---
id: wiki-2026-0508-data-twins
title: Data Twins (Digital Twins)
category: 10_Wiki/Topics
status: verified
canonical_id: self
aliases: [Digital Twin, Cyber-Physical Twin, Virtual Replica]
duplicate_of: none
source_trust_level: A
confidence_score: 0.88
verification_status: applied
tags: [digital-twin, iot, simulation, industry-4-0, modeling]
raw_sources: []
last_reinforced: 2026-05-10
github_commit: pending
tech_stack:
  language: Python/C++
  framework: Azure Digital Twins / Omniverse / Modelica
---

# Data Twins (Digital Twins)

## 매 한 줄
> **"매 digital twin 의 핵심: live data binding + physics-aware simulation + bidirectional sync"**. 매 2002 Michael Grieves 의 PLM 컨셉 으로 시작, 매 NASA Apollo 13 의 ground simulation 이 ancestor. 매 2026 현재 NVIDIA Omniverse, Azure Digital Twins, AWS IoT TwinMaker, 매 LLM-grounded 산업 simulation 으로 manufacturing / smart-city / healthcare 의 mainstream.

## 매 핵심

### 매 3 fidelity levels
- **Descriptive twin**: 매 static data + dashboard.
- **Predictive twin**: 매 ML / physics simulation — 매 forecast.
- **Prescriptive twin**: 매 optimize + actuate back to physical asset.

### 매 components
- **Sensor layer**: 매 IoT (MQTT, OPC UA, CAN bus).
- **Time-series store**: 매 InfluxDB, Timescale, AWS Timestream.
- **Twin graph**: 매 ontology (DTDL, asset hierarchy).
- **Simulation kernel**: 매 Modelica, Omniverse PhysX, OpenFOAM.
- **Closed-loop controller**: 매 actuator command back.

### 매 응용
1. Manufacturing (Siemens, GE Predix — turbine twin).
2. Smart city (Singapore Virtual Singapore, Shanghai twin).
3. Healthcare (heart twin for surgery planning).
4. Supply chain (warehouse / fleet simulation).
5. Building / HVAC optimization (BIM + live sensor).

## 💻 패턴

### DTDL (Digital Twins Definition Language)
```json
{
  "@context": "dtmi:dtdl:context;3",
  "@id": "dtmi:com:example:Turbine;1",
  "@type": "Interface",
  "displayName": "Turbine",
  "contents": [
    { "@type": "Telemetry", "name": "rpm", "schema": "double" },
    { "@type": "Telemetry", "name": "tempC", "schema": "double" },
    { "@type": "Property", "name": "model", "schema": "string" },
    { "@type": "Command", "name": "shutdown" }
  ]
}
```

### Azure Digital Twins (Python SDK)
```python
from azure.digitaltwins.core import DigitalTwinsClient
from azure.identity import DefaultAzureCredential

client = DigitalTwinsClient(url, DefaultAzureCredential())
twin = {
  "$metadata": {"$model": "dtmi:com:example:Turbine;1"},
  "rpm": 3500.0, "tempC": 78.5, "model": "T-900"
}
client.upsert_digital_twin("turbine-42", twin)
# Query
for t in client.query_twins("SELECT * FROM digitaltwins WHERE tempC > 80"):
    print(t)
```

### MQTT ingestion → twin update
```python
import paho.mqtt.client as mqtt, json

def on_message(client, userdata, msg):
    data = json.loads(msg.payload)
    update_twin(data["device_id"], {"rpm": data["rpm"]})

c = mqtt.Client()
c.on_message = on_message
c.connect("broker.local", 1883)
c.subscribe("plant/+/telemetry")
c.loop_forever()
```

### Physics simulation (FMU via Modelica)
```python
from fmpy import simulate_fmu
result = simulate_fmu("turbine.fmu",
                      start_time=0, stop_time=60, step_size=0.01,
                      input=[("inlet_pressure", input_signal)])
```

### NVIDIA Omniverse (USD asset twin)
```python
import omni.usd
from pxr import UsdGeom, Gf
stage = omni.usd.get_context().get_stage()
turbine = UsdGeom.Xform.Define(stage, "/World/Turbine")
turbine.AddRotateYOp().Set(Gf.Vec3f(0, rpm * dt * 6, 0))  # live RPM
```

### Anomaly-driven actuation (closed loop)
```python
def control_loop(twin):
    if twin.tempC > 95:
        send_command(twin.id, "reduce_load", value=20)
        log(f"Twin {twin.id} thermal protection triggered")
```

### LLM-augmented twin Q&A
```python
import anthropic
client = anthropic.Anthropic()
def ask_twin(twin_state, question):
    return client.messages.create(
        model="claude-opus-4-7-20260101",
        max_tokens=512,
        system="You are an expert in industrial twin diagnostics.",
        messages=[{"role": "user",
                   "content": f"State: {twin_state}\nQ: {question}"}]
    ).content[0].text
```

## 매 결정 기준
| 상황 | Approach |
|---|---|
| Asset monitoring only | Descriptive (dashboard) |
| Predictive maintenance | Predictive (ML on telemetry) |
| Autonomous operation | Prescriptive (closed-loop) |
| 3D / VR walkthrough | Omniverse / USD |
| Cloud-managed | Azure Digital Twins / AWS TwinMaker |
| Edge constraints | Local twin + sync (KubeEdge) |

**기본값**: 매 industrial use 의 Azure Digital Twins + DTDL ontology, 매 3D viz 의 Omniverse.

## 🔗 Graph
- 부모: [[Cyber-Physical Systems]]
- 변형: [[Predictive Maintenance]]
- Adjacent: [[클라우드_인프라_및_IaC_운영_표준|IoT]] · [[Edge Computing]] · [[Time Series]]

## 🤖 LLM 활용
**언제**: 매 twin schema (DTDL) drafting, 매 anomaly explanation, 매 operator natural-language query, 매 simulation scenario generation.
**언제 X**: 매 hard-real-time control loop — 매 LLM latency 의 unfit. 매 deterministic control 은 PID / MPC.

## ❌ 안티패턴
- **Twin = dashboard 의 단순 rebrand**: 매 simulation / closed-loop 없으면 그냥 monitoring.
- **No data quality validation**: 매 garbage sensor → garbage twin.
- **Twin without versioning**: 매 schema drift / model evolution 의 disaster.
- **Tight coupling to vendor**: 매 vendor lock — 매 DTDL / OPC UA 같은 standards 사용.
- **Ignoring security**: 매 closed-loop = attacker 의 actuation = physical damage.
- **One twin for everything**: 매 hierarchical decomposition (asset → system → plant) 의 사용.

## 🧪 검증 / 중복
- Verified (Grieves 2002, NASA twin paradigm, Microsoft DTDL spec, ISO 23247, NVIDIA Omniverse docs 2025).
- 신뢰도 B+ (terminology / scope 의 industry variation).

## 🕓 Changelog
| 날짜 | 변경 |
|---|---|
| 2026-05-08 | Phase 1 |
| 2026-05-10 | Manual cleanup — Digital twin patterns + Omniverse / DTDL |