Advanced framework for interpreting bidirectional pile load tests using nonlinear load transfer models and metaheuristic optimization

Abstract

This study aims to propose an advanced automated framework for interpreting the load-displacement transfer mechanism observed in bidirectional pile load tests (BPLTs) based on nonlinear load transfer (NLT) models and an Forensic-Based Investigation (FBI) algorithm. Instrumented strain gauge data are directly utilized as main inputs in the proposed FBI-optimized NLT framework to facilitate model calibration and interpretation. The efficiency of the proposed framework is confirmed by validating it against experimental results obtained from BPLTs of three drilled shafts socketed into weak sandstone at the Tamsui River Estuary under O-cell testing. The analyzed results demonstrate that the proposed automated framework is highly capable of calibrating the model parameters, and the FBI-optimized NLT models effectively capture critical aspects of the load transfer mechanism as well as the equivalent head-load displacement curves and axial pile capacity. Among soil-pile interface models adopted in the proposed framework, the disturbed state concept (DSC) and generalized nonlinear softening (GNS) models prove most effective and outperform others in representing the stress-strain behavior at soil-pile interfaces. Furthermore, the proposed FBI-optimized NLT models provide better alternatives to simplified, commonly used approaches (i.e., Osterberg and Loadtest) with a comprehensive characterization of the load transfer mechanism to estimate head-down load-displacement curves and axial capacity of piles, where these simplified approaches use simple assumptions.