BIM Cost Estimation with DDC CWICR

Generate accurate cost estimates from BIM models using AI classification and the DDC CWICR construction cost database.

Business Case

Problem: Traditional cost estimation:

• Manual and time-consuming (weeks for detailed estimate)
• Subjective and inconsistent between estimators
• Requires specialized knowledge
• Difficult to update with design changes

Solution: Automated BIM-to-cost pipeline:

• Extract quantities directly from model
• AI classifies elements to work items
• Vector search finds matching prices in CWICR
• Complete estimate in hours, not weeks

ROI: 80% reduction in estimation time, consistent methodology

System Architecture

┌──────────────────────────────────────────────────────────────────────────┐
│                  BIM TO COST ESTIMATION PIPELINE                          │
├──────────────────────────────────────────────────────────────────────────┤
│                                                                           │
│   ┌─────────┐     ┌─────────┐     ┌─────────┐     ┌─────────────────┐   │
│   │ BIM     │     │ DDC     │     │ AI      │     │ DDC CWICR       │   │
│   │ Model   │────►│Converter│────►│ LLM     │────►│ Vector Search   │   │
│   │.rvt/.ifc│     │         │     │         │     │ (Qdrant)        │   │
│   └─────────┘     └─────────┘     └─────────┘     └─────────────────┘   │
│                        │              │                    │             │
│                        ▼              ▼                    ▼             │
│                   ┌─────────┐    ┌─────────┐         ┌──────────┐       │
│                   │ .xlsx   │    │ Work    │         │ Matched  │       │
│                   │ QTO     │    │ Items   │         │ Rates    │       │
│                   └─────────┘    └─────────┘         └──────────┘       │
│                        │              │                    │             │
│                        └──────────────┼────────────────────┘             │
│                                       ▼                                  │
│                              ┌─────────────────┐                        │
│                              │ COST ESTIMATE   │                        │
│                              │                 │                        │
│                              │ • By element    │                        │
│                              │ • By trade      │                        │
│                              │ • By phase      │                        │
│                              │ • Resources     │                        │
│                              └─────────────────┘                        │
│                                                                           │
└──────────────────────────────────────────────────────────────────────────┘

DDC CWICR Database

Database Overview:
  work_items: 55,719
  resources: 27,672
  languages: 9 (AR, DE, EN, ES, FR, HI, PT, RU, ZH)
  fields_per_item: 85
  embedding_model: text-embedding-3-large (3072d)
  vector_db: Qdrant

Collections:
  - ddc_cwicr_ar  # Arabic (Dubai prices)
  - ddc_cwicr_de  # German (Berlin prices)
  - ddc_cwicr_en  # English (Toronto prices)
  - ddc_cwicr_es  # Spanish (Barcelona prices)
  - ddc_cwicr_fr  # French (Paris prices)
  - ddc_cwicr_hi  # Hindi (Mumbai prices)
  - ddc_cwicr_pt  # Portuguese (São Paulo prices)
  - ddc_cwicr_ru  # Russian (St. Petersburg prices)
  - ddc_cwicr_zh  # Chinese (Shanghai prices)

Pipeline Stages

Python Implementation

import pandas as pd
import numpy as np
from qdrant_client import QdrantClient
from qdrant_client.models import Filter, FieldCondition, MatchValue
from openai import OpenAI
from typing import List, Dict, Optional
from dataclasses import dataclass
import json

@dataclass
class WorkItem:
    """Matched work item from CWICR"""
    cwicr_code: str
    description: str
    unit: str
    unit_price: float
    labor_cost: float
    material_cost: float
    equipment_cost: float
    productivity: float  # units per hour
    currency: str
    confidence: float

@dataclass
class CostLineItem:
    """Single line item in estimate"""
    bim_type: str
    work_item: WorkItem
    quantity: float
    quantity_unit: str
    total_cost: float
    labor_cost: float
    material_cost: float
    equipment_cost: float
    phase: str
    trade: str


class BIMCostEstimator:
    """BIM to cost estimation using DDC CWICR"""

    def __init__(
        self,
        qdrant_url: str,
        qdrant_api_key: str = None,
        openai_api_key: str = None,
        language: str = "EN"
    ):
        self.qdrant = QdrantClient(url=qdrant_url, api_key=qdrant_api_key)
        self.openai = OpenAI(api_key=openai_api_key)
        self.language = language
        self.collection = f"ddc_cwicr_{language.lower()}"

    def get_embedding(self, text: str) -> List[float]:
        """Generate embedding for text"""
        response = self.openai.embeddings.create(
            model="text-embedding-3-large",
            input=text,
            dimensions=3072
        )
        return response.data[0].embedding

    def search_cwicr(
        self,
        query: str,
        limit: int = 5,
        category_filter: str = None
    ) -> List[WorkItem]:
        """Search CWICR database for matching work items"""

        # Get embedding
        query_vector = self.get_embedding(query)

        # Build filter if category specified
        query_filter = None
        if category_filter:
            query_filter = Filter(
                must=[
                    FieldCondition(
                        key="category",
                        match=MatchValue(value=category_filter)
                    )
                ]
            )

        # Search
        results = self.qdrant.search(
            collection_name=self.collection,
            query_vector=query_vector,
            query_filter=query_filter,
            limit=limit
        )

        # Parse results
        work_items = []
        for r in results:
            payload = r.payload
            work_items.append(WorkItem(
                cwicr_code=payload.get('code', ''),
                description=payload.get('description', ''),
                unit=payload.get('unit', ''),
                unit_price=float(payload.get('unit_price', 0)),
                labor_cost=float(payload.get('labor_cost', 0)),
                material_cost=float(payload.get('material_cost', 0)),
                equipment_cost=float(payload.get('equipment_cost', 0)),
                productivity=float(payload.get('productivity', 1)),
                currency=payload.get('currency', 'USD'),
                confidence=r.score
            ))

        return work_items

    def decompose_bim_type(
        self,
        bim_type: str,
        category: str
    ) -> List[str]:
        """Use LLM to decompose BIM type into work items"""

        prompt = f"""
Decompose this BIM element type into construction work items:

BIM Type: {bim_type}
Category: {category}

List the individual work activities needed to construct this element.
For example, "Brick Wall 240mm" decomposes into:
- Masonry: Brick laying
- Mortar: Cement mortar for joints
- Plaster: Internal plaster finish
- Paint: Wall painting

Return a JSON array of