this method using Python programming language will be compared with the results obtained using the finite element method using Ansys/Workbench software. We will then be able to determine the advantages and disadvantages of the two methods, ANNs and FEM, and draw the appropriate lessons.