Automation in Geotechnical Engineering: AI-Enabled Foundation Design Workflow-Paper

Paper

Pdf Summary

The paper presents a proof-of-concept framework for automating routine geotechnical foundation design using interconnected AI agents. Built with the n8n workflow platform, the system combines large language models (LLMs) with deterministic calculations to interpret site data, select suitable analyses, and generate design outputs.<br /><br />The workflow begins with a Site Characterisation Agent, which processes borehole and laboratory data, assigns representative soil parameters using configured correlations, and produces a structured ground model. A Task Routing Agent interprets a designer’s natural-language brief, identifies the appropriate analytical method, formats required inputs, and requests any missing information. Specialist agents then perform assessments; a worked example uses a Pile Capacity Agent to calculate pile resistance and determine whether a proposed pile length meets the design load. The paper also describes using a vector database to retrieve national code requirements and examples to guide consistent report writing.<br /><br />In the example, the AI-derived pile capacity result was checked against manual calculations. The authors argue that the main value of AI lies less in carrying out standard engineering calculations than in selecting the right tools and assembling appropriate input data. The modular framework could be expanded to other foundation types, additional investigation data, and commercial geotechnical software.<br /><br />The study is simplified and does not represent the complexity of many real projects. Because LLMs can produce incorrect or fabricated outputs, engineers must verify results and apply professional judgment. The proposed approach is intended to reduce repetitive work and improve consistency, while shifting engineers’ effort toward oversight and higher-level decisions—not replacing engineering responsibility.

Keywords

geotechnical foundation design

AI agents

n8n workflow

large language models

pile capacity analysis

site characterisation

deterministic calculations

ground model

vector database

engineering automation