Soil liquefaction—the process where saturated soil loses its structure and transforms to a fluid-like state—can have devastating outcomes, as evidenced by the Great East Japan Earthquake in 2011. Large-scale liquefaction during this disaster damaged thousands of houses in the Tokyo Bay area, posing a formidable challenge to infrastructure in Japan.
To prevent this, chemical grouting is considered to be effective where grout is injected into the soil to substitute soil pore water with a solidifying chemical, thus improving the soil structure. But, in heterogeneous soil with low-permeability areas, achieving uniform and reliable chemical grout permeation is difficult.
A previous study presented an innovative technique named Finite Element Method (FEM) to analyze the permeation behavior of chemical grout in both homogeneous and heterogeneous soils, showing that grout tends to bypass areas of low permeability, affecting soil remediation negatively. Therefore, prediction and optimization of permeation performance are important for successful chemical grouting in complex soil types.
To this end, Professor Shinya Inazumi from the College of Engineering, Shibaura Institute of Technology (SIT), Japan, led a team of researchers from SIT and the Asian Institute of Technology, Thailand, to integrate AI-based predictive modeling into FEM-based permeation analysis. They developed a practical framework to assess grout permeation behavior in soils with low-permeability zones, and their paper was published in the journal Results in Engineering on June 1, 2025.
“Unlike previous studies that relied solely on traditional permeation analysis methods to highlight the effects of soil heterogeneity on chemical grout permeation efficiency, this study is unique by combining traditional permeation analysis with advanced AI techniques such as neural networks and gradient boosting decision trees,” explains Prof. Inazumi about the novelty of the study.
After creating a soil model with low-permeability regions, a 2D FEM-based permeation analysis was performed to calculate the permeation velocity, based on which permeation risk and range were evaluated. Parameters that contributed to a permeation risk of greater than 10% were identified and used as inputs for multiple regression analysis to predict the permeation risk.
The FEM-derived permeation datasets were also used to train two AI-based predictive models: a neural network and a gradient boosting decision tree.
Results showed that FEM-based permeation analysis demonstrated an average permeation rate of 94.5% and a worst-case value of 81% when 5.5% of the soil was of low permeability. The AI-based predictive models showed an average permeation rate of 96% and a worst-case drop of 83%.
A comparative validation of AI-based predictive models with FEM simulations and previous studies suggests a high predictive accuracy of R2= 0.849. This demonstrates that AI-based models can be integrated with numerical modeling when trained on well-structured datasets.
Moreover, AI-based models delivered predictions in less than two seconds, compared to FEM simulations which took almost 30–40 minutes to process predictions. Simple regression models developed in this study using inputs like soil geometry and proximity to low-permeability zones accurately estimated permeation risks without the need for exhaustive computational steps.
Additionally, the FEM-based analysis revealed that the low-permeability zones block flow patterns and reduce permeation velocities in the soil.
To achieve further reliable prediction models, experimental validation with field data is necessary. With more diverse training data, the AI-based predictive models can be more accurate, reducing the risk of overestimation.
“These findings contribute to the field by proposing a practical framework that helps engineers to predict grout permeation behavior more efficiently, even in complex and heterogeneous soils,” concludes Prof. Inazumi.
Incorporating more physical properties of the grouting process, like grout pressure, grout rheology, injection conditions, and grain size distributions into FEM simulations can additionally improve the practical field applicability of this novel integrative framework in the future.
Overall, this study may help address critical challenges associated with the liquefaction of soil in earthquake-prone regions of the world, like Japan.
More information:
Khin Nyein Chan Kyaw et al, Integration of FEM-based permeation analysis and AI-based predictive models for improved chemical grout permeation assessment in heterogeneous soils, Results in Engineering (2025). DOI: 10.1016/j.rineng.2025.105071
Shibaura Institute of Technology
Citation:
AI-augmented models improve chemical grouting predictions in complex soils (2025, May 28)
retrieved 28 May 2025
from https://techxplore.com/news/2025-05-ai-augmented-chemical-grouting-complex.html
This document is subject to copyright. Apart from any fair dealing for the purpose of private study or research, no
part may be reproduced without the written permission. The content is provided for information purposes only.
