Modelo de previsão de recalques em estacas hélice contínua, metálicas e escavadas através de redes neurais artificiais

Abstract

This work presents a method for pile settlement prediction using an artificial neural network (ANN) known as perceptron. Results from 199 SPT and static load tests carried out on continuous flight auger, steel and bored piles were used in the model development. The ANN model allows the prediction of settlements as a function of the pile type and geometry (diameter and length), the stratigraphy and characteristics of soils defined by SPT test results, and the applied load. The architecture that presented the best performance was A10:14:8:4:2:1, with a correlation coefficient of 0.94. The results were compared to those obtained in previous works and by other authors that also developed settlement prediction models using ANNs. This comparison has shown that the proposed model is able to predict settlements more accurately.