@article{alqahtani_gitushi_echekki_2024, title={A Data-Based Hybrid Chemistry Acceleration Framework for the Low-Temperature Oxidation of Complex Fuels}, volume={17}, ISSN={["1996-1073"]}, url={https://doi.org/10.3390/en17030734}, DOI={10.3390/en17030734}, abstractNote={The oxidation of complex hydrocarbons is a computationally expensive process involving detailed mechanisms with hundreds of chemical species and thousands of reactions. For low-temperature oxidation, an accurate account of the fuel-specific species is required to correctly describe the pyrolysis stage of oxidation. In this study, we develop a hybrid chemistry framework to model and accelerate the low-temperature oxidation of complex hydrocarbon fuels. The framework is based on a selection of representative species that capture the different stages of ignition, heat release, and final products. These species are selected using a two-step principal component analysis of the reaction rates of simulation data. Artificial neural networks (ANNs) are used to model the source terms of the representative species during the pyrolysis stage up to the transition time. This ANN-based model is coupled with C0–C4 foundational chemistry, which is used to model the remaining species up to the transition time and all species beyond the transition time. Coupled with the USC II mechanism as foundational chemistry, this framework is demonstrated using simple reactor homogeneous chemistry and perfectly stirred reactor (PSR) calculations for n-heptane oxidation over a range of composition and thermodynamic conditions. The hybrid chemistry framework accurately captures correct physical behavior and reproduces the results obtained using detailed chemistry at a fraction of the computational cost.}, number={3}, journal={ENERGIES}, author={Alqahtani, Sultan and Gitushi, Kevin M. and Echekki, Tarek}, year={2024}, month={Feb} }
 @article{taassob_kumar_gitushi_ranade_echekki_2024, title={A PINN-DeepONet framework for extracting turbulent combustion closure from multiscalar measurements}, volume={429}, ISSN={["1879-2138"]}, url={https://doi.org/10.1016/j.cma.2024.117163}, DOI={10.1016/j.cma.2024.117163}, journal={COMPUTER METHODS IN APPLIED MECHANICS AND ENGINEERING}, author={Taassob, Arsalan and Kumar, Anuj and Gitushi, Kevin M. and Ranade, Rishikesh and Echekki, Tarek}, year={2024}, month={Sep} }
 @article{gitushi_echekki_2024, title={Comparisons of Different Representative Species Selection Schemes for Reduced-Order Modeling and Chemistry Acceleration of Complex Hydrocarbon Fuels}, volume={17}, ISSN={["1996-1073"]}, url={https://doi.org/10.3390/en17112604}, DOI={10.3390/en17112604}, abstractNote={The simulation of engine combustion processes, such as autoignition, an important process in the co-optimization of fuel-engine design, can be computationally expensive due to the large number of thermo-chemical scalars needed to describe the full chemical system. Yet, the inherent correlations between the different chemical species during oxidation can significantly reduce the complexity of representing this system. One strategy is to select a subset of representative species that accurately captures the combustion process at a fraction of the computational cost of the full system. In this study, we compare the performance of four different techniques to select these species. They include the two-step principal component analysis (PCA) approach, directed relation graphs (DRGs), the global pathway selection (GPS) approach, and the manifold-informed species selection method. A parametric study of the representative species selection is carried out on data from the simulation of homogeneous and perfectly stirred reactors by investigating seven cumulative variances and 47 different cut-off percentages for the two-step PCA, and 65 and 51 thresholds for the DRGs and GPS, respectively. Results show that these selection methods capture key important species that can accurately describe the chemical system and track each stage of oxidation. The two-step PCA is sensitive to the cumulative variance, and DRGs and GPS are sensitive to the choice of target variables. By selecting key representative species and reducing the number of thermo-chemical scalars, these three methods can be used to develop computationally efficient hybrid chemistry schemes.}, number={11}, journal={ENERGIES}, author={Gitushi, Kevin M. and Echekki, Tarek}, year={2024}, month={Jun} }
 @article{ranade_gitushi_echekki_2023, title={Deep Learning of Joint Scalar PDFs in Turbulent Flames from Sparse Multiscalar Data}, volume={11}, ISSN={["1563-521X"]}, DOI={10.1080/00102202.2023.2283816}, abstractNote={We investigate the reconstruction of multi-dimensional joint scalar probability density functions starting with sparse multiscalar data in turbulent flames using deep operator networks (DeepONets). The formulation is based on 1) a dimensionality reduction to reduce the complexity of modeling the PDF using principal component analysis (PCA), and 2) the implementation of the deep operator neural network (or DeepONet) to construct a generalizable joint PDF of the principal components (PCs). We demonstrate the accuracy and generalizability of the DeepONet on the Sandia turbulent non-premixed flames D, E and F where the DeepONet is trained on data for flame E and is evaluated for the 3 flames. The DeepONet-based approach is able to construct multiple PDF shapes based on the PCs for all conditions. Combined with a recently proposed method for evaluating thermochemical scalar means conditioned on the PCs, the convolution of these means with the joint PDFs yields good predictions of thermochemical scalars’ unconditional means profiles in the flames.}, journal={COMBUSTION SCIENCE AND TECHNOLOGY}, author={Ranade, Rishikesh and Gitushi, Kevin M. and Echekki, Tarek}, year={2023}, month={Nov} }
 @article{gitushi_blaylock_hecht_2023, title={Simulations for Planning of Liquid Hydrogen Spill Test}, volume={16}, ISSN={["1996-1073"]}, DOI={10.3390/en16041580}, abstractNote={In order to better understand the complex pooling and vaporization of a liquid hydrogen spill, Sandia National Laboratories is conducting a highly instrumented, controlled experiment inside their Shock Tube Facility. Simulations were run before the experiment to help with the planning of experimental conditions, including sensor placement and cross wind velocity. This paper describes the modeling used in this planning process and its main conclusions. Sierra Suite’s Fuego, an in-house computational fluid dynamics code, was used to simulate a RANS model of a liquid hydrogen spill with five crosswind velocities: 0.45, 0.89, 1.34, 1.79, and 2.24 m/s. Two pool sizes were considered: a diameter of 0.85 m and a diameter of 1.7. A grid resolution study was completed on the smaller pool size with a 1.34 m/s crosswind. A comparison of the length and height of the plume of flammable hydrogen vaporizing from the pool shows that the plume becomes longer and remains closer to the ground with increasing wind speed. The plume reaches the top of the facility only in the 0.45 m/s case. From these results, we concluded that it will be best for the spacing and location of the concentration sensors to be reconfigured for each wind speed during the experiment.}, number={4}, journal={ENERGIES}, author={Gitushi, Kevin Mangala and Blaylock, Myra and Hecht, Ethan S. S.}, year={2023}, month={Feb} }
 @article{gitushi_blaylock_klebanoff_2022, title={Hydrogen gas dispersion studies for hydrogen fuel cell vessels II: Fuel cell room releases and the influence of ventilation}, volume={47}, ISSN={["1879-3487"]}, DOI={10.1016/j.ijhydene.2022.04.263}, abstractNote={Results are presented for computational fluid dynamics (CFD) modeling for varying hydrogen leaks within a hydrogen vessel's Fuel Cell Rack inside a Fuel Cell Room. In the limiting case of no room ventilation, modeling shows that the flammable region produced by the hydrogen leak is initially limited by self-induced entrainment and recirculation of air caused by the buoyant rising of hydrogen. Locally and at shorter times (minutes), this effect can be even more influential in limiting the size of the flammable envelope than Fuel Cell Room ventilation. Interestingly, the more diffuse detectable (but sub-flammable) region is not self-limited. This indicates the recirculation pattern required for the self-limiting effect requires a sufficient concentration of hydrogen to establish and differentiate the rising hydrogen mass from the surrounding air, thereby establishing the recirculation pattern that self-limits the flammable region at short times. Modeling results with the Fuel Cell Room ventilation activated shows that several seconds after a hydrogen leak is initiated, the flammable region reaches a steady state, with only minor fluctuations due to the air currents created by ventilation. The expected trends with ventilation rate are found: for a given leak size, a decreasing flammable envelope is found as ventilation is increased and for a given level of ventilation, an increasing hydrogen leak rate produces a larger flammable region. For the cases and ventilation rates examined, flammable H2/air mixtures greater than 4% clear the Fuel Cell Room within 1.5 s after the hydrogen leak is turned off. The CFD modeling results for the detectable level of hydrogen that would trigger an alarm showed that higher ventilation rates might have the unintended consequence of making a hydrogen leak harder to detect, depending on the location of the gas detector in the Fuel Cell Room For the hydrogen leak rates considered in this study, we find that a ventilation rate of 15 ACH provides timely hydrogen evacuation while allowing the leak to be detected by the ceiling-mounted hydrogen monitor (for most monitor locations).}, number={50}, journal={INTERNATIONAL JOURNAL OF HYDROGEN ENERGY}, author={Gitushi, K. M. and Blaylock, M. L. and Klebanoff, L. E.}, year={2022}, month={Jun}, pages={21492–21505} }
 @article{gitushi_ranade_echekki_2022, title={Investigation of deep learning methods for efficient high-fidelity simulations in turbulent combustion}, volume={236}, ISSN={["1556-2921"]}, url={https://doi.org/10.1016/j.combustflame.2021.111814}, DOI={10.1016/j.combustflame.2021.111814}, abstractNote={Turbulent combustion modeling often faces a trade-off between the so-called flamelet-like models and PDF-like models. Flamelet-like models, are characterized by a choice of a limited set of prescribed moments, which are transported to represent the manifold of the composition space and its statistics. PDF-like approaches are designed to directly evaluate the closure terms associated with the nonlinear chemical source terms in the energy and species equations. They generate data on the fly, which can be used to accelerate the simulation of PDF-like based models. Establishing key ingredients for implementing acceleration schemes for PDF-like methods by constructing flamelet-like models on the fly can potentially result in computational saving while maintaining the ability to resolve closure terms. These ingredients are investigated in this study. They include a data-based dimensional reduction of the composition space to a low-dimensional manifold using principal component analysis (PCA). The principal components (PCs) serve as moments, which characterize the manifold; and conditional means of the thermo-chemical scalars are evaluated in terms of these PCs. A second ingredient involves adapting a novel deep learning framework, DeepONet, to construct joint PCs' PDFs as alternative methods to presumed shapes common in flamelet-like approaches. We also investigate whether the rotation of the PCs into independent components (ICs) can improve their statistical independence. The combination of these ingredients is investigated using experimental data based on the Sydney turbulent nonpremixed flames with inhomogeneous inlets. The combination of constructed PDFs and conditional mean models are able to adequately reproduce unconditional statistics of thermo-chemical scalars, and establish acceptable statistical independence between the PCs, which simplify further the modeling of the joint PCs' PDFs.}, journal={COMBUSTION AND FLAME}, publisher={Elsevier BV}, author={Gitushi, Kevin M. and Ranade, Rishikesh and Echekki, Tarek}, year={2022}, month={Feb} }