The rise of digital twins

Staff Writer

Nov 7, 2022

Model breaking down digital twin process

Part Two: A space ripe for innovation & industry collaboration

Supporting smart habitats and building operations

As part of a NASA-funded project, faculty members Mario Bergés and Burcu Akinci are designing systems for a smart habitat that supports life in deep space. The habitat will rely heavily on autonomous technologies, including detecting and fixing faults and predicting the impacts of anything that goes wrong. Of course, planning for every scenario is impossible.

“If I tell a robot, ‘Here are the specific solutions to fix these specific faults,’ it'll work up until any of my assumptions about what faults are or how to solve them fall apart,” says Bergés. “Say the temperature sensor shows that a room is very hot and then you take corrective action, but really the power to that sensor was awry so the readings were wrong. That's a situation that we can't anticipate entirely.”

To address new and unexpected issues, autonomous technologies will need to reason through what is happening and how to fix it. Doing that will require knowledge of the principles of physics as well as detailed knowledge of the habitat, including its composition and how each piece of technology behaves. “All of that context and domain knowledge should be embedded in the digital twin,” Bergés explains.

“To actually make this habit smart, there needs to be a thread that ties everything together,” adds Akinci. “That's where the digital twin comes in.”

Akinci and Bergés are also collaborating on a project with ANSYS using the company’s digital twin platform. For that project, CEE’s Porter Hall is serving as a testbed for integrating ANSYS’ physics-based simulations with building data for facility operations, including issue diagnosis, prognosis, and self-correction.

Increasing worker safety & efficacy

While typical home construction produces significant waste, modular home construction offers an efficient, affordable, and sustainable alternative. However, creating houses and building components in busy, packed factories is complex, with many potential hazards.

To help, faculty member Pingbo Tang is working with two manufacturers, Module and DMI Companies, to model human and machine behavior on production lines. To do so, he is using factory field notes, videos, and control system logs. The finished digital twin platform will generate data-driven suggestions for improving human-machine efficiency, safety, and health as well as adjusting workspaces for new components and designs.

Analyzing human behavior and decision-making in operations is a theme across much of Tang’s work. For example, working with the Federal Aviation Administration and NASA’s Ames Research Center, he created a digital twin of the Los Angeles airport air traffic. That simulation is allowing researchers to study how air traffic controllers make decisions as well as how AI learning from data-driven airport operation simulations could help prevent future mistakes.

To model how intelligence systems should help humans in these complex situations, we need algorithms for capturing and analyzing human behavior together with machine behavior.

Pingbo Tang , Associate Professor, Civil and Environmental Engineering

Elsewhere, Tang is modeling the actions of nuclear power plant workers with Arizona Public Services in the hopes of identifying what activities are essential during plant outages. Eventually, AI could learn from nuclear plants’ communication and operation histories and alert workers to take appropriate actions in similar contexts.

“My work is complementing machine learning with human learning. Can a machine observe and learn when and why experienced human operators perform better than automatic control systems?” says Tang. “To model how intelligence systems should help humans in these complex situations, we need algorithms for capturing and analyzing human behavior together with machine behavior.”

Faculty member Sean Qian's research on "improving urban mobility, emissions, and infrastructure systems working," involves large-scale dynamic network modeling and data analytics for transportation systems. In a recent collaboration with Honda USA, he replicated the flow of individual, ridesharing, and public transit vehicles through Columbus, Ohio. Now, he is using that digital twin to analyze the impact of increased electric vehicle usage on power grids and mobility systems as well as develop strategies to mitigate loads to both systems.

Qian is also modeling how people and traffic move through cities for a project with the technology company Fujitsu, using the Washington DC metro area as a case study. In that work, he is using data and algorithms to infer how people determine the need to travel, the destination, transportation mode, parking, and departure time. “We want to know not just how but why they’re using infrastructure,” says Qian. With this knowledge, cities could make changes that improve mobility, safety, emissions, and more.

Additionally, the model includes things like event traffic, accidents, and extreme weather to provide information for response and resource allocation plans. “Say we set the digital twin so that an incident on the highway closes two lanes at 4 pm. We may assume a fraction of people learn about it from their smartphones and others have no idea what's going on,” says Qian. “The simulation can show what happens if we focus on dispatching response teams faster. Maybe we invest in rerouting people and disseminating accurate, timely information. We might tune nearby signal timing. Once we know how people respond in these situations, these levels can all be examined in the digital twin.”