Skip to content

🛡️ BIMGUARD AI

Unified Compliance Engine for BIM Validation

Group 5 — Letícia · Malak · Marc · Mark · Osama

FMP Plenary 1 | MAICEN Module 10

THE GOAL: Automating Complex Compliance Validation

BIMGUARD streamlines the validation of complex 3D geometry and project requirements to ensure absolute adherence to Building Codes and ISO 19650 standards.

  • $15.8B: Annual U.S. losses from interoperability failures (NIST GCR 04-867)
  • 📐 Code Conformance: Automating architectural cross-checks against Building Codes (e.g., fire protection walls, accessibility and door swings).
  • 📋 BEP & EIR Alignment: Ensuring strict adherence to ISO 19650 naming, classification, and parameter rules directly within the validation pipeline.

OUR SOLUTION: A "Digital Inspector" for openBIM

BIMGUARD AI bridges unstructured documents (BEPs, EIRs, codes) and structured IFC models through a dual-core engine.

🧠 NLP Core — Rule Extraction

LLMs (OpenAI/Anthropic) parse BEP, EIR & code PDFs → convert clauses into structured, human-reviewed JSON/SHACL rules.

Focus areas: Naming Rules · Classification · Parameters

📐 Geometry Core — Halo Engine

Generates volumetric "Halo" clearance zones around end-of-line and in-line assets. Accounts for Final LOD gaps (e.g., seismic bracing), optimum centre-to-centre service distances, and prevents galvanic corrosion by verifying pipe material distances.

Focus areas: Maintenance Access · Service Spacing · Material Checking

USE CASE: Weekly Coordination & Compliance

On "model drop day," BIMGUARD AI automates critical checkpoints:

1. Gatekeeper Check Validates uploaded IFC against BEP and EIR requirements — flags missing parameters, naming violations (ISO 19650), and incorrect classification codes.

2. Geometry & Clearance Checks Generates "Halo" zones around assets for maintenance access. Validates architectural code compliance (fire ratings, accessibility) and cross-checks service spacing/materials.

Output: BCF Issues communicating breached codes/standards, assigned to correct parties with trackable progress.

PROGRESS: Where We Are Now

✅ Completed

  • Problem framing & methodology submitted
  • Next.js 16 + Tailwind v4 frontend scaffold
  • 3D IFC viewer (@thatopen/components)
  • FastAPI backend with Docker setup
  • MkDocs project documentation site

🚧 In Progress

  • Rule Studio — PDF ingestion → rule extraction
  • Properties panel & element inspection
  • Halo engine prototype with IfcOpenShell
  • BCF export integration and issue tracking
  • Compliance analysis workflow UI

Tech Stack: Next.js · FastAPI · IfcOpenShell · Three.js · OpenAI API · IFC4 RV

OPEN DISCUSSION: Questions for the Cohort

We'd appreciate your thoughts on these areas:

  • 🎯 Scope Boundary: How deep should the LLM-driven rule extraction go? Full automation vs. a structured suggestion interface where experts curate rules?
  • ⚖️ Validation Approach: What metrics best demonstrate compliance accuracy — precision/recall on expert-annotated BEPs/EIRs, or real-world pilot results from live projects?
  • 🌍 Multi-Jurisdiction Rules: Building codes vary by country and region. Any experience with modular rule-package architectures for handling jurisdiction-specific requirements?

🛡️ Thank You

We welcome your feedback and questions.

📂 github.com/osama-ata/bim-guard | Group 5 — MAICEN M10