Pile in the Unsaturated Cracked Substrate with Reliability Assessment based on Neural Networks

Abstract

In this study, we present reliability assessments for a monopile socketed into a vertical fissured unsaturated substrate correlated with the degree of saturation. The pile was loaded by concentrated lateral force in a three-dimensional space. Procedures were prepared in the numerical analysis software FLAC 3D, fast lagrangian analysis of continua. FISH, a scripting language embedded within FLAC, was used to control the depth of cracks in unsaturated soil and manage a significant number of the calculations. The presented approach expands knowledge about piles loaded laterally and also the considerable influence of environmental changes. In the reliability part of the study, the excessive horizontal displacement of the pile head was examined as an implicit function. The analytical model of vertical water flux in the partially saturated soil was applied in connection with the iterative solution of the cracked depth of the substrate. The suction profile was estimated for a layer above the groundwater level. An important aspect of this paper is that the reliability analyses were prepared for this task in complex environmental conditions and with a horizontal load. All limit state functions were approximated following the response surface methodology, which was defined by neural networks based on modified perceptrons and was compared with well-known polynomials functions.