Deep neural operator-driven real-time inference to enable digital twin solutions for nuclear energy systems

Kazuma Kobayashi, Syed Bahauddin Alam

Research output: Contribution to journalArticlepeer-review


This paper focuses on the feasibility of deep neural operator network (DeepONet) as a robust surrogate modeling method within the context of digital twin (DT) enabling technology for nuclear energy systems. Machine learning (ML)-based prediction algorithms that need extensive retraining for new reactor operational conditions may prohibit real-time inference for DT across varying scenarios. In this study, DeepONet is trained with possible operational conditions and that relaxes the requirement of continuous retraining - making it suitable for online and real-time prediction components for DT. Through benchmarking and evaluation, DeepONet exhibits remarkable prediction accuracy and speed, outperforming traditional ML methods, making it a suitable algorithm for real-time DT inference in solving a challenging particle transport problem. DeepONet also exhibits generalizability and computational efficiency as an efficient surrogate tool for DT component. However, the application of DeepONet reveals challenges related to optimal sensor placement and model evaluation, critical aspects of real-world DT implementation. Addressing these challenges will further enhance the method’s practicality and reliability. Overall, this study marks an important step towards harnessing the power of DeepONet surrogate modeling for real-time inference capability within the context of DT enabling technology for nuclear systems.

Original languageEnglish (US)
Article number2101
JournalScientific reports
Issue number1
StatePublished - Jan 24 2024

ASJC Scopus subject areas

  • General


Dive into the research topics of 'Deep neural operator-driven real-time inference to enable digital twin solutions for nuclear energy systems'. Together they form a unique fingerprint.

Cite this