.png){kind=icon}
.png){kind=icon}
Return
Data-driven tabular foundation model for rock yield strength criterion
.png){kind=icon}
OA
AI
H
B
S
DOI：10.1016/j.mtcomm.2026.115571.png)
.png){kind=icon}
.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Understanding the rock yield strength mechanism is essential for stability analysis, reinforcement design, and safe construction in rock engineering, particularly in numerical simulations. However, building a reasonable rock yield strength criterion for describing the rock failure mechanism is difficult owing to the rock type variability, complexity of the stress state, and extreme rock environment. This study developed a novel data-driven rock yield strength criterion by incorporating the tabular foundation model and true triaxial compressive test data. A synthesized dataset was generated using the generalized Zhang-Zhu (GZZ) model. Based on a synthetic dataset, the Tabular Prior-data Fitted Network (TabPFN), a tabular foundation model, was leveraged to approximate the yield function and describe the rock failure strength criterion. The developed yield criterion was illustrated and validated by the true triaxial compressive test. The results were compared with the actual laboratory test and GZZ model. The correlation coefficient between the actual yield surface and the one predicted by the developed criterion is 0.9999. The correlation coefficient and the mean absolute error between the actual first principal stress and that predicted by the developed criterion are 0.9663 and 35.4419 MPa, respectively. The results showed that the developed model characterized the complex, nonlinear yield function well and revealed the yield mechanism underlying the intact rock data. This provides a scientific and promising approach for characterizing the strength properties and understanding the failure mechanism of intact rock.
Keywords:
Intact rock
Strength criterion
Data-driven
Tabular foundation model
Prior-data Fitted Networks (PFN)
AI Summary
Key information extracted from the uploaded paper, including a brief overview, abstract, background, key highlights, visual analysis, and future outlook.
Journal
IF:
4.5
Papers:
1.5W
Citations:
3.7W
Organization
Cited Papers