arXiv subjects

Yared W. Bekele

Publications and source records attributed to Yared W. Bekele.

6 recordsLinked to original sources

arXiv2026

GeoMCP: A Trustworthy Framework for AI-Assisted Analytical Geotechnical Engineering

Analytical methods underpin geotechnical engineering practice, yet their implementation remains fragmented across error-prone spreadsheets and opaque proprietary software. While Large Language Models (LLMs) offer transformative potential for streamlining engineering workflows, their statistical nature fundamentally conflicts with the strict determinism required for safety-critical calculations. Their tendency to hallucinate formulas, misinterpret units, or alter methodologies between sessions creates a critical trust gap. This paper introduces GeoMCP, a framework built to bridge this gap via a key insight: engineering methods should be represented as structured data, not embedded code. GeoMCP captures analytical methods as "method cards", declarative JSON files defining formulas, units, applicability limits, and literature citations. A constrained symbolic engine executes these cards with verified dimensional consistency, while structured "Agent Skills" guide LLMs to apply engineering judgment and orchestrate the analysis. By exposing these verified capabilities through the Model Context Protocol (MCP), GeoMCP shifts the role of the AI from an unreliable calculator to an intelligent orchestrator. Validated against an official JRC Eurocode~7 worked example, the framework demonstrates computational parity with traditional approaches while ensuring complete mathematical transparency. Ultimately, GeoMCP provides a blueprint for transitioning the industry from isolated legacy software to an interoperable, AI-ready ecosystem where engineers can leverage modern AI without surrendering professional responsibility.

cs.CE↗

arXiv2025

GeoSim.AI: AI assistants for numerical simulations in geomechanics

The ability to accomplish tasks via natural language instructions is one of the most efficient forms of interaction between humans and technology. This efficiency has been translated into practical applications with generative AI tools now allowing users to get things done through natural language queries. The emergence of advanced Large Language Models (LLMs) marks a pivotal shift in this direction. With ongoing advancements in the field of generative AI, integrating natural language commands into sophisticated technical fields in science and engineering is becoming increasingly feasible. This paper introduces GeoSim.AI - a suite of AI assistants for numerical simulations in geomechanics - thereby demonstrating the transformative potential of generative AI in geotechnical engineering. We investigate how AI assistants powered by LLMs can streamline the process of creating complex simulation inputs and interpreting results by translating natural language instructions or image inputs into precise technical commands and scripts. This approach aims to bridge the gap between human intent and the intricate requirements of numerical modeling tools, potentially revolutionizing how researchers and engineers interact with simulation software. We present demonstrations involving AI assistants for performing slope stability analyses in various software packages. The demonstrations highlight the potential of this technology to significantly enhance productivity and accessibility in computational geomechanics. GeoSim.AI is under active development, continuously expanding the suite of AI assistants for various numerical simulation problems in geotechnical engineering.

cs.CE↗

arXiv2024

Physics-informed neural networks with curriculum training for poroelastic flow and deformation processes

Physics-Informed Neural Networks (PINNs) have emerged as a highly active research topic across multiple disciplines in science and engineering, including computational geomechanics. PINNs offer a promising approach in different applications where faster, near real-time or real-time numerical prediction is required. Examples of such areas in geomechanics include geotechnical design optimization, digital twins of geo-structures and stability prediction of monitored slopes. But there remain challenges in training of PINNs, especially for problems with high spatial and temporal complexity. In this paper, we study how the training of PINNs can be improved by using an idealized poroelasticity problem as a demonstration example. A curriculum training strategy is employed where the PINN model is trained gradually by dividing the training data into intervals along the temporal dimension. We find that the PINN model with curriculum training takes nearly half the time required for training compared to conventional training over the whole solution domain. For the particular example here, the quality of the predicted solution was found to be good in both training approaches, but it is anticipated that the curriculum training approach has the potential to offer a better prediction capability for more complex problems, a subject for further research.

cs.CE↗

arXiv2020

Physics-informed deep learning for flow and deformation in poroelastic media

A physics-informed neural network is presented for poroelastic problems with coupled flow and deformation processes. The governing equilibrium and mass balance equations are discussed and specific derivations for two-dimensional cases are presented. A fully-connected deep neural network is used for training. Barry and Mercer's source problem with time-dependent fluid injection/extraction in an idealized poroelastic medium, which has an exact analytical solution, is used as a numerical example. A random sample from the analytical solution is used as training data and the performance of the model is tested by predicting the solution on the entire domain after training. The deep learning model predicts the horizontal and vertical deformations well while the error in the predicted pore pressure predictions is slightly higher because of the sparsity of the pore pressure values.

cs.CE↗

arXiv2020

Deep Learning for One-dimensional Consolidation

Neural networks with physical governing equations as constraints have recently created a new trend in machine learning research. In line with such efforts, a deep learning model for one-dimensional consolidation where the governing equation is applied as a constraint in the neural network is presented here. A review of related research is first presented and discussed. The deep learning model relies on automatic differentiation for applying the governing equation as a constraint. The total loss is measured as a combination of the training loss (based on analytical and model predicted solutions) and the constraint loss (a requirement to satisfy the governing equation). Two classes of problems are considered: forward and inverse problems. The forward problems demonstrate the performance of a physically constrained neural network model in predicting solutions for one-dimensional consolidation problems. Inverse problems show prediction of the coefficient of consolidation. Terzaghi's problem with varying boundary conditions are used as example and the deep learning model shows a remarkable performance in both the forward and inverse problems. While the application demonstrated here is a simple one-dimensional consolidation problem, such a deep learning model integrated with a physical law has huge implications for use in, such as, faster real-time numerical prediction for digital twins, numerical model reproducibility and constitutive model parameter optimization.

cs.CE↗

arXiv2017

On Mixed Isogeometric Analysis of Poroelasticity

Pressure oscillations at small time steps have been known to be an issue in poroelasticity simulations. A review of proposed approaches to overcome this problem is presented. Critical time steps are specified to alleviate this in finite element analyses. We present a mixed isogeometric formulation here with a view to assessing the results at very small time steps. Numerical studies are performed on Terzaghi's problem and consolidation of a layered porous medium with a very low permeability layer for varying polynomial degrees, continuities across knot spans and spatial discretizations. Comparisons are made with equal order simulations.

math.NA↗