Название: Digital Twin: A Dynamic System and Computing Perspective
Автор: Ranjan Ganguli, Sondipon Adhikari, Souvik Chakraborty, Mrittika Ganguli
Издательство: CRC Press
Год: 2023
Страниц: 252
Язык: английский
Формат: pdf (true)
Размер: 28.3 MB

The digital twin of a physical system is an adaptive computer analog which exists in the cloud and adapts to changes in the physical system dynamically. This book introduces the computing, mathematical, and engineering background to understand and develop the concept of the digital twin. It provides background in modeling/simulation, computing technology, sensor/actuators, and so forth, needed to develop the next generation of digital twins. Concepts on cloud computing, Big Data, IoT, wireless communications, high-performance computing, and blockchain are also discussed.

A digital twin is an avatar of a real physical system which exists in the computer. In contrast to a computer model of a physical system which attempts to closely match the behavior of a physical system in a temporally static sense, the digital twin also tracks the temporal evolution of the physical system. The evolution of the computer replica with time is a key attribute of the digital twin. Some researchers have defined the digital twin at the conceptual level. However, these definitions have been very general due to the attempt to keep a very large number of systems within the ambit of the definitions. In this chapter, we seek to introduce the digital twin concept at a high level, while presenting the background needed to understand further chapters in the book.

The digital twin has theoretical foundations in information science, production engineering, Data Science and Computer Science. For most physical systems,
the virtual model will be a computer program which solves partial differential equations or matrix equations. This simulation model of the system must be verified and validated, typically with experimental data. The virtual model may need to be updated at this stage and the model fidelity improved to minimize the discrepancy between the physical and virtual model. Uncertainty analysis of the virtual model is typically conducted to account for deviations in the physical system properties. Statistical measures can be used to quantify the deviation between the physical model and the virtual model, and optimization methods can be used to minimize this difference. The second phase of the digital twin called data fusion involves collecting data from the system, typically using sensors. Some examples of sensors include pressure sensors, light sensors, accelerometers, gyroscopes, temperature sensors and motion sensors. Sensors can also be connected to prototyping boards such as the Arduino Uno and Raspberry Pi 2 which then allow them (sensors) to become part of the Internet of Things (IoT). Algorithms based on pattern recognition methods such as Machine Learning, fuzzy logic, etc., can also be used for data processing or feature extraction.

The third phase of the digital twin involving interaction and collaboration implies that there must be information flow between the physical model and the data fusion function of the digital twin. It is also possible that inputs generated by the virtual model are communicated to the physical model via actuators. Thus, the virtual model should incorporate changes in the physical model communicated through sensor data.

The final function of the digital twin is service, which is the reason for which the digital twin exists, such as “structure monitoring, lifetime forecasting, in-time manufacturing etc.” Estimation of service conditions and life of the physical system along with the prediction of required maintenance, downtime and replacement are some of the applications of the digital twin concept.

The idea of using modeling and simulation for the analysis, design and optimization of engineering systems is now ubiquitous. However, these ideas were applied for systems in general and not for a specific specimen of a system, such as a particular aircraft. Digital twins have become possible due to the Internet of Things (IoT), wireless communications, high-performance computing capability in the cloud framework and the ability to handle Big Data using signal processing and Machine Learning. Furthermore, areas such as cybersecurity have become important for the successful and safe deployment of digital twins.

Digital twin is defined as a virtual representation of a physical system. Digital twin can be easily represented in software as a simple multi-dimensional matrix array. In many instances, these arrays having multiple dimensions are called tensors. While the creation of the digital twin often needs the solutions of physics-based models such as partial differential equation, the computer considers the digital twin to be a tensor representation of the physical reality. The Chapter 2 addresses the aspects of the digital twin which are closely related to computer implementation. In fact, this compute content is critical for the successful manifestation of digital twin in the industry. Later chapters will focus on the mathematical infrastructure behind the digital twin concept. As shown, telemetry collection from sensors on the physical system requires an IoT defined framework including a system of databases and data-lakes, analysis using Machine Learning and Deep Learning which can be hosted at the end-point or in the cloud on popular cloud provider solution sites such as Amazon AWS, Microsoft Azure and Google GCE.

Features:

Provides background material needed to understand digital twin technology
Presents computational facet of digital twin
Includes physics-based and surrogate model representations
Addresses the problem of uncertainty in measurements and modeling
Discusses practical case studies of implementation of digital twins, addressing additive manufacturing, server farms, predictive maintenance, and smart cities

This book is aimed at graduate students and researchers in Electrical, Mechanical, Computer, and Production Engineering.