Generative AI and Digital twins - simulating and modelling complex, cyber physical systems

?Abstract

This post and the subsequent post are inspired by a set of posts from Dirk Hartmann

Dirk is a part of our #AI for #Digitaltwins course at the #universityofoxford and we share many common ideas and views on the future of digital twins

Dirk shared

???????? ???? ???????? ???????????????????? ???????????????? ???? ?? ?????? ?????????????

???????????????????? ???????????????????????? – ?????????????????? ???? ?????? ?????? ???????? ?!?

and he proposes that?

?????????????????? ???????????????????? ?????? ????????????: To address this, we need approaches that leverage existing knowledge or integrate physics reasoning within models to reduce data requirements. hashtag#DigitalTwins curating knowledge beyond data are key to achieve scalable solutions.

Some promising technologies include:

??? ???????????? ???????????????????? ????????????: Combining physics-based models and data-based models.

?? ??????????????-???????????????? ????????????????????????: Incorporating physics knowledge in the learning function.

?? ?????????????????? ????????????: Allowing reasoning on well-curated data (e.g., as prototyped in our HiSimcenter Demo).

?? ??????????-?????????????????? ?????????????? ????????????????: Unearthing first-principle relations (e.g., AI-Descartes, AI Feynman, AI-Lorentz).

All of the above are valid and you should look at these processes. Certainly, we are looking at hybrid simulation models and knowledge graphs. But I also want to propose some additional ideas.

In part one of this two part post, I propose that:

• digital twins can ideally model cyber physical systems?
• Digital twins can model complex systems.?
• The exact mechanism to do so is related to simulation.??
• simulation cannot easily handle emergent behaviour

Following this, in part two I propose the role of AI in enhancing simulation and the approach I am following most closely in my work at Oxford,??

Understanding Digital Twins

At a minimum, a digital twin is a digital replica of a physical entity. Hence, a digital twin ecosystem should comprise of a physical entity that you wish to model, a digital equivalent and connectivity between the two to maintain / synchronise the digital and physical entities? Ideally,? this connectivity should be real-time. The twin also needs some form of data visualisation capabilities to convey this information so that action could be undertaken.????

Beyond this, the digital twin could also include other functionality

• Modeling and Simulation??
• Lifecycle Management
• Analytics and Insights?
• Scalability and security?
• AR/VR and immersive visualization
• Advanced insights like anomaly detection
• High fidelity models
• Lifecycle management?
• Real time collaboration?
• Handling complex business processes like sustainability??

More broadly, we can think of digital twins as a mechanism to model cyber physical systems.?

?Modelling cyber physical systems

A Cyber-Physical System (CPS) is an integrated system that combines computational elements with physical processes. These systems involve the close interaction between digital (cyber) components such as computers, networks, and software, and physical components like sensors, actuators, and mechanical devices. The aim is to enable real-time monitoring, control, and optimization of complex processes and operations.

Key Characteristics of Cyber-Physical Systems include

• Integration of Cyber and Physical Components
• Computational Elements: Includes processors, software, and networks that handle data processing, computation, and communication.
• Physical Elements: Comprises sensors, actuators, and other hardware that interact with the physical environment.
• Real-Time Data Processing
• Sensing: Sensors collect real-time data from the physical environment.
• Actuation: Actuators receive commands from the cyber system to perform physical actions.
• Feedback
• Automation and control#
• Autonomous operation?

?There are a number of domains that use cyber physical systems: Smart Grids, Autonomous Vehicles, Smart Manufacturing, Healthcare Systems and Building Automation Systems.?

Increasingly, Digital twins will also enable more sophisticated simulations of complex systems, such as entire cities or large industrial plants, providing insights into system behavior under various conditions.?

What is a complex system?

A complex system is a system composed of interconnected and interdependent components that exhibit collective behavior not obvious from the individual behavior of the components. These systems are characterized by intricate interrelationships and interactions that give rise to emergent properties, which are properties that cannot be predicted simply by analyzing the system's parts. Complex systems are found in various domains, including natural, social, and engineered environments.

Key Characteristics of Complex Systems are Interconnected Components, Nonlinearity, Emergent Behavior, Adaptability and Evolution, Dynamism, Diversity. Examples of complex systems are Internet, Climate System etc.? Modelling complex systems poses some unique challenges such as high complexity, computation, unpredictability etc.

?A digital twin can be used to represent complex systems. ? Examples of Digital Twins for Complex Systems include Smart Cities, Healthcare Systems, Manufacturing Systems, Energy Systems, Power Grids etc.????

So far, we have have proposed that digital twins can ideally model cyber physical systems and complex systems. The exact mechanism to do so is related to simulation.??

What is simulation and what are the techniques of simulation?

Simulation is a technique used to model the operation of a system, process, or phenomenon through the use of mathematical models, computer programs, and other methodologies.

?It involves creating a digital or physical representation of a real-world system? to study its behavior under various conditions without interacting with the real system.?

Simulations are used to analyze complex systems, predict outcomes, optimize processes, and train individuals in various fields.??

Simulation techniques? include - Discrete Event Simulation (DES); Continuous Simulation; Agent-Based Simulation (ABS); Monte Carlo Simulation; System Dynamics (SD) etc.?

A flight simulator is an example of a simulation. .?

Digital twins use simulations to predict? conditions that may not be easily observed such as boundary or failure conditions.?

Rules are an integral part of simulations and are built into the simulation model. These rules define how the system behaves, how different elements interact, and how various processes unfold over time. The rules ensure that the simulation accurately represents the real-world system or phenomenon being studied.?

For example, for manufacturing simulation, Machine Operation Rules: Define the operation times, failure rates, and maintenance schedules of machines.

Emergent behaviour

So far, we have said that?

digital twins can ideally model cyber physical systems and complex systems. The exact mechanism to do so is related to simulation.??

Now, we propose that simulation cannot easily handle emergent behaviour

Complex systems often exhibit emergent behavior, where the system's collective behavior is not easily inferred from individual components. Capturing and understanding these phenomena is challenging. The unpredictable nature of emergent behavior makes designing models that can anticipate and accurately represent such phenomena difficult. Hence, we see complex systems failing in unpredictable and catastrophic ways as we saw in crowdstrike.?

Here, we propose that AI could be used to enhance simulation especially to cater for emergent phenomenon

We will discuss this more in the next section about how AI can be used to enhance simulation.?