Graduate-Project

Operational Campus Digital Twin from BIM

UC Berkeley · Civil & Environmental Engineering · M.S. Construction Systems

A web-based operational digital twin for UC Berkeley’s campus buildings, centered on the Grimes Engineering Center. The system integrates BIM model analysis, real-time energy metering (BMO), and interactive web visualization — demonstrating a complete “BIM-to-Operations” pipeline at zero infrastructure cost.

Live Demo: doe2park.github.io/Graduate-Project

System Overview

BMO Obvius Meters (154 devices, 26 online)
  → GitHub Actions (Python, every 15 min)
    → JSON Data Store (live + 24hr history + daily archive + weekly report)
      → GitHub Pages (free HTTPS hosting)
        → 5 Web Interfaces
           ├── Campus Map (ArcGIS 2D)
           ├── 3D Viewer (CesiumJS + Google Photorealistic 3D Tiles)
           ├── Energy Dashboard (Chart.js)
           ├── Weekly Report
           └── Grimes MEP Viewer (Three.js + WebXR)
  → Cloudflare Workers AI (Llama 3.1 chatbot)

Live Interfaces

Interface	URL	Technology
Campus Map	grimes-campus-map-arcgis.html	ArcGIS JS SDK 4.29
3D Campus View	campus-3d.html	CesiumJS + Google 3D Tiles
Energy Dashboard	campus-energy-dashboard.html	Chart.js 4.4
Weekly Report	weekly-report.html	Vanilla JS
Grimes MEP Viewer	grimes-xr.html	Three.js + WebXR
AI Chatbot	campus-chatbot.ucb-dt.workers.dev	Cloudflare Workers AI

Campus Map

Interactive 2D satellite map displaying live energy data for 64 UC Berkeley buildings.

64 buildings with coordinates verified from berkeley.edu/map
26 buildings with live kW readings from BMO metering, auto-refreshed every 60 seconds
Category-colored markers with energy-intensity outline (green → gold → red)
Click any building for popup with department, year built, floors, energy data
24-hour time slider with play button — replay campus energy patterns
AI chatbot powered by Cloudflare Workers AI (Llama 3.1 8B) with local NLP fallback
- 20+ intents, fuzzy string matching, handles typos and varied phrasing
- Korean keyword support
- Time-slider aware — answers reflect currently selected time
Links to berkeley.edu building pages, energy dashboard, and 3D viewer
Grimes beacon marker guides users to the full BIM-to-WebXR digital twin

3D Campus Viewer

Photorealistic 3D visualization of UC Berkeley campus with live energy overlay.

Google Photorealistic 3D Tiles via CesiumJS — real building textures, trees, terrain
OSM Buildings fallback when Google Tiles unavailable
Energy-coded ground circles: size proportional to kW, color by intensity
Floating labels on building rooftops (visible on zoom-in, fade on zoom-out)
24-hour time slider with stats panel synchronization
- Total kW, buildings online, daily cost, CO₂ all update with slider position
Click any building for info popup (name, department, kW, year, floors, cost, CO₂)
Enhanced rendering: maximumScreenSpaceError: 4, FXAA, device pixel ratio scaling

Energy Dashboard

Real-time analytics with 6 interactive charts.

Stats cards: Campus load (kW), daily cost ($), avg kW/floor, daily energy (MWh)
Energy by Category: Donut chart — Engineering, Science, Humanities, Professional, Student Life, Libraries
Top 10 Buildings: Horizontal bar chart ranked by current kW
Campus Load History: 24-hour line chart with hover crosshair and tooltip
Cost per Building: Top consumers by estimated $/day
Energy Efficiency (kW/floor): Normalized by floor count — identifies truly inefficient buildings
Building Age vs Energy: Scatter plot — year built vs current kW with building metadata on hover
Auto-refresh every 60 seconds
shortName() mapping with 30+ entries and fuzzy matching

Weekly Report

Auto-generated 7-day energy summary.

Weekly stats: average kW, total MWh, cost, CO₂
Daily breakdown cards with peak times
Peak vs average comparison charts
Building weekly rankings table
Generated from daily JSON snapshot archive

Grimes MEP Viewer

3D MEP (Mechanical, Electrical, Plumbing) walkthrough of Grimes Engineering Center with live energy data.

Three.js renderer with WebXR support (desktop, mobile, VR headset)
8 MEP system layers: HVAC Ductwork, HVAC Equipment, Plumbing/Piping, Fire Protection, Electrical, Diffusers/Grilles, Pumps/Valves, Spiral Ductwork
Toggle individual systems on/off via legend
View modes: System (color by type), Temperature, Flow, Energy (heatmap)
Live BMO data for electrical components (3 meters: Roof Electric, 480/277V, 208/120V)
WASD + QE movement, mouse orbit, click any component for sensor panel
First-person walkthrough mode

Live Data (BMO Metering — ~70% of UI)

Meter	Voltage	Data Points
Meter #3 (Roof Electric)	480V	kW, current, voltage, power factor, frequency, kWh delivered, peak demand
Meter #76 (HVAC/Lighting)	480/277V	kW, current, voltage
Meter #77 (Outlets/IT)	208/120V	kW, current, voltage
Water meter	—	flow rate (cf/m), total consumption (cf)
Steam meter	—	total consumption (gal)

BIM-Validated Floor Energy Breakdown (~15% of UI)

Derived from analysis of 741,796 BIM elements in the Revit/Navisworks model. 93 electrical panelboards were identified and mapped to BMO meters by floor:

Floor	Meter Sources	Panelboards	Description
SRV (Roof)	Meter #3 × 100%	0 (direct AHU feed)	Rooftop air handling units, exhaust fans
L3	Meter #77 × 18%	11 × 208V	Upper floor outlets, IT equipment
L2	Meter #77 × 35%	22 × 208V	Mid floor outlets, lab equipment
L1	Meter #76 × 45% + Meter #77 × 31%	14 × 480V + 19 × 208V	HVAC, lighting, outlets
B1	Meter #76 × 55% + Meter #77 × 16%	17 × 480V + 10 × 208V	MSB, mechanical room, workshop

Key finding: 480/277V panelboards (Meter #76) exist only on B1 and L1. L2 and L3 have exclusively 208/120V distribution. Main switchboards (MSB-1, MSB-2) are located on B1.

Simulated Data (~5% of UI)

HVAC, plumbing, and fire protection sensor readings shown on component click are simulated placeholders. These are ready for future BAS (Building Automation System) integration. All simulated values are labeled “SIM” in the UI.

Data Architecture

Automated Pipeline

File	Description	Update Frequency
`data/campus_energy.json`	Live snapshot — 26 buildings × kW, cost, CO₂, anomaly flags, prediction	Every 15 min
`data/campus_energy_history.json`	24-hour rolling history — up to 96 data points per building	Every 15 min
`data/daily/YYYY-MM-DD.json`	Permanent daily archive — all readings preserved for trend analysis	Append each cycle
`data/weekly_report.json`	7-day summary — avg/peak/min kW, cost, CO₂ per building	Every 15 min

BMO Connection

System: BMO Obvius (BuildingManager Online)
Auth: HTTP Basic Authentication
Database: dbU216ucberkelF682
Devices discovered: 154 AcquiSuite devices
Currently reporting: 26 buildings
Offline: 12 buildings (hardware disconnected — requires Facilities Services)
kW columns parsed: “kW total (kW)”, “Active Power Total (kW)”, “kW total”, “kW”, “kW del-rec (kW)”, “Power Total (kW)”, “kW del (kW)”

Features

Anomaly detection: Flags buildings consuming >150% of their 2-hour rolling average
Prediction: Weighted moving average for short-term kW prediction
CO₂ estimation: kW × 24h × 0.21 kg/kWh (California grid average, CARB)
Cost estimation: kW × 24h × $0.15/kWh (PG&E commercial rate)
Daily snapshots: Permanent JSON files for long-term historical analysis

AI Chatbot

Deployed as a Cloudflare Worker using Workers AI (free tier, no API key required).

Model: Meta Llama 3.1 8B Instruct (@cf/meta/llama-3.1-8b-instruct)
Endpoint: https://campus-chatbot.ucb-dt.workers.dev
Context: Sends current energy data + building metadata with each query
Time-aware: Reflects current time slider position
Fallback: Local fuzzy NLP engine with 20+ intents when AI proxy is unavailable
Languages: English and Korean

BIM Model Analysis

The Grimes Engineering Center BIM model (Revit → Navisworks NWD) contains 741,796 elements across 48 NWC files from multiple design and construction teams:

File Code	Firm / Role	Discipline
`RT`	Rutherford & Chekene	Electrical design
`FMB`	Mechanical designer	HVAC / plumbing design
`CDC`	Mechanical contractor	HVAC / plumbing construction BIM
`PRIBUSS PRAGMATIC`	Pribuss Engineering + Pragmatic PE	Fire protection design + construction
`SOM`	Skidmore, Owings & Merrill	Architecture
`XL`	—	3D Grids / coordination

MEP Files (35 NWC)

Electrical:     00/01/02/03-UCBBE-RT-ELEC.nwc (power distribution)
                00/01/02/03-UCBBE-RT-ELEC-IW.nwc (internal wiring)
                00/01/02/03-UCBBE-RT-ELEC-LTG.nwc (lighting)
                RF-UCBBE-RT-ELEC.nwc (roof electrical)
Fire Protection: 00/01/02/03-UCBBE-PRIBUSS PRAGMATIC-FP.nwc
Mechanical:     00/01/02/03/04-UCBBE-FMB-MECH DUCT.nwc
                00/01/02/03-UCBBE-FMB-MP.nwc (mechanical piping)
                00/01/02/03-UCBBE-FMB-PL.nwc (plumbing)
                B1/L1/L2/L3-UCBB-CDC-MF.nwc (mechanical fabrication)

Technology Stack

Component	Technology	Cost
2D Campus Map	ArcGIS JS SDK 4.29	Free (developer tier)
3D Campus View	CesiumJS 1.119 + Google Photorealistic 3D Tiles	Free (Cesium ion)
MEP 3D Viewer	Three.js r160 + WebXR	Free (open source)
Energy Dashboard	Chart.js 4.4.1	Free (open source)
AI Chatbot	Cloudflare Workers AI (Llama 3.1 8B)	Free (Workers free tier)
Data Pipeline	Python 3.12 + GitHub Actions	Free (GitHub free tier)
Hosting	GitHub Pages	Free
Metering	BMO Obvius	UC Berkeley infrastructure
Total monthly cost		$0

Repository Structure

Graduate-Project/
├── grimes-campus-map-arcgis.html    # Campus 2D map with chatbot
├── campus-3d.html                   # CesiumJS 3D viewer
├── campus-energy-dashboard.html     # Chart.js dashboard
├── weekly-report.html               # Weekly report viewer
├── grimes-xr.html                   # Grimes MEP 3D viewer
├── grimes-mep-only.glb              # MEP 3D model (1.7MB)
├── grimes-mep-compressed.glb        # Full building model (57MB)
├── bmo_fetch_campus.py              # BMO data fetcher (26 buildings)
├── generate_weekly_report.py        # Weekly report generator
├── data/
│   ├── campus_energy.json           # Live energy snapshot
│   ├── campus_energy_history.json   # 24hr rolling history
│   ├── weekly_report.json           # 7-day summary
│   └── daily/                       # Permanent daily archives
│       └── YYYY-MM-DD.json
└── .github/
    └── workflows/
        └── bmo-campus-energy.yml    # 15-min cron job

Security Notes

The following files contain sensitive data and must NOT be committed to the public repository:

.env — BMO credentials
data/building_data.json — internal building configuration
raw_csv/ — raw BMO meter exports

Only code files and auto-generated output files (e.g., campus_energy.json) are safe to push.

Future Work

Restore 12 offline BMO meters (Facilities Services coordination)
Energy efficiency index: kW/sqft normalization with actual floor area data
Weather correlation: temperature vs. energy consumption using NOAA data
Grimes MEP time-linked visualization: pipe colors change by real-time load
Sub-metering integration: individual panelboard monitoring (93 panels identified)
BAS integration: real HVAC/plumbing sensor data to replace simulated values
Monthly automated PDF reports for Facilities Services
User testing with campus facility managers

Author

Yoonsung Chung M.S. Construction Systems · UC Berkeley CEE yoonsung_chung@berkeley.edu

Acknowledgments

UC Berkeley Facilities Services — BMO metering system access
Prof. Kenichi Soga — Thesis advisor
Tianyu — BMO data coordination
BIM model contributors: SOM (Architecture), Rutherford & Chekene (Electrical), Pribuss Engineering + Pragmatic PE (Fire Protection)