Clustering-Based Fuzzy Model for Predicting Anchor Bearing Capacity

Abstract

Despite the widespread usage of the anchoring technique in geotechnical projects in Brazil, mainly in underground works and retaining walls, limited research has been carried out on the determination of load capacity in anchors. Ground anchored walls in Brazil are often dimensioned by methodologies that are not satisfactorily adapted to the reality of the national soil. In this context, this work proposes a simplified computational methodology for forecasting the load capacity of anchors in geotechnical projects. To this end, a fuzzy model is proposed using specific selected characteristics of the soils and geometry of tieback wall systems, whose measurements are easier to be obtained when compared with the load capacity measurement itself. For validation, calibration, and application of the model, the results of 40 performance tests (load tests) were used. Thirty-two samples were used to train the algorithm whereas eight were used to test the model. The results obtained using the proposed approach were close to the extrapolated values of the performance tests according to the existing empirical method, showing that the model is a viable alternative for the design of anchors. Finally, the methodology presented in this work seems to be a first step for the prediction of the anchor bearing capacity and therefore it will certainly benefit from the addition of new test databases of different locations around the world.