Engineering Design / ConsultingShanghai Geotechnical Investigations & Design Institute2026-08-03

MOI Turns Engineering Archives into Reusable Knowledge for SGIDI

MOI digitizes survey reports, scans, cross-page tables, and drawings while preserving layout semantics and source evidence, making historical archives searchable and reviewable.

Shanghai Geotechnical Investigations & Design Institute

SGIDI (Shanghai Geotechnical Investigations & Design Institute Group) is one of China's leading geotechnical engineering and urban infrastructure consultancies, providing geological surveys, foundation design, underground engineering and construction supervision. Over decades of project delivery it has accumulated a substantial engineering archive.

Multimodal
Documents + drawings
Page/Region
Traceable evidence
Human-in-loop
Confidence review
API
Downstream delivery

The challenge

Engineering documents-geological survey reports, soil stratigraphy tables, foundation drawings, structural calculations-contain critical project data, but this information is locked inside unstructured formats: scanned PDFs, CAD files, handwritten field notes, and complex multi-column tables with domain-specific notation. Manually extracting key parameters (soil layer properties, bearing capacities, design loads), classifying document types, and tagging them for retrieval is extremely labor-intensive and error-prone. Engineers spend significant time searching archives rather than doing engineering work, and institutional knowledge embedded in historical projects remains largely inaccessible.

The solution

SGIDI deployed an AI Drawing Structuring & Classification System on OmniFabric, automating key information extraction from engineering documents, intelligent classification by document type, and structured tagging for fast retrieval and downstream analysis.

The outcome

Complex engineering tables-soil stratigraphy, physical-mechanical parameters, bearing capacity data-are automatically extracted and structured, making decades of project data instantly queryable.

Documents are intelligently classified by type (survey report, design drawing, supervision record) and tagged with domain metadata (project ID, region, soil conditions), eliminating manual filing effort.

Cross-project search by soil type, design parameter, or geological condition lets engineers instantly locate relevant historical precedents-turning the archive from a storage burden into a knowledge asset.

Solution Architecture

Data sources
  • Geological Survey Reports (PDF / Scanned)
  • Soil Parameter & Stratigraphy Tables
  • Foundation & Structural Design Drawings (CAD)
  • Construction Supervision & Field Records
MatrixOne Intelligence
  • OCR & Multi-Format Document Parsing (PDF, CAD, Scans)
  • Complex Table Recognition & Parameter Extraction
  • Document Type Classification (LLM + Rule Engine)
  • Domain-Specific Tagging (Project, Soil Type, Region)
  • Structured Index & Knowledge Base Construction
Business applications
  • Structured Parameter Database (Soil, Foundation, Load)
  • Auto-Classified & Tagged Document Library
  • Cross-Project Search & Historical Reference
  • Data Foundation for AI-Assisted Design & Review

Related solutions and products