Engineers have developed an optimization-based computational framework to predict the ultimate vertical bearing capacity of foundation piles installed in high-fill mine spoil-dump sites using incomplete test measurements. In these challenging geotechnical environments, conventional Osterberg-cell (O-cell) pile load tests are routinely terminated before a test pile reaches its ultimate bearing state. This early termination is typically driven by the inherent loading-capacity limitations of the internal testing apparatus and the strict operational requirement to avoid destructive failure on active engineering piles, preventing engineers from directly observing true ultimate capacity.
To overcome this measurement constraint, the research team implemented a back-analysis method utilizing the limited-memory Broyden–Fletcher–Goldfarb–Shanno algorithm with box constraints (L-BFGS-B). The formulation relies on fundamental load-transfer theory to numerically simulate the interaction between the pile and surrounding soil strata under axial stress. By constraining parameter ranges within realistic engineering boundaries, the optimization routine systematically solves for geotechnical resistance values that match the available field observations without requiring physical pile failure.
The mathematical representation incorporates a depth-dependent ideal elastoplastic model to capture shaft friction resistance along the length of the pile, alongside a bilinear hardening model designed to describe end-bearing resistance at the pile toe. Through iterative inversion, the algorithm back-analyzes soil-layer-specific coefficients governing depth-dependent shaft stiffness and their associated elastic-limit displacements at the pile–soil interface, while simultaneously evaluating toe initial stiffness, hardening stiffness, and base elastic-limit displacement.
The parameters are calibrated within the prescribed bounds by minimizing the mean squared residual between measured displacement responses from O-cell field monitoring and calculated analytical values. Once these underlying interface properties are accurately determined, the framework reconstructs the complete load–displacement behavior for both the upper and lower pile segments across the full operational loading spectrum.
Following segment reconstruction, the researchers utilized the equal-displacement method to convert the segmented bi-directional O-cell responses into an equivalent conventional pile-head load–displacement curve. This synthesized top-down curve enables direct derivation of the ultimate vertical bearing capacity as if standard static load testing had been executed to failure in the field.
To validate the predictive accuracy of the inversion technique, the authors established ABAQUS numerical simulation models representing both the internal O-cell configuration and standard conventional static load tests. Comparative analysis confirmed that the proposed analytical approach effectively reproduced the observed O-cell responses, while the back-analyzed axial-force distributions closely matched numerical simulation benchmarks.
The predicted ultimate vertical bearing capacity demonstrated close agreement with the conventional static-loading numerical simulation, yielding a relative deviation of approximately 5.9%. The researchers concluded that the L-BFGS-B inversion framework provides an efficient engineering methodology for extracting load-transfer parameters from incomplete O-cell testing datasets and determining single-pile bearing performance in high-fill spoil dumps.
Subject of Research: Prediction of foundation pile vertical bearing capacity in high-fill areas using L-BFGS-B inversion
Article Title: Prediction of single-pile bearing capacity in high-fill areas from O-cell tests using L-BFGS-B inversion
Article References: Xiao, Y., Chen, M., Wang, J., Zhi, C., Wei, H., Zhao, H., & Liu, C. (2026). Prediction of single-pile bearing capacity in high-fill areas from O-cell tests using L-BFGS-B inversion. Scientific Reports. https://doi.org/10.1038/s41598-026-80176-y
Image Credits: AI Generated
DOI: 10.1038/s41598-026-80176-y
Keywords: pile load tests, O-cell testing, bearing capacity, L-BFGS-B algorithm, load-transfer theory, high-fill areas, spoil-dump sites, geotechnical engineering, elastoplastic model, bilinear hardening, pile shaft resistance, pile toe resistance, equal-displacement method, ABAQUS simulation, numerical modeling, displacement residual
Cite Scienmag News
Denise Maddox. (November 22, 2026). Optimization Algorithm Predicts Pile Bearing Capacity from Incomplete Load Test Data. Scienmag. https://scienmag.com/optimization-algorithm-predicts-pile-bearing-capacity-from-incomplete-load-test-data/
Denise Maddox. "Optimization Algorithm Predicts Pile Bearing Capacity from Incomplete Load Test Data." Scienmag, 22 November 2026, https://scienmag.com/optimization-algorithm-predicts-pile-bearing-capacity-from-incomplete-load-test-data/. Accessed 22 November 2026.
Denise Maddox. "Optimization Algorithm Predicts Pile Bearing Capacity from Incomplete Load Test Data." Scienmag. November 22, 2026. https://scienmag.com/optimization-algorithm-predicts-pile-bearing-capacity-from-incomplete-load-test-data/








