Paul G. A. Cizmas
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Paul G. A. Cizmas has a total of 30 co-authors. Here are the top co-authors by number of shared publications:
- David N. Liliedahl
Department of Aerospace Engineering, Texas A&M University, College Station, TX 77843-3141
2 shared publications - Elizabeth Hope Krath
Sandia National Laboratories (SNL-CA), Livermore, CA (United States)
2 shared publications Forrest L. Carpenter
Department of Aerospace Engineering, Texas A&M University , 701 H.R. Bright Bldg., TAMU 3141, College Station, Texas 77843-3141, USA
2 shared publications- Forrest L. Carpenter
Department of Aerospace Engineering, Texas A&M University, College Station, TX 77843-3141
2 shared publications - Forrest L. Carpenter
Texas A&M University, College Station, Texas 77843-3141
2 shared publications Joaquin I. Gargoloff
Texas A&M University, College Station, Texas 77843-3141
2 shared publications
Latest Publications
- Open Access
- 9 July 2026
- Published by Elsevier in Proceedings of the Combustion Institute
- Vol. 42
Characterization of an ethylene–oxygen predetonator for high-speed combustors
- 9 July 2026
- Published by Elsevier in Proceedings of the Combustion Institute
- Vol. 42
Open AccessCharacterization of an ethylene–oxygen predetonator for high-speed combustors
- Open Access
- 1 June 2026
- Published by Elsevier in SSRN Electronic Journal
Trajectory-Optimized Time Reparameterization for Learning-Compatible Reduced-Order Modeling of Stiff Dynamical Systems
Stiff dynamical systems present a fundamental challenge for machine-learning reduced-order models, as explicit time integration becomes unstable while implicit schemes are computationally expensive during training. This limitation restricts the applicability of neural ordinary differential equation (NODE) models in engineering domains such as chemical kinetics, and control systems, where multiscale dynamics are prevalent. This work investigates time reparameterization as a stiffness-mitigation mechanism for NODE-based models and introduces a novel learning framework: trajectory-optimized time reparameterization. The proposed approach casts the time transformation as an optimization problem posed in arc-length coordinates, selecting a traversal-speed profile to directly penalize acceleration in the transformed time domain. By targeting the smoothness of the dynamics integrated during training, this method produces time maps and reparameterized trajectories that are better conditioned and easier to learn. The approach is evaluated on three representative stiff systems: a parameterized stiff linear system, the van der Pol oscillator, and the HIRES chemical kinetics model. Under identical training regimens, the method consistently improves model learnability and prediction accuracy compared with existing time reparameterization strategies. Quantitative results demonstrate reductions in training loss of one to two orders of magnitude, particularly in regimes with severe or monotone stiffness. These results highlight that the efficacy of stiffness mitigation is critically dependent on the regularity of the learned time transformation. The proposed optimization-based approach provides a practical and computationally efficient framework for applying NODE-based models to engineering applications involving stiff dynamics.
- 1 June 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessTrajectory-Optimized Time Reparameterization for Learning-Compatible Reduced-Order Modeling of Stiff Dynamical Systems
Stiff dynamical systems present a fundamental challenge for machine-learning reduced-order models, as explicit time integration becomes unstable while implicit schemes are computationally expensive during training. This limitation restricts the applicability of neural ordinary differential equation (NODE) models in engineering domains such as chemical kinetics, and control systems, where multiscale dynamics are prevalent. This work investigates time reparameterization as a stiffness-mitigation mechanism for NODE-based models and introduces a novel learning framework: trajectory-optimized time reparameterization. The proposed approach casts the time transformation as an optimization problem posed in arc-length coordinates, selecting a traversal-speed profile to directly penalize acceleration in the transformed time domain. By targeting the smoothness of the dynamics integrated during training, this method produces time maps and reparameterized trajectories that are better conditioned and easier to learn. The approach is evaluated on three representative stiff systems: a parameterized stiff linear system, the van der Pol oscillator, and the HIRES chemical kinetics model. Under identical training regimens, the method consistently improves model learnability and prediction accuracy compared with existing time reparameterization strategies. Quantitative results demonstrate reductions in training loss of one to two orders of magnitude, particularly in regimes with severe or monotone stiffness. These results highlight that the efficacy of stiffness mitigation is critically dependent on the regularity of the learned time transformation. The proposed optimization-based approach provides a practical and computationally efficient framework for applying NODE-based models to engineering applications involving stiff dynamics.
- 8 April 2026
- Published by Taylor & Francis in Tribology Transactions
- Vol. 69 (3) , 562-572
An Experimental and Numerical Investigation of the Rotordynamic Force Coefficients of a Gas Labyrinth Seal: Part 2—Numerical Simulation
Accurate prediction of rotordynamic coefficients and leakage rate of labyrinth seals is important for estimating the rotordynamic stability and performance of turbomachinery. Computational Fluid Dy...
- 8 April 2026
- Published by Taylor & Francis in Tribology Transactions
- Vol. 69 (3) , 562-572
An Experimental and Numerical Investigation of the Rotordynamic Force Coefficients of a Gas Labyrinth Seal: Part 2—Numerical Simulation
Accurate prediction of rotordynamic coefficients and leakage rate of labyrinth seals is important for estimating the rotordynamic stability and performance of turbomachinery. Computational Fluid Dy...
- 31 January 2026
- Published by Elsevier in Powder Technology
- Vol. 474
Reduced-order modeling of multiphase flow in fluidized beds using proper orthogonal decomposition and neural ODEs
- 31 January 2026
- Published by Elsevier in Powder Technology
- Vol. 474
Reduced-order modeling of multiphase flow in fluidized beds using proper orthogonal decomposition and neural ODEs
- 1 March 2024
- Published by AIP Publishing in Physics of Fluids
- Vol. 36 (3)
Prediction of unsteady flows in turbomachinery cascades using proper orthogonal decomposition
A reduced-order model based on the proper orthogonal decomposition (POD) method is used to assess whether it can predict the highly unsteady flow in turbomachinery cascades. Three cases are considered: a transonic fan rotor called the National Aeronautics and Space Administration Rotor 67, a cascade of oscillating compressor blades called the 10th Standard Configuration, and a cascade of oscillating turbine blades called the 11th Standard Configuration. The governing equations used in this POD model, called zeta-POD, are written using the specific volume instead of density. This formulation of the governing equations allows the precomputation of the coefficients of the system of ordinary differential equations prior the flow simulation. Consequently, a significant speedup is obtained compared to the full-order model and a traditional, density-based POD. In addition, the accuracy of the zeta-POD model is compared against the full-order model and the density-based POD.
- 1 March 2024
- Published by AIP Publishing in Physics of Fluids
- Vol. 36 (3)
Prediction of unsteady flows in turbomachinery cascades using proper orthogonal decomposition
A reduced-order model based on the proper orthogonal decomposition (POD) method is used to assess whether it can predict the highly unsteady flow in turbomachinery cascades. Three cases are considered: a transonic fan rotor called the National Aeronautics and Space Administration Rotor 67, a cascade of oscillating compressor blades called the 10th Standard Configuration, and a cascade of oscillating turbine blades called the 11th Standard Configuration. The governing equations used in this POD model, called zeta-POD, are written using the specific volume instead of density. This formulation of the governing equations allows the precomputation of the coefficients of the system of ordinary differential equations prior the flow simulation. Consequently, a significant speedup is obtained compared to the full-order model and a traditional, density-based POD. In addition, the accuracy of the zeta-POD model is compared against the full-order model and the density-based POD.
- Open Access
- 1 December 2023
- Published by AIP Publishing in Physics of Fluids
- Vol. 35 (12)
A multi-limiter method for simulating complex flows on large grids
This paper presents a novel method for aiding the convergence of complex flows simulations on large grids. In this work, large grids are considered those that are of the order of 100 × 106 cells or more. These large grids, which, for example, model the flow around a powered airplane, are typically difficult to generate and modify. If a high-fidelity solution is needed, as in the case of predicting sonic boom, the amount of numerical dissipation must be minimized. This reduced dissipation, however, may result in solution divergence, triggered by improperly resolved large gradients due to grid deficiency. Using a highly dissipative limiter for the entire grid could avoid solution divergence, but the resulting solution may be less accurate. To avoid the loss of solution accuracy due to too much numerical dissipation, we present an approach in which the flow solver uses two limiters: a low-dissipation limiter for the vast majority of the computational domain, and a high-dissipation limiter for the regions where the grid is not adequate for capturing the flow variation and would otherwise lead to the divergence of the solution. The paper demonstrates that (1) the overall solution is not modified by the local use of the high-dissipation limiter, and (2) for a case in which the use of low-dissipation limiter (Venkatakrishnan) resulted in solution divergence, a converged solution is obtained by using the low-dissipation limiter on the vast majority of the grid and a high-dissipation limiter (van Albada) on the region where the mesh was problematic. The effect of a local high-dissipation limiter on the overall solution is explored on a double-wedge canonical airfoil. The multi-limiter approach is then used to predict the near field pressure of the X-59 airplane.
- 1 December 2023
- Published by AIP Publishing in Physics of Fluids
- Vol. 35 (12)
Open AccessA multi-limiter method for simulating complex flows on large grids
This paper presents a novel method for aiding the convergence of complex flows simulations on large grids. In this work, large grids are considered those that are of the order of 100 × 106 cells or more. These large grids, which, for example, model the flow around a powered airplane, are typically difficult to generate and modify. If a high-fidelity solution is needed, as in the case of predicting sonic boom, the amount of numerical dissipation must be minimized. This reduced dissipation, however, may result in solution divergence, triggered by improperly resolved large gradients due to grid deficiency. Using a highly dissipative limiter for the entire grid could avoid solution divergence, but the resulting solution may be less accurate. To avoid the loss of solution accuracy due to too much numerical dissipation, we present an approach in which the flow solver uses two limiters: a low-dissipation limiter for the vast majority of the computational domain, and a high-dissipation limiter for the regions where the grid is not adequate for capturing the flow variation and would otherwise lead to the divergence of the solution. The paper demonstrates that (1) the overall solution is not modified by the local use of the high-dissipation limiter, and (2) for a case in which the use of low-dissipation limiter (Venkatakrishnan) resulted in solution divergence, a converged solution is obtained by using the low-dissipation limiter on the vast majority of the grid and a high-dissipation limiter (van Albada) on the region where the mesh was problematic. The effect of a local high-dissipation limiter on the overall solution is explored on a double-wedge canonical airfoil. The multi-limiter approach is then used to predict the near field pressure of the X-59 airplane.
- 1 May 2022
- Published by American Institute of Aeronautics and Astronautics (AIAA) in Journal of Aircraft
- Vol. 59 (3) , 595-611
UNS3D Near-Field Predictions for the Third AIAA Sonic Boom Prediction Workshop
This paper presents the results of turbulent flow simulations prepared for the Third AIAA Sonic Boom Prediction Workshop. Solutions were prepared for two geometries that exhibited shock–plume inter...
- 1 May 2022
- Published by American Institute of Aeronautics and Astronautics (AIAA) in Journal of Aircraft
- Vol. 59 (3) , 595-611
UNS3D Near-Field Predictions for the Third AIAA Sonic Boom Prediction Workshop
This paper presents the results of turbulent flow simulations prepared for the Third AIAA Sonic Boom Prediction Workshop. Solutions were prepared for two geometries that exhibited shock–plume inter...
- 2 December 2021
- Published by Cambridge University Press (CUP)
- p. 447-448
ICAO Standard Atmosphere
Get up to speed with this robust introduction to the aerothermodynamics principles underpinning jet propulsion, and learn how to apply these principles to jet engine components. Suitable for undergraduate students in aerospace and mechanical engineering, and for professional engineers working in jet propulsion, this textbook includes consistent emphasis on fundamental phenomena and key governing equations, providing students with a solid theoretical grounding on which to build practical understanding; clear derivations from first principles, enabling students to follow the reasoning behind key assumptions and decisions, and successfully apply these approaches to new problems; practical examples grounded in real-world jet propulsion scenarios illustrate new concepts throughout the book, giving students an early introduction to jet and rocket engine considerations; and online materials for course instructors, including solutions, figures, and software resources, to enhance student teaching.
- 2 December 2021
- Published by Cambridge University Press (CUP)
- p. 447-448
ICAO Standard Atmosphere
Get up to speed with this robust introduction to the aerothermodynamics principles underpinning jet propulsion, and learn how to apply these principles to jet engine components. Suitable for undergraduate students in aerospace and mechanical engineering, and for professional engineers working in jet propulsion, this textbook includes consistent emphasis on fundamental phenomena and key governing equations, providing students with a solid theoretical grounding on which to build practical understanding; clear derivations from first principles, enabling students to follow the reasoning behind key assumptions and decisions, and successfully apply these approaches to new problems; practical examples grounded in real-world jet propulsion scenarios illustrate new concepts throughout the book, giving students an early introduction to jet and rocket engine considerations; and online materials for course instructors, including solutions, figures, and software resources, to enhance student teaching.
- Open Access
- 9 April 2021
- Published by MDPI AG in Applied Sciences
- Vol. 11 (8) , 3377
Several Cases for the Validation of Turbulence Models Implementation
This paper presents several test cases that were used to validate the implementation of two turbulence models in the UNS3D code, an in-house code. The two turbulence models used were the Shear Stress Transport model and the Spalart–Allmaras model. These turbulence models were explored using the numerical results generated by three computational fluid dynamics codes: NASA’s FUN3D and CFL3D, and UNS3D. Four cases were considered: a flat plate case, an airfoil near-wake, a backward-facing step, and a turbine cascade known as the Eleventh Standard Configuration. The numerical results were compared among themselves and against experimental data.
- 9 April 2021
- Published by MDPI AG in Applied Sciences
- Vol. 11 (8) , 3377
Open AccessSeveral Cases for the Validation of Turbulence Models Implementation
This paper presents several test cases that were used to validate the implementation of two turbulence models in the UNS3D code, an in-house code. The two turbulence models used were the Shear Stress Transport model and the Spalart–Allmaras model. These turbulence models were explored using the numerical results generated by three computational fluid dynamics codes: NASA’s FUN3D and CFL3D, and UNS3D. Four cases were considered: a flat plate case, an airfoil near-wake, a backward-facing step, and a turbine cascade known as the Eleventh Standard Configuration. The numerical results were compared among themselves and against experimental data.
- 1 February 2021
- Published by Elsevier in Journal of Computational Physics
- Vol. 426
An efficient proper orthogonal decomposition based reduced-order model for compressible flows
- 1 February 2021
- Published by Elsevier in Journal of Computational Physics
- Vol. 426
An efficient proper orthogonal decomposition based reduced-order model for compressible flows