table of content

1 Bridging the physical and digital worlds

1.1 What is a digital twin?

1.2 Technologies enabling digital twins

1.2.1 Internet of things

1.2.2 Cloud computing

1.2.3 Machine learning and artificial intelligence

1.2.4 Visualization technologies and game engines

1.2.5 Agentic AI and autonomous decision-making

1.3 Digital twin capability levels

1.3.1 Descriptive digital twin

1.3.2 Informative digital twin

1.3.3 Predictive digital twin

1.3.4 Comprehensive digital twin

1.3.5 Autonomous digital twin

1.4 What are digital twins used for?

1.4.1 Accelerating product development

1.4.2 Reducing costs through predictive maintenance

1.4.3 Optimizing performance and operational efficiency

1.4.4 Training and simulation

1.5 Digital twins across industries

1.5.1 Mining, energy, and industrial

1.5.2 Automotive

1.5.3 Agriculture

1.5.4 Defense

1.5.5 Healthcare

1.5.6 Infrastructure

1.5.7 Smart buildings, cities, and states

1.5.8 What a digital twin is not good for

1.6 Building your first digital twin: Key considerations and challenges

1.6.1 A clear definition of success

1.6.2 Data quality

1.6.3 Skills gap

1.6.4 Build vs. buy

1.6.5 Governance and security

2 Mapping physical systems to digital representations

2.1 Defining your objectives

2.1.1 Identify specific challenges and opportunities

2.1.2 Establish measurable performance indicators and baselines

2.1.3 Map decision-making improvements

2.1.4 Define a minimum viable digital representation

2.2 A home digital twin to optimize energy and water use

2.2.1 Mapping the home to a digital representation

2.2.2 Start with what you have

2.3 Information sources for digital representation

2.3.1 Historical records

2.3.2 Photographs

2.3.3 Videos

2.3.4 Engineering documents

2.3.5 External systems

2.4 Spatial and geometric representations

2.4.1 2D geometric models

2.4.2 3D geometric models

2.5 Spatial mapping and reference systems

2.5.1 Coordinate systems and projections

2.5.2 Coordinate transformations and conversions

2.6 Deciding what you need

3 Sensing the real world

3.1 How sensors work

3.1.1 The sensor signal chain

3.1.2 Key sensor characteristics

3.2 Selecting the right sensors for your digital twin

3.2.1 Defining what you’re measuring

3.2.2 Matching sensor capabilities to real-world constraints

3.2.3 Verifying integration before purchasing

3.2.4 Choosing sensors for a home digital twin

3.3 Connecting sensors: Communication technologies

3.3.1 Short-range technologies (<100 m)

3.3.2 Long-range technologies (>100 m)

3.3.3 Choosing the right technology

3.3.4 Universal data transport architecture

3.4 Real-world data collection strategies

3.4.1 Edge processing

3.4.2 When electronic sensing isn’t enough

3.4.3 Manual sensing strategies

3.5 Data processing

3.5.1 Understanding IoT data characteristics

3.5.2 Message decoding

3.5.3 Sensor fusion

3.6 Scalability and device management

3.7 Bringing it all together

3.7.1 Data flow walkthrough

3.7.2 Architectural principles

4 Data integration and management

4.1 Types of data

4.1.1 Reference data

4.1.2 Time-series data

4.1.3 Unstructured and semi-structured data

4.1.4 Spatial data

4.1.5 Derived data

4.2 Data sources

4.2.1 OT data sources

4.2.2 IT data sources

4.2.3 The convergence of IT and OT

4.2.4 External data sources

4.3 Data structures and storage

4.3.1 Relational and transactional data

4.3.2 Columnar and analytical storage

4.3.3 Time-series data storage

4.3.4 Semi-structured and document data

4.3.5 Object storage, data lakes, and lakehouses

4.3.6 Specialized storage systems

4.3.7 Data life cycle management

4.4 Data ingestion

4.4.1 Batch data ingestion

4.4.2 Streaming data ingestion

4.4.3 API integration

4.4.4 ETL and data transformation

4.5 Data governance and compliance

4.5.1 Data classification and sensitivity

4.5.2 Regulatory compliance

4.5.3 Access control and authentication

4.5.4 Data quality

4.6 Selecting the right combination

4.7 Data store for a home digital twin

4.7.1 Storing time-series data

4.7.2 Storing object data

4.7.3 Storing relationship data

5 Modeling reality

5.1 Making sense of data

5.1.1 From measurements to decisions

5.1.2 The knowledge engineering challenge

5.2 Understanding context in digital twins

5.2.1 Types of context

5.2.2 How to contextualize data

5.2.3 The importance of context in generative AI

5.3 Ontologies as a blueprint for understanding

5.3.1 Core components of an ontology for a digital twin

5.3.2 Defining ontologies for digital twins

5.3.3 Choosing an ontology for your model

5.4 Knowledge graphs

5.4.1 Graph theory

5.4.2 Building a knowledge graph as a labeled property graph

5.4.3 Graph traversals

5.4.4 Building a knowledge graph of my home

5.4.5 Choosing a knowledge graph service

5.5 Standards and interoperability

5.5.1 A framework for implementing standards in your digital twin

5.5.2 Practical standards assessment

5.5.3 A practical assessment framework

5.5.4 Real-world example: The Asset Administration Shell

5.6 Serving the model of reality

5.6.1 Context-aware data retrieval

5.6.2 Building an API layer for my home digital twin

5.6.3 The power of a GraphQL schema

6 2D Visualization and insight

6.1 Why graphical visualization matters in digital twins

6.1.1 Data-to-visual mapping principles

6.1.2 User experience considerations

6.2 2D visualization techniques

6.2.1 Charts, graphs, and data visualization

6.2.2 Dashboards and control panels

6.2.3 Schematic diagrams

6.2.4 Spatial 2D visualization

6.2.5 Heatmaps

6.3 2D visualization performance considerations

6.3.1 Choosing the right rendering technology

6.3.2 Data decimation and aggregation for time-series data

6.4 Building effective visualizations with Grafana

6.4.1 Why Grafana works well for digital twins

6.4.2 Choosing a visualization platform

7 Spatial context and 3D insight

7.1 3D visualization fundamentals for digital twins

7.1.1 More complex models

7.1.2 Texturing the mesh

7.1.3 Web-based 3D visualization

7.2 Creating a mesh-based 3D geometric model

7.2.1 Serving the model via the web

7.2.2 Spatially locating the model

7.3 Data integration

7.3.1 The importance of spatially locating data in 3D

7.3.2 Annotation and markup

7.3.3 3D visualization performance considerations

7.4 Capturing reality in 3D

7.4.1 360-degree photography

7.4.2 Photogrammetry

7.4.3 Neural Radiance Fields

7.4.4 Gaussian splatting

7.4.5 Light detection and ranging

7.4.6 Simultaneous localization and mapping

7.4.7 Universal Scene Description and the NVIDIA Omniverse

7.5 Extended reality

7.5.1 Virtual reality: Full immersion

7.5.2 Augmented reality: Contextual overlay

7.5.3 Mixed reality: Blending worlds

7.6 Building effective visualizations

7.6.1 Security and access control

7.6.2 Usability and user experience

7.6.3 Data-to-visualization pipeline and scalability

8 Integrating inference and intelligence

8.1 A practical data pipeline

8.1.1 Cleaning and normalization

8.1.2 Time alignment and resampling

8.1.3 Feature extraction

8.1.4 Storage

8.2 Anomaly and fault detection

8.2.1 Rule-based detection and its limitations

8.2.2 Statistical approaches

8.2.3 Unsupervised learning

8.3 Predictive modeling and machine learning

8.3.1 Training a forecasting model

8.3.2 Applying AutoML

8.3.3 Supervised learning

8.4 Generative AI

8.4.1 Retrieval-augmented generation

8.4.2 Agentic AI

8.4.3 Building an agent for the home digital twin

8.4.4 When agents complement or replace control strategies

8.4.5 Multi-agent systems

8.5 Choosing the right approach

8.5.1 Operational viability

8.5.2 Safety and human in the loop

8.5.3 Cost

9 Predicting outcomes with simulation

9.1 Modeling and simulating physical systems

9.1.1 The digital twin advantage

9.1.2 From static models to dynamic simulation

9.1.3 Continuous versus discrete event simulation

9.2 Continuous systems

9.2.1 Simulating a continuous system

9.2.2 The initial value problem

9.3 Acausal modeling for digital twins

9.3.1 How it works

9.3.2 Modeling tank drainage in OpenModelica

9.3.3 Executing simulations

9.3.4 Integrating sensor data with continuous simulation

9.4 Simulating processes with discrete event simulation

9.4.1 Why DES matters for digital twins

9.4.2 DES with Python using SimPy

9.4.3 Example application of DES in digital twins

9.5 Finite element and field simulation

9.5.1 Using FEM for virtual sensing in a digital twin

9.5.2 Example application of FEA in digital twins

9.6 Computational fluid dynamics

9.6.1 CFD with Python for lightweight simulation

9.6.2 Example application of CFD in digital twins

9.7 Reduced-order models

9.7.1 ROMs for virtual sensing

9.7.2 An example ROM

10 Digital twins in production

10.1 Selecting the right deployment platform

10.1.1 Public cloud providers

10.1.2 Digital twin development platforms and engines

10.1.3 Fully managed digital twin platforms

10.2 Security as a cross-cutting concern

10.2.1 Expanding the threat model

10.2.2 Identity everywhere

10.2.3 Policy as code

10.3 Digital twin operations

10.3.1 Observability

10.3.2 Model lifecycle management

10.3.3 Change management

10.3.4 Cost management

10.3.5 Software dependency management

10.4 Aligning to strategy and measuring value

10.4.1 Operational impact

10.4.2 User adoption and trust

10.4.3 Financial value

10.5 Governance, ethics, and responsibility

10.5.1 Decision authority

10.5.2 Accountability and ownership

10.5.3 Ethical use

10.5.4 Auditability and traceability

10.6 A production readiness checklist

Appendices

Appendix A: Building a LoRaWAN network

A.1 What you’ll need

A.1.1 LoRaWAN gateway

A.1.2 LoRaWAN network server

A.1.3 LoRaWAN sensors

A.2 Step 1: Configuring AWS IoT Core for LoRaWAN

A.3 Step 2: Configuring your gateway

A.4 Step 3: Configuring device and service profiles

A.4.1 Creating the device profile

A.4.2 Creating the service profile

A.5 Step 4: Adding a destination for your sensor data

A.6 Step 5: Adding a sensor

Appendix B: Building a custom IoT sensor

B.1 Hardware

B.2 Cloud configuration

B.2.1 Step 1: Create a thing

B.2.2 Step 2: Generate certificates

B.2.3 Step 3: Attach an access policy

B.2.4 Step 4: Download certificates

B.3 Device software

B.4 Verifying the connection

Appendix C: Capturing a 3D model using photogrammetry

C.1 What you’ll need

C.2 Image capture

C.2.1 Preparing your camera

C.2.2 Manual photo capture

C.2.3 Automated photo capture

C.3 Processing the data

C.4 Creating 3D tiles and georeferencing them

Overview

4 Data integration and management

Data is presented as the lifeblood of a digital twin, making integration and management the foundation for accuracy, trust, and impact. The chapter catalogs the main data types—reference, timeseries, unstructured and semi-structured objects, spatial, and derived data (including features and embeddings)—explaining how each contributes context, behavior, and semantics. It then maps the principal sources: operational technology (SCADA, historians, IIoT), information technology (ERP, finance, BI), and external feeds, emphasizing that coherent combination across these domains turns raw signals into decision-ready intelligence.

To operationalize this, the chapter surveys data structures and pipelines matched to access patterns: relational, columnar, graph, document, and key-value models; ingestion modes spanning batch for historical context, streaming with sliding windows for live signals, and API integrations for on-demand retrieval; and ETL/ELT to cleanse, standardize, and enrich. It addresses the realities of IT/OT convergence—security constraints, air-gapped networks, proprietary protocols—and recommends mirrored data, demilitarized zones, protocol translation, and layered architectures to protect control systems while enabling analytics.

Storage and governance complete the blueprint. The chapter advocates polyglot persistence—timeseries databases for telemetry, analytical warehouses or columnar files for large-scale queries, data lakes/lakehouses for heterogeneous assets, transactional stores for reference data, and specialized systems such as object, graph, vector, and feature stores—combined with lifecycle tiering (hot/warm/cold) to balance performance and cost. Robust governance spans data classification, regulatory compliance, zero-trust access control, and rigorous data quality management. Selection criteria focus on performance, scale, consistency, cost, and integration complexity, illustrated by a practical example that pairs serverless key-value storage for timeseries, object storage for media, and a graph database for relationships in a home digital twin.

Examples of unstructured data stored in my home digital twin, with an image taken by an IoT camera shown to the left, and a device manual on the right.

A 3D mesh model of my home that I will use in my digital twin with the coordinates displayed for three sensors I have deployed to measure water usage around the home. Listing 4.2 shows how I can use the coordinates to find objects within five meters of the person.

The relationship between a raw data store, derived data in a feature store, and the training and serving of ML models.

An example of how higher dimensionality raw data can be converted to a compact format that encodes the similarity in operating parameters between Pump A and C, while showing the difference in Pump B.

The Purdue model of IT/OT convergence segments the network into layers with a DMZ acting as a buffer zone between the OT and IT networks.

A relational model of a work order showing tables and properties representing entities, and their relationships.

A house, its rooms, and their devices modelled as a graph structure.

A sliding window of length four over a data stream. As the window slides through time, older readings fall out, and new readings are added.

An example lifecycle policy for data stored in Amazon S3 that transitions data to cheaper storage and eventual deletion as it ages.

Summary

There are a number of different types of data that are found in most digital twins including timeseries, relational, object, spatial, and derived data which includes features and embeddings.
Digital twins source data from both internal data stores, and from external sources.
Both operational systems and information technology systems within the organization are important sources of data that digital twins leverage, with these traditionally separate systems beginning to converge.
There are different strategies digital twins adopt to ingest data including batch-based, streaming, and API-based integrations.
There are a range of modern data storage technologies that can be adopted to store the types of data a digital twin uses.
Digital twins aggregate data from many sources and governance and management of compliance of this data is an important consideration.

FAQ

What types of data do digital twins handle, and why do they matter?

Digital twins typically work with five key data types: reference data (stable asset/configuration context), timeseries data (timestamped operational signals), unstructured and semi-structured data (documents, images, videos, configs), spatial data (2D/3D location and geometry), and derived data (features and embeddings created from raw data). Each plays a distinct role in making the twin accurate, contextual, and useful for analysis and decision-making.

Where does the data for a digital twin come from?

Three main domains feed a twin: operational technology (OT) systems such as SCADA/DCS/PLCs and IIoT devices; information technology (IT) systems like ERP, finance, procurement, and BI; and external sources including third-party APIs and open datasets. Together they provide real-time state, business context, and environmental context.

Why is integrating OT data challenging, and how is it typically addressed?

OT environments prioritize safety and determinism, often run air-gapped networks, and use proprietary protocols, making direct queries risky and impractical. Common patterns include mirroring/process historians, data diodes or unidirectional gateways, and an industrial DMZ with protocol converters and security gateways, all aligned to the Purdue model to safely bridge OT and IT.

How should I choose among relational, columnar, graph, document, and key-value structures?

Use relational models for consistency and flexible ad-hoc queries; columnar layouts for efficient analytics and aggregations; graph structures for traversing rich relationships (assets-locations-dependencies); document stores for flexible, evolving schemas; and key-value stores for predictable, high-throughput lookups and caching. Pick based on access patterns and consistency needs.

When should I use batch, streaming, or API-based ingestion?

Batch is ideal for slow-moving enterprise data and historical backfills with complex transforms. Streaming suits live sensor/telemetry where you need rolling aggregates (e.g., sliding windows) and alerts. APIs enable targeted, on-demand collection (REST/GraphQL/gRPC/WebSockets) that can be scheduled or event-driven without constant ingestion.

What is the role of ETL/ELT in a digital twin?

ETL/ELT standardizes and enriches data: cleansing, unit conversion, validation, aggregation, and contextual joins (e.g., adding asset metadata). ETL transforms before loading; ELT loads raw data first and transforms in the target system (common with cloud warehouses). Both ensure consistent, analytics-ready data for the twin.

Which storage systems best fit common digital twin workloads?

Timeseries databases (e.g., InfluxDB, TimescaleDB, Timestream) optimize temporal writes, compression, aggregations, and time partitioning. Analytical warehouses/lakehouses (e.g., Redshift, BigQuery, Snowflake, Delta/Parquet on object storage) handle large-scale analytics and diverse formats. Transactional stores (relational/NoSQL) manage reference and configuration data. Specialized systems include object storage for media, graph databases for relationships, vector databases for embeddings, and feature stores for ML.

How do I manage data across its lifecycle (hot, warm, cold)?

Keep recent “hot” data on fast storage for low-latency decisions and dashboards; move older “warm” data to cost-efficient stores with compression/aggregation for reporting and model training; archive “cold” data to low-cost tiers (e.g., Glacier/Archive) with lifecycle policies, accepting longer retrieval times.

What are the essentials of data governance and compliance for digital twins?

Define classification (public, internal, confidential, restricted); meet regulatory obligations (e.g., GDPR, HIPAA, PCI) including minimization, portability, deletion, and audit trails; enforce RBAC/ABAC with zero-trust principles; and implement strong data quality rules, monitoring, and remediation to keep the twin reliable and safe.

What does a practical polyglot storage design look like in practice?

A typical design pairs purpose-picked stores: e.g., timeseries in a key-value or timeseries DB (partitioned by sensor and timestamp), object storage for documents and media, a graph DB for relationships, and an analytical layer (warehouse or Parquet/Delta in a lakehouse) for large-scale queries. The home twin example uses DynamoDB for timeseries, S3 for objects/exports, and Memgraph for relationships, trading some specialization for simplicity, scalability, and cost control.

pro $24.99 per month

access to all Manning books, MEAPs, liveVideos, liveProjects, and audiobooks!
choose one free eBook per month to keep
exclusive 50% discount on all purchases
renews monthly, pause or cancel renewal anytime

lite $19.99 per month

access to all Manning books, including MEAPs!

team

5, 10 or 20 seats+ for your team - learn more

eBook

pdf, ePub, online

$55.99 $27.99

you save $28.00 (50%)

include audio $24.99 $12.49

pro $24.99 per month

access to all Manning books, MEAPs, liveVideos, liveProjects, and audiobooks!
choose one free eBook per month to keep
exclusive 50% discount on all purchases
renews monthly, pause or cancel renewal anytime

lite $19.99 per month

access to all Manning books, including MEAPs!

team

5, 10 or 20 seats+ for your team - learn more

eBook

$55.99 $27.99

you save $28.00 (50%)

include audio $24.99 $12.49

eBook

pdf, ePub, online

$55.99 $27.99

you save $28.00 (50%)

include audio $24.99 $12.49

pro $24.99 per month

access to all Manning books, MEAPs, liveVideos, liveProjects, and audiobooks!
choose one free eBook per month to keep
exclusive 50% discount on all purchases
renews monthly, pause or cancel renewal anytime

lite $19.99 per month

access to all Manning books, including MEAPs!

team

5, 10 or 20 seats+ for your team - learn more