Digital Twins in Action

Get the eBook free when you register your print book at Manning.

"Makes the theory tangible and the tools accessible, from a single sensor to a city-scale system."
--Anthony Townsend, Cornell University

A "digital twin" is a virtual replica of a physical system that allows operations engineers to perform predictive maintenance, run risk-free simulations, and make smart, data-driven decisions. Digital Twins in Action teaches you how to combine methods from software development, data science, robotics, and visual design to create a powerful integrated platform. It uses both standard open sources tools and commercial frameworks. Grounded firmly in practical reality, it clarifies a topic clouded by marketing speak and vague definitions.

This hands-on book for software developers and automation engineers guides you step-by-step as you build a small-scale twin you can run in your office or home environment. Author and industry veteran Greg Biegel introduces essential mental models like the "Opportunity-to-Data" Mapping Chain and the Minimum Viable Digital Representation (MVDR) to ensure your design serves concrete goals. Across 10 in-depth chapters, you will learn to prioritize diverse information sources, extract legacy documents using OCR, deploy physical sensor networks over LoRaWAN or Wi-Fi, and organize real-time streams with standardized MQTT topic hierarchies. Reviewer Sanjay Basu, PhD, VP of Gen AI and GPU Infrastructure at Oracle remarked, "Digital Twins in Action takes you from a five-dollar temperature sensor to an AI agent making autonomous decisions!"

Because accurate visual simulations are an integral part of a digital twin, this book dives into modeling and visualization techniques that make virtual systems accessible to both humans and machines. You'll master data integration across relational databases, Parquet files, and Delta tables. Then, you'll construct a semantic model of reality using ontologies and knowledge graphs in Memgraph, render responsive 2D dashboards using Grafana, and locate 3D mesh models geospatially on an interactive digital globe using CesiumJS and WebGL.

Modern digital twins rely on advanced data analysis, integrating inference and simulation. You will build supervised image classifiers to act as virtual sensors, use unsupervised Isolation Forests for anomaly detection, and apply AutoGluon for time-series forecasting. You'll also explore continuous and discrete event simulations and learn how to construct autonomous AI agents that reason, plan, and optimize physical behaviors. Finally, you will learn to secure the cyber-physical boundary so you can confidently deploy, monitor, and scale a resilient, auditable digital twin.

What's inside

- Model physical environments using semantic knowledge graphs and ontologies
- How to deploy hybrid IoT sensor networks and stream real-time data
- Integrate predictive machine learning, physics simulations, and AI agents
- The five-level digital twin maturity model

About the reader

For software developers, data engineers, and architects who know the basics of machine learning. No prior experience in visual design or industrial automation required.

About the author

Greg Biegel has designed and built digital twins in production for over 20 years. His extensive industry experience includes six years spent developing a state-of-the-art industrial digital twin platform from the ground up for a leading Australian energy producer.

Table of Contents

1 Bridging the physical and digital worlds
2 Mapping physical systems to digital representations
3 Sensing the real world
4 Data integration and management
5 Modeling reality
6 2D Visualization and insight
7 Spatial context and 3D insight
8 Integrating inference and intelligence
9 Predicting outcomes with simulation
10 Digital twins in production
A Building a LoRaWAN network
B Building a custom IoT sensor
C Capturing a 3D model using photogrammetry