⚡ electrical engineering

A Simplified Method for Seismic Fragility Analysis of Geogrid-Encased Stone Column Composite Foundations

This study proposes a machine learning-based probabilistic framework to analyze the seismic fragility of geogrid-encased stone column foundations, revealing that native soil shear strength and vertical seismic coefficients are the most critical factors influencing failure probability and guiding design toward high-strength geogrids, adequate footing width, and high area replacement ratios.

Original authors: Xiaocong Cai, Ling Zhang, Jinpeng Tan, Zijian Yang

Published 2026-08-26

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Xiaocong Cai, Ling Zhang, Jinpeng Tan, Zijian Yang

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

In the quiet, unsteady ground beneath many cities, soft soil poses a persistent challenge for engineers. This material, often found in river valleys and coastal plains, is heavy, squishy, and prone to turning into a liquid-like slurry when shaken by an earthquake. To build safely on such ground, engineers often install stone columns: vertical pillars of crushed rock driven deep into the earth to stiffen the soil and carry the weight of buildings. However, in extremely soft ground, these stone pillars can sometimes bulge outward and fail under pressure, much like a sausage casing bursting when squeezed too hard. To solve this, engineers wrap the stone columns in a strong, grid-like fabric called geogrid. This fabric acts as a tight belt, holding the stone together and allowing the foundation to withstand much greater forces. While the strength of these reinforced foundations under normal conditions is well understood, their behavior during the violent shaking of an earthquake remains a complex puzzle. The ground does not just shake side-to-side; it also jolts up and down, creating a chaotic mix of forces that can overwhelm even the strongest designs.

A team of researchers at Hunan University has developed a new way to predict how likely these reinforced foundations are to fail during an earthquake. Instead of relying on a single, rigid calculation that assumes all soil and materials are perfect, the team created a flexible framework that accounts for the natural variations found in the real world. They combined a proven method for calculating the maximum weight a foundation can hold with a powerful computer technique known as machine learning. This approach allowed them to run thousands of simulations, testing how the foundation would react when variables like the strength of the soil, the size of the columns, and the intensity of the earthquake changed. By doing so, they moved beyond asking simply "will it hold?" to asking "what are the odds it will hold?" This shift provides a much clearer picture of risk, helping designers understand which factors truly matter when protecting a building from seismic disaster.

The researchers began by refining the mathematical rules that describe how a geogrid-encased stone column foundation behaves under seismic stress. They verified these rules against data from physical experiments, including large-scale shaking table tests where miniature foundations were subjected to simulated earthquakes, as well as real-world field tests. Once they confirmed their calculations matched reality, they built a digital model to explore the vast range of possibilities. They fed the model with realistic ranges for key factors: the strength of the geogrid fabric, the diameter of the stone columns, the friction and stickiness of the surrounding soil, and the specific forces of an earthquake, including both horizontal and vertical shaking. The model then calculated the probability of failure for millions of different combinations of these factors.

The results revealed a landscape of risk where some factors are far more critical than others. The researchers found that the failure probability is most sensitive to the strength of the surrounding natural soil and the ratio of vertical to horizontal shaking. If the native soil is weak, the foundation is far more likely to fail, regardless of how strong the stone columns are. Similarly, when an earthquake shakes the ground up and down with significant force relative to the side-to-side shaking, the risk of collapse rises sharply. In contrast, the diameter of the stone columns and the inherent stickiness of the stone material itself had a surprisingly small effect on the overall safety. The study showed that simply making the columns thicker does not necessarily make the foundation safer; in some cases, larger columns can actually increase the risk because they attract more of the earthquake's shaking force.

Perhaps most importantly, the study identified the specific design choices that offer the best protection. The researchers concluded that the most effective strategy is not to rely on massive stone pillars, but to use a dense arrangement of smaller columns wrapped in high-strength geogrid fabric. A wider foundation base also significantly reduces the risk of failure by spreading the load more evenly. The team's analysis suggests that while improving the stone columns themselves is helpful, the true key to safety lies in the quality of the natural soil and the strength of the fabric belt holding the stones together. By focusing on these specific elements—using strong geogrids, ensuring a wide foundation, and packing the columns closely together—engineers can design foundations that are far more resilient to the unpredictable violence of an earthquake. This new method provides a practical, probabilistic tool for designers to quantify these risks, moving the field from guesswork toward a precise understanding of safety in a shaking world.

Technical Summary: A Simplified Method for Seismic Fragility Analysis of Geogrid-Encased Stone Column Composite Foundations

Problem Statement
Geogrid-encased stone columns (GESCs) are increasingly utilized to reinforce soft soil deposits in seismically active regions, offering improved stiffness and bearing capacity compared to ordinary stone columns. However, while static performance is well-documented, the seismic response of GESC composite foundations remains complex and insufficiently quantified. Existing methodologies often rely on deterministic safety factors or computationally expensive finite-element dynamic analyses, which fail to adequately account for the inherent uncertainties in geotechnical systems—such as spatial variability in soil properties, material strength randomness, and seismic demand. Consequently, there is a lack of a practical, probabilistic framework to quantify the seismic fragility (failure probability) of GESC foundations under varying earthquake intensities.

Methodology
This study proposes a machine learning-based probabilistic assessment framework to evaluate the seismic fragility of GESC composite foundations. The methodology integrates three core components:

Analytical Modeling: An ultimate seismic bearing capacity ( $q_u$ ) model is established using the limit equilibrium method and the equivalent area method. The model incorporates the apparent cohesion ( $c_g$ ) provided by geogrid encasement and accounts for horizontal ( $k_h$ ) and vertical ( $k_v$ ) seismic coefficients. The derived seismic bearing capacity factors ( $N_{cE}, N_{qE}, N_{\gamma E}$ ) are validated against existing analytical solutions, shaking table tests, and field test data, demonstrating high accuracy.
Probabilistic Simulation: To address parameter uncertainty, the study employs Monte Carlo simulation. Key input variables, including seismic coefficients, column diameter ( $D_c$ ), geogrid tensile strength ( $T$ ), area replacement ratio ( $m$ ), and shear strength parameters of both the column ( $\phi_c, c_c$ ) and surrounding soil ( $\phi_s, c_s$ ), are treated as random variables with defined statistical distributions (primarily normal and uniform).
Machine Learning Integration: A neural network model (architecture: 9-32-16-8-1) is trained in MATLAB to predict the failure probability ( $P_f$ ). The model uses the analytical $q_u$ calculations compared against a representative vertical load demand ( $P_v$ ) to classify failure states. The model's performance is verified using loss curves, accuracy curves, and Receiver Operating Characteristic (ROC) analysis.

Key Results
A comprehensive parametric study was conducted to evaluate the influence of twelve parameters on the failure probability ( $P_f$ ):

Influential Parameters: The failure probability is most significantly governed by the shear strength of the surrounding soil ( $\phi_s$ ) and the ratio of vertical to horizontal seismic coefficients ( $k_v/k_h$ ).
Correlations: $P_f$ shows a positive correlation with geogrid tensile strength ( $T$ ), area replacement ratio ( $m$ ), shear strength parameters ( $\phi_c, c_c, \phi_s, c_s$ ), footing width ( $B$ ), and overburden pressure ( $q_1$ ). Conversely, $P_f$ exhibits a negative correlation with the ratio $k_v/k_h$ , column diameter ( $D_c$ ), and vertical load demand ( $P_v$ ).
Negligible Factors: The study found that column diameter ( $D_c$ ) and column cohesion ( $c_c$ ) have a negligible effect on the failure probability compared to other factors.
Soil vs. Column: The seismic reliability of the foundation is more critically influenced by the inherent properties of the native soil than by the enhancements provided by the columns themselves.

Significance and Claims
The paper claims to provide a simplified yet rigorous analytical methodology that efficiently generates seismic fragility curves for GESC foundations while explicitly incorporating parameter uncertainty. By moving beyond deterministic safety factors, the proposed framework offers a more realistic assessment of risk under seismic events.

The authors conclude that design efforts should prioritize the native soil conditions over column-specific enhancements. Specifically, the study recommends focusing on:

Utilizing high-strength geogrids.
Ensuring adequate foundation width.
Maintaining a high area replacement ratio.
Employing a dense arrangement of small-diameter columns.

This approach aims to comprehensively enhance the seismic reliability of GESC composite foundations, addressing the gap between complex numerical simulations and practical design requirements.

More like this