By Tinoco, J.; Gomes Correia, A.; Cortez, P.
Advances in Transportation Geotechnics II - Proceedings of the 2nd International Conference on Transportation Geotechnics, ICTG 2012
Jet Grouting (JG) technology, one of the most efficient soft soils improvement methods, has been widely applied in important geotechnical works due to its versatility. However, there is still an important limitation to overcome related with the absence of rational approaches for its design. In the present work, three different Data Mining (DM) techniques, i.e., Artificial Neuronal Networks (ANN), Support Vector Machines (SVM) and multiple regression are trained in order to predict elastic young modulus (E0) of JG mixtures. It is shown that the complex relationships between E0 and its contributing factors can be learned using DM tools, particularly by SVM and ANN algorithms. By performing a detailed sensitivity analysis, understandable knowledge is extracted from the trained models, in terms of the relative importance of the attributes, as well as its effect in E0 prediction. In addition, the mathematical expression proposed by Eurocode 2 to estimate concrete stiffness, is adapted to JG material. Its low performance is assessed and compared with those achieved by DM models.