About

Pavel V. Tsvetkov is a Professor of Nuclear Engineering at Texas A&M University and serves as the Director of the Graduate Program in Nuclear Engineering. He holds a Ph.D. in Nuclear Engineering from Texas A&M University, obtained in 2002, and an M.S. in Theoretical & Experimental Reactor Physics from Moscow State Engineering Physics Institute, earned in 1995. His research interests encompass system analysis and optimization methods, complex engineered systems, system design, and symbiotic nuclear energy systems. He is particularly focused on waste minimization, sustainability, high temperature gas reactors (HTGRs), cogeneration systems, and direct nuclear energy conversion systems. Recognized for his contributions to the field, he received the George Armistead, Jr ’23 Faculty Excellence Teaching Award.

Research topics

• Computer Science
• Engineering
• Artificial Intelligence
• Nuclear engineering
• Reliability engineering
• Physics
• Composite material
• Chemistry
• Embedded system
• Materials science
• Thermodynamics
• Environmental science
• Mechanics
• Mathematics
• Systems engineering

Selected publications

• A New NCSP Nuclear Criticality Safety Training and Pipeline Program [Slides]

  2025-06-15

  report
  PublisherDOI

• Deterministic and probabilistic Deep Learning in predicting reactor physics of subcritical system driven by a pulsed source

  Progress in Nuclear Energy · 2025-07-28

  articleOpen accessSenior author

  Widely reported are the non-ideal response of standard techniques in reactivity measurement when the Subcritical Assembly (SCA) is far from critical. This emanates from the loose applicability of fundamental mode assumption in Point Reactor Kinetics from which the analytical formulae relating detector response to reactivity were derived. This work evaluated the potential of Deep Learning (DL) in overcoming these biasing effects particularly in an SCA driven by a Pulsed Neutron Source (PNS). Deterministic DL models processing core map and detector temporal response were trained using data from neutronics calculations, and subsequently compared with Area-ratio, and Slope-fit methods in a simulated PNS experiment. Results show the robustness of DL against spatial effect that severely affected Area-ratio method leading to severe underestimation, a non-conservative scenario in criticality safety. As illustration, k e f f D L = 0.94158 is approximately equal to k e f f M C N P = 0.94093 ± 0.00034 ; meanwhile k e f f A r e a − r a t i o = 0.435 ± 0.00900 ≪ k e f f M C N P . Furthermore, DL did not indicate increasing bias as system becomes deeply subcritical unlike Slope-fit method. These advantages also extended to probabilistic variants based on Bayesian Neural Networks, which were found to be well-calibrated due to matching predicted and observed confidence levels. These findings suggest the strong potential of deploying DL in an operational context, helping assure safety margins in SCAs and safe approach to criticality in research reactors. • Reactor physics prediction in pulsed-source subcritical assembly by Deep Learning. • Bayesian Neural Network applied in a simulated pulsed-source experiment. • Robustness to reactivity bias of Deep Learning compared to Area-ratio and Slope-fit. • Trade-offs between Area-ratio and Slope-fit methods were explored. • Agreement of predicted and observed confidence levels of Bayesian Neural Network.

  PublisherDOI

• Dynamics modeling of molten salt reactor with reduced and expanded representations of delayed neutron precursors

  Annals of Nuclear Energy · 2025-04-12 · 2 citations

  article
  PublisherDOI

• Adjoint Monte Carlo Method for the Multiplicity Moment Distributions of Subcritical Multiplying Systems [Poster]

  2025-04-24

  reportOpen access
  PublisherOA PDFDOI

• Advancements and challenges of machine learning and deep learning in autonomous control of nuclear reactors

  Annals of Nuclear Energy · 2025-06-13 · 2 citations

  articleOpen accessSenior author

  This review paper explores recent advancements in the application of machine learning (ML) and deep learning technologies for autonomous control in nuclear reactors. It covers intelligent diagnosis systems using ML, deep learning algorithms, and hybrid approaches for reactor condition assessment. In the area of intelligent control, traditional methods such as fuzzy control, proportional-integral-derivative (PID) control, and Model Predictive Control (MPC), coupled with neural networks, are discussed, as well as deep reinforcement learning (DRL) for controlling a nuclear reactor. Key challenges are identified, including system integration, cybersecurity, and regulatory adaptation. The review highlights the need for future research on integrating intelligent diagnosis and control systems in real-world reactors, particularly advanced and small modular reactors. It also stresses the importance of considering cybersecurity during the design phase of autonomous control systems and updates of regulatory frameworks to accommodate AI-driven technologies in nuclear power plant operations.

  PublisherDOI

• Potential of deep learning methods to enhance satellite-based monitoring of nuclear power plants focusing on remote operation evaluations

  Annals of Nuclear Energy · 2025-03-14 · 2 citations

  article
  PublisherDOI

• Exploratory AI Applications to Enhance Fiberoptics Performance in Nuclear Monitoring

  2025-11-01

  article1st authorCorresponding

  Fiberoptic-based sensing is a promising approach for monitoring extreme, hard-to-access nuclear environments like reactor internals and spent fuel storage facilities, where structural integrity, operational performance, and nuclear security are vital. Technologies such as fiber Bragg gratings (FBGs) and distributed sensing of temperature and radiation enable real-time monitoring to meet performance and security needs. However, challenges including signal noise, limited resolution, degradation in harsh conditions, calibration dependency, and complex data processing limit applicability for emerging nuclear systems. Agent-based AI data analysis offers an innovative solution, particularly for real-time sensor calibration in rapidly varying environments. Collaborative autonomous agents process distributed fiberoptic data, perform real-time signal denoising, and achieve high-precision anomaly detection critical for integrity and security. AI and agent-based AI enable adaptive deployment and perofrmance, real-time calibration of fiberoptic sensors, ensuring accuracy in dynamic conditions, unlike traditional static methods. Reinforcement learning methods offer optimization options for sensor performance metrics, while swarm intelligence leads to apporaches enhancing data fusion across multimodal networks. This study explores the state of the art in fiberoptic applications for nuclear monitoring and identifies performance bottlenecks and ways to address them. It proposes an agent-based AI framework to enhance signal processing and measurement accuracy in multimodal domains.

  PublisherDOI

• Differentiable hybrid neural network approach for enhancing reactor dynamics simulations

  Annals of Nuclear Energy · 2025-08-06 · 1 citations

  article
  PublisherDOI

• Deep learning-based enhancement of integrated process control performance in nuclear reactor operations

  Annals of Nuclear Energy · 2025-08-20 · 1 citations

  articleSenior author
  PublisherDOI

• A Digital Twin-Based Simulator for Small Modular and Microreactors

  2024-12-15 · 1 citations

  article

  This paper reports preliminary work done on a mechanistic-based model digital twin (DT) for Gen IV reactors. The case study is a conceptual 4.5 MWth Small Modular Lead-cooled Fast (LFR) Research Reactor whose design incorporated aspects from all the three existing families of Gen IV LFRs. The back end for the DT exploited a modular approach consisting of the Neutronics and Thermohydraulic Coupling of the reactor core. This modular approach gives room for subsequent modification and/or addition of new blocks as the design concept matures without perturbing the entire system. After benchmarking simulation results with data from literature, the system's GUI demonstrated the capability to perform and visualize common operational transients either as a stand-alone simulator or in real-time using the MQTT broker. Insights derived from this virtual environment could contribute towards the ongoing refinement of LFR technology thus accelerating development through design testing, visualization, and optimization.

  PublisherDOI

Frequent coauthors

• Jason A. Hearne

  Texas A&M University

  19 shared

• Ayodeji Babatunde Alajo

  Missouri University of Science and Technology

  15 shared

• David Ames

  Sandia National Laboratories California

  15 shared

• Mario Mendoza

  Texas A&M University

  13 shared

• Ronald Daryll E. Gatchalian

  Texas A&M University

  12 shared

• Tom G. Lewis

  9 shared

• Alan Waltar

  United States Department of Energy

  8 shared

• Megan L. Pritchard

  6 shared

Awards & honors

• George Armistead, Jr ’23 Faculty Excellence Teaching Award