Kazuma Kobayashi
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Kazuma Kobayashi has a total of 13 co-authors. Here are the top co-authors by number of shared publications:
- Syed Bahauddin Alam
National Center for Supercomputing Applications Plasma & Radiological Engineering University of Illinois at Urbana-Champaign 61801 Urbana IL USA
5 shared publications Jaewan Park
Mechanical Science and Engineering Department, The Grainger College of Engineering, University of ILlinois Urbana-Champaign, Urbana, USA
2 shared publications- Souvik Chakraborty
Department of Applied Mechanics Department of Applied Mechanics Yardi School of Artificial Intelligence (ScAI) Indian Institute of Technology Indian Institute of Technology Delhi
2 shared publications Diab Abueidda
National Center for Supercomputing Applications, University of Illinois at Urbana-Champaign, Urbana, IL, USA
1 shared publicationFarid Ahmed
Nuclear, Plasma, & Radiological Engineering Department, The Grainger College of Engineering, University of Illinois Urbana-Champaign, Urbana, USA
1 shared publication- Farid Ahmed
Nuclear, Plasma & Radiological Engineering, University of Illinois Urbana-Champaign, Urbana, Illinois, USA
1 shared publication
Latest Publications
- Open Access
- 19 September 2026
- Published by Springer Nature in Nature Communications
- Vol. 17 (1)
Virtual sensing to enable real-time monitoring of inaccessible locations & unmeasurable parameters
Real-time monitoring of safety-critical interior states remains an open problem across energy, environmental, and industrial systems where direct instrumentation is infeasible. Existing approaches based on governing equations, discrete state vectors, or fixed sensor locations fail to provide mesh-independent, field-level reconstructions at arbitrary interior coordinates under real-time constraints. Here we introduce neural operator-based virtual sensing as a general framework for recovering inaccessible interior fields from sparse, heterogeneous measurements and operating conditions, instantiated with MIMONet, a multi-input, multi-output neural operator that fuses heterogeneous inputs and decodes coupled physical fields through a shared latent representation. Across three engineering-grade evaluations of escalating complexity, from confined recirculating flows to pressurized water reactor subchannels and compact power-system heat exchangers, MIMONet achieves ≤5% relative reconstruction error, while generating full-field predictions with sub-millisecond inference on an NVIDIA H200, at least five orders of magnitude faster than traditional high-fidelity CFD simulations, with calibrated uncertainty and noise resilience demonstrated on the heat-exchanger case. We validate the framework on three independent real-world datasets spanning electrochemical energy, atmospheric science, and physical oceanography: current-density mapping in a hydrogen fuel cell, wind-speed profiling on instrumented meteorological towers, and North Atlantic ocean-state from a global ocean reanalysis. MIMONet improves over classical virtual-sensing baselines by 72–84%: 84% on fuel-cell internal-field recovery from 10% sensor coverage and 83% on hub-height wind-speed recovery from lower-level tower sensors. Most notably, it recovers sea-surface height, a field with zero direct sensors, from its learned coupling to observed temperature and salinity at 4.1 × lower error than the best classical predictor, a cross-modal field recovery beyond the reach of classical interpolation. These results establish operator-based virtual sensing as a practical route to real-time field observability in systems where the states that matter most cannot be directly measured.
- 19 September 2026
- Published by Springer Nature in Nature Communications
- Vol. 17 (1)
Open AccessVirtual sensing to enable real-time monitoring of inaccessible locations & unmeasurable parameters
Real-time monitoring of safety-critical interior states remains an open problem across energy, environmental, and industrial systems where direct instrumentation is infeasible. Existing approaches based on governing equations, discrete state vectors, or fixed sensor locations fail to provide mesh-independent, field-level reconstructions at arbitrary interior coordinates under real-time constraints. Here we introduce neural operator-based virtual sensing as a general framework for recovering inaccessible interior fields from sparse, heterogeneous measurements and operating conditions, instantiated with MIMONet, a multi-input, multi-output neural operator that fuses heterogeneous inputs and decodes coupled physical fields through a shared latent representation. Across three engineering-grade evaluations of escalating complexity, from confined recirculating flows to pressurized water reactor subchannels and compact power-system heat exchangers, MIMONet achieves ≤5% relative reconstruction error, while generating full-field predictions with sub-millisecond inference on an NVIDIA H200, at least five orders of magnitude faster than traditional high-fidelity CFD simulations, with calibrated uncertainty and noise resilience demonstrated on the heat-exchanger case. We validate the framework on three independent real-world datasets spanning electrochemical energy, atmospheric science, and physical oceanography: current-density mapping in a hydrogen fuel cell, wind-speed profiling on instrumented meteorological towers, and North Atlantic ocean-state from a global ocean reanalysis. MIMONet improves over classical virtual-sensing baselines by 72–84%: 84% on fuel-cell internal-field recovery from 10% sensor coverage and 83% on hub-height wind-speed recovery from lower-level tower sensors. Most notably, it recovers sea-surface height, a field with zero direct sensors, from its learned coupling to observed temperature and salinity at 4.1 × lower error than the best classical predictor, a cross-modal field recovery beyond the reach of classical interpolation. These results establish operator-based virtual sensing as a practical route to real-time field observability in systems where the states that matter most cannot be directly measured.
- Open Access
- 4 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 104-107
Neuroscience Inspired Graph Operators Towards Edge-Deployable Virtual Sensing for Irregular Geometries
Real-time virtual sensing can enhance instrumentation in engineering systems but requires sparse-to-dense reconstruction on irregular grids under strict computational constraints. Neural operators have shown promise, but few address power consumption. Spiking networks offer a solution through neuromorphic deployment, but existing spiking neuron models degrade significantly in regression tasks. To address this, we present the Variable Spiking Graph Neural Operator (VS-GNO), combining spectral-spatial graph operator learning with the Variable Spiking Neuron (VSN) and an energy-error balance loss term. Relative to a non-spiking baseline of \(0.4\%\) relative L2 error, VS-GNO achieves \(0.71\%\) error with \(15\%\) average spiking and under \(3\%\) error with \(7\%\) spiking in its spectral-only form, and \(1.04\%\) error with \(24.5\%\) average spiking and \(2.15\%\) error at \(15\%\) spiking in the full architecture. These results demonstrate a controllable energy-accuracy tradeoff for edge deployment on irregular geometries.- 4 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 104-107
Open AccessNeuroscience Inspired Graph Operators Towards Edge-Deployable Virtual Sensing for Irregular Geometries
Real-time virtual sensing can enhance instrumentation in engineering systems but requires sparse-to-dense reconstruction on irregular grids under strict computational constraints. Neural operators have shown promise, but few address power consumption. Spiking networks offer a solution through neuromorphic deployment, but existing spiking neuron models degrade significantly in regression tasks. To address this, we present the Variable Spiking Graph Neural Operator (VS-GNO), combining spectral-spatial graph operator learning with the Variable Spiking Neuron (VSN) and an energy-error balance loss term. Relative to a non-spiking baseline of \(0.4\%\) relative L2 error, VS-GNO achieves \(0.71\%\) error with \(15\%\) average spiking and under \(3\%\) error with \(7\%\) spiking in its spectral-only form, and \(1.04\%\) error with \(24.5\%\) average spiking and \(2.15\%\) error at \(15\%\) spiking in the full architecture. These results demonstrate a controllable energy-accuracy tradeoff for edge deployment on irregular geometries. - Open Access
- 3 July 2026
- Published by Springer Nature in Communications Engineering
- Vol. 5 (1)
Network architecture follows coupling in multiphysics systems: single vs. multiple branches in DeepONet and S-DeepONet
Real-time prediction of complex physical systems requires surrogate models that learn from data while representing strong multiphysics coupling. Deep Operator Networks have shown success in single-physics problems, yet their effectiveness in capturing nonlinear interactions in coupled systems (such as thermo-mechanical or electro-thermal coupling) remains underexplored. Here we pose a practical question: should the architecture of a neural operator reflect the strength of physical coupling it aims to model? We compare single-branch and multi-branch designs, in both feedforward and sequential recurrent forms, across three representative systems: a reaction-diffusion problem with heterogeneous sources, a nonlinear thermo-electrical problem with temperature-dependent conductivity and Joule heating, and a viscoplastic thermo-mechanical model of steel solidification. Single-branch networks consistently outperform multi-branch variants in tightly coupled regimes by encouraging shared latent representations, whereas multi-branch designs remain favorable for decoupled or single-physics tasks. Once trained, these surrogates deliver full-field predictions up to 1.8 × 104 times faster than physics-based solvers.
- 3 July 2026
- Published by Springer Nature in Communications Engineering
- Vol. 5 (1)
Open AccessNetwork architecture follows coupling in multiphysics systems: single vs. multiple branches in DeepONet and S-DeepONet
Real-time prediction of complex physical systems requires surrogate models that learn from data while representing strong multiphysics coupling. Deep Operator Networks have shown success in single-physics problems, yet their effectiveness in capturing nonlinear interactions in coupled systems (such as thermo-mechanical or electro-thermal coupling) remains underexplored. Here we pose a practical question: should the architecture of a neural operator reflect the strength of physical coupling it aims to model? We compare single-branch and multi-branch designs, in both feedforward and sequential recurrent forms, across three representative systems: a reaction-diffusion problem with heterogeneous sources, a nonlinear thermo-electrical problem with temperature-dependent conductivity and Joule heating, and a viscoplastic thermo-mechanical model of steel solidification. Single-branch networks consistently outperform multi-branch variants in tightly coupled regimes by encouraging shared latent representations, whereas multi-branch designs remain favorable for decoupled or single-physics tasks. Once trained, these surrogates deliver full-field predictions up to 1.8 × 104 times faster than physics-based solvers.
- Open Access
- 12 July 2024
- Published by Springer Nature in npj Materials Degradation
- Vol. 8 (1) , 1-10
Sensor degradation in nuclear reactor pressure vessels: the overlooked factor in remaining useful life prediction
Sensor degradation poses a critical yet ‘often overlooked’ challenge in accurately predicting the remaining useful life (RUL) of nuclear reactor pressure vessels (RPVs), hindering safe and efficient plant operation. This paper introduces an approach to RUL estimation that explicitly addresses sensor degradation, a significant departure from conventional methods. We model neutron embrittlement, a dominant degradation process in RPV steel, as a Wiener process and leverage real-world surveillance capsule data for insightful parameterization. Maximum likelihood estimation is utilized to characterize the degradation dynamics in the model. A Kalman filter then seamlessly integrates the degradation model with sensor measurements, effectively compensating for degradation-induced errors and providing refined state estimates. These estimates power a robust RUL prediction framework. Our results expose the profound impact of sensor degradation on conventional RUL predictions. By directly confronting sensor degradation, our method yields substantially more accurate and reliable RUL estimates. This work marks a significant advancement in the field of materials degradation, offering a powerful tool to optimize nuclear power plant safety and longevity.
- 12 July 2024
- Published by Springer Nature in npj Materials Degradation
- Vol. 8 (1) , 1-10
Open AccessSensor degradation in nuclear reactor pressure vessels: the overlooked factor in remaining useful life prediction
Sensor degradation poses a critical yet ‘often overlooked’ challenge in accurately predicting the remaining useful life (RUL) of nuclear reactor pressure vessels (RPVs), hindering safe and efficient plant operation. This paper introduces an approach to RUL estimation that explicitly addresses sensor degradation, a significant departure from conventional methods. We model neutron embrittlement, a dominant degradation process in RPV steel, as a Wiener process and leverage real-world surveillance capsule data for insightful parameterization. Maximum likelihood estimation is utilized to characterize the degradation dynamics in the model. A Kalman filter then seamlessly integrates the degradation model with sensor measurements, effectively compensating for degradation-induced errors and providing refined state estimates. These estimates power a robust RUL prediction framework. Our results expose the profound impact of sensor degradation on conventional RUL predictions. By directly confronting sensor degradation, our method yields substantially more accurate and reliable RUL estimates. This work marks a significant advancement in the field of materials degradation, offering a powerful tool to optimize nuclear power plant safety and longevity.
- Top Cited PapersOpen Access
- 25 January 2024
- Published by Elsevier in Engineering Applications of Artificial Intelligence
- Vol. 131
Improved generalization with deep neural operators for engineering systems: Path towards digital twin
- 25 January 2024
- Published by Elsevier in Engineering Applications of Artificial Intelligence
- Vol. 131
Top Cited PapersOpen AccessImproved generalization with deep neural operators for engineering systems: Path towards digital twin