Digital twin projects look to slash nuclear O&M costs

Nine projects, financed by the U.S. Department of Energy, aim to decimate operations and maintenance (O&M) costs in the next generation of advanced nuclear reactors.

The $27-million funding for the projects forms part of the Advanced Research Projects Agency-Energy’s (ARPA-E) Generating Electricity Managed by Intelligent Nuclear Assets (GEMINA) program which looks at technologies including digital twins, artificial intelligence, advanced control systems, predictive maintenance and model-based fault detection.

Current generation nuclear plants already have higher O&M costs than other generating technologies, partly because of their age, so the projects will focus on the new generation, says ARPA-E Program Director Rachel Slaybaugh.

“If you design the next set of reactors with that same O&M philosophy you will continue to have high costs so, in the design process, let’s start thinking about how we can do this differently so they’re really lean and efficient,” Slaybaugh says.

Some of the projects will look directly at digital twin technology, whereby a complex, manipulatable simulation of an existing asset is created for operations, testing and investigation. The technique is already in use in less regulated industries and propels advances in everything from automotive design to NASA space missions.

A digital twin includes physical data from the original piece it is copying, historical knowledge on how that piece reacts under certain conditions, simulated data and machine learning, Slaybaugh explains. 

“Advance reactors obviously don’t exist yet so a lot of what we’re focusing on in the GEMINA program is, instead of a fully-simulated system, we can use what are called hardware-in-the-loop systems,” she says. 

The nuclear core could be simulated to get a heat profile, and the results would drive a heater on a physical flow loop with the fluids of interest. The thermal hydraulic measurements are then capturing the real physics instead of only having simulations, she says.

Taken from J. Buongiorno, M. Corradini, J. Parsons, D. Petti, “The Future of Nuclear Energy in a Carbon Constrained-World, An MIT Interdisciplinary Study”, MIT Energy Initiative, September 2018.

Safer and Cheaper

The GEMINA project aims to bring digital technology into the decision-making process to maintain safety and help keep the O&M costs down, says Nuclear Instrumentation and Controls Engineer at the Oak Ridge National Laboratory (ORNL) Pradeep Ramuhalli. 

“The digital twin not only helps determine what the condition of components is and what the predicted condition will be, but also we use the digital twin in a way that allows one to do planning, or ‘what if’ analysis,” Ramuhalli says.

ORNL will form part of a General Electric Research-led team, alongside Exelon Generation and the University of Tennessee-Knoxville, which will build a digital twin using General Electric Hitachi’s BWRX-300 SMR as a reference design.

GE is no stranger to creating digital twins and has developed and deployed well over 1.2 million digitals twins for products from everything from the transportation to energy sectors. 

Recent technological advances have brought the potential of a digital twin to the nuclear energy application area and accurate digital twin representations will help make new plants more efficient, economical and potentially be useful for training, Ramuhalli adds. 

The program comes alongside the DOE’s recently announced Advanced Reactor Demonstration Program (ARDP), which has called on participants to design, build and demonstrate two commercially-viable advanced reactors within the next five to seven years and is part of the U.S.’s push to regain leadership in the global nuclear industry. 

A Touch of Humility

Abhinav Saxena, a Senior Artificial Intelligence Scientist at GE Research and project leader on the AI-enabled predictive maintenance digital twins project, previously worked as a research scientist at NASA on algorithmic approaches to fast and reliable prognosis of failures in complex aerospace systems. 

His team is developing a Humble AI framework, a layer around all machine learning and AI models which assesses an AI model’s region of competency to establish trust in a model prediction before an automated action can be taken. 

That will help cut costs by moving from a time-based to a condition-based predictive maintenance framework with GEH’s SMR, says Saxena.

The Humble AI framework defaults to a known safe operation mode when the algorithms fail to recognize a certain situation, and so ensures systemic handling of uncertainties, data and model assurance.

The ARPA-E project forces the industry to take advanced methods used in other areas and apply them to nuclear, which due to heavy regulation, has not moved as rapidly with AI advancements, says Saxena.

“How far can we push one of these frontiers using cutting edge technology? That’s really where we want to bring in AI and digital twins; what if we can build these tools so well, we can push towards increasing more automation, or possible autonomy, in certain aspects?” he says.

High Fidelity Simulation

A Massachusetts Institute of Technology (MIT) team is also looking the GEH SMR and aims to assemble, validate and exercise high-fidelity digital wings of the BWRX-300 systems.

“Digital twin could mean many things. The one we’re using uses very high fidelity simulation to apply to certain narrow specific areas – predictive maintenance, in particular early fault detection, so we’re very focused on the high-fidelity tools application,” says MIT Nuclear Science and Engineering Professor Emilio Baglietto. 

Baglietto is working with fellow NSE Professor Koroush Shirvan to advance and demonstrate predictive maintenance approaches and model-based fault system detection techniques for the BWRX-300 but their work will extend to all reactors where a flowing liquid is present. 

The MIT team is concentrating on real cost analysis by producing accurate models of systems within the reactor.

“We can’t be kind of accurate, we have to be very accurate and that’s the challenge. We think we are at the level of the accuracy and level of certainty that we can do it,” says Baglietto. 

Other GEMINA projects include Electric Power Research Institute (EPRI) examining moving from a "maintain and repair" approach to a "replace and refurbish" approach, X-energy looking at O&M techniques via a digital twin for its  Xe-100 nuclear reactor design and Moltex Energy USA developing a multi-physics plant digital twin environment for its SSR-W. 

 By Paul Day