Surrogate-Assisted Reliability, Optimization, and Parameter Back-Analysis in Geotechnical Engineering: A State-of-the-Art Synthesis
Year: 2026
Type: preprint
Abstract: Surrogate metamodeling in geomechanics has reached a critical structural turning point. While legacy approximation frameworks built upon static, linearizing response surfaces historically optimized computational workflows, their structural failure is evident when navigating the high-dimensional parameter spaces, severe multi-physics couplings, and localized non-linear failure envelopes typical of real-world infrastructure assets. A highly rigorous, state-of-the-art critical synthesis is delivered in this paper, breaking away from redundant historical tracking to chart the 2024–2026 data-driven frontier. A definitive, comparative taxonomy of six core metamodel variants is established: Polynomial Response Surfaces (PRSM), Radial Basis Functions (RBF), Kriging (KSM), Artificial Neural Networks (ANN), Support Vector Regression (SVR), and Moving Least Squares (MLSM). Crucially, the limitations of traditional frameworks are exposed, and the paradigm shift toward sequential active learning loops, multi-fidelity information fusion, and Physics-Informed Neural Networks (PINNs) is decoded. The operational execution of these advanced hybrid architectures is dissected across three core pillars: low-probability geotechnical reliability analysis under spatial variability (Random Fields), generative multi-objective design optimization, and real-time parameter inverse back-analysis for digital twin deployment. Finally, current structural bottlenecks — including curse-of-dimensionality limits, numerical noise filtering, and automated hyper-parameter tuning (AutoML) — are critically cross-examined. By bridging the gap between historical foundational mechanics and modern intelligent computing, a definitive strategic roadmap is provided for resilient and predictive geotechnical engineering.
Source: Research Square
Authors: Serges Mendomo Meye, Pieride Mabe Fogang
Language: English
Subfield: Civil and Structural Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7204279787