Data and Code for Adaptive, Physics-Guided and Reliability-Aware Surrogate Modeling for Slope Stability and Failure Response

Description

This dataset contains the numerical data and supporting computational materials used to develop and validate an adaptive, physics-guided and reliability-aware surrogate framework for slope stability and failure-response assessment. The study considers factor of safety together with seven additional numerical-response descriptors related to failure localization, critical-response position and effective-stress conditions. The deposited materials support surrogate-model development, validation and reproducibility of the reported analyses. FLAC3D is proprietary software and is not included in the repository.

Steps to reproduce

Install Python and the required packages listed in requirements.txt. To reproduce the final surrogate, run FINAL_SURROGATE_FRAMEWORK.py using FINAL_150_FOS_DATASET.csv for the FoS branch and FINAL_101_V43_PHYSICS_RICH_MASTER.csv for the seven physics-response models. The framework automatically generates the seven physics-guided features from the six engineering inputs (H, β, c, φ, γ and ru). The FoS model uses StandardScaler followed by RBF-SVR (C = 3.0, gamma = 0.03, epsilon = 0.03), while the seven additional responses are predicted using target-specific ExtraTrees regressors. The supplied trained model can also be used directly for prediction. BLIND_VALIDATION_20_FLAC3D_INPUTS.csv contains the 20 previously unseen forward-validation cases, and BV20_PRED.csv contains their frozen surrogate predictions. The FLAC3D validation runner requires a licensed installation of FLAC3D 7 and the Itasca Python environment. Detailed commands and file descriptions are provided in README.txt.

Institutions

• Central South University

  Hunan, Changsha

Categories

Licence