arXiv subjects

Pinaki Pal

Publications and source records attributed to Pinaki Pal.

At least 37 records · Page 2Linked to original sources

arXiv2024

Computational Study on the Impact of Gasoline-Ethanol Blending on Autoignition and Soot/NOx Emissions under Gasoline Compression Ignition Conditions

Computational fluid dynamics (CFD) simulations of a single-cylinder gasoline compression ignition engine are performed to investigate the impact of gasoline-ethanol blending on autoignition, nitrogen oxide (NOx), and soot emissions under low-load conditions. A four-component toluene primary reference fuel (TPRF) + ethanol (ETPRF) surrogate (with 10% ethanol by volume; E10) is employed to represent the test gasoline (RD5-87). A 3D engine CFD model employing finite-rate chemistry with a skeletal kinetic mechanism, adaptive mesh refinement (AMR), and hybrid method of moments (HMOM) is adopted to capture in-cylinder combustion and soot/NOx emissions. The engine CFD model is validated against experimental data for three gasoline-ethanol blends: E10, E30 and E100, with varying ethanol content by volume. Model validation is carried out for multiple start-of-injection (SOI) timings (-21, -27, -36, and -45 crank angle degrees after top-dead-center (aTDC)) with respect to in-cylinder pressure, heat release rate, combustion phasing, NOx and soot emissions. For late injection timings (-21 and -27oaTDC), E30 yields higher soot than E10; while the trend reverses for early injection cases (-36 and -45oaTDC). E100 yields the lowest amount of soot among all fuels irrespective of SOI timing. Further, E10 shows a non-monotonic trend in soot emissions with SOI timing: SOI-36>SOI-45>SOI-21>SOI-27, while soot emissions from E30 exhibit monotonic decrease with advancing SOI timing. NOx emissions from various fuels follow a trend of E10>E30>E100. NOx emissions increase as SOI timing is advanced for all fuels, with an anomaly for E10 and E100 where NOx decreases when SOI is advanced beyond -36oaTDC. Detailed analysis of the numerical results is performed to investigate the emission trends and elucidate the impact of chemical composition and physical properties on autoignition and emissions characteristics.

math.NA↗

arXiv2024

A Posteriori Evaluation of a Physics-Constrained Neural Ordinary Differential Equations Approach Coupled with CFD Solver for Modeling Stiff Chemical Kinetics

The high computational cost associated with solving for detailed chemistry poses a significant challenge for predictive computational fluid dynamics (CFD) simulations of turbulent reacting flows. These models often require solving a system of coupled stiff ordinary differential equations (ODEs). While deep learning techniques have been experimented with to develop faster surrogate models, they often fail to integrate reliably with CFD solvers. This instability arises because deep learning methods optimize for training error without ensuring compatibility with ODE solvers, leading to accumulation of errors over time. Recently, NeuralODE-based techniques have offered a promising solution by effectively modeling chemical kinetics. In this study, we extend the NeuralODE framework for stiff chemical kinetics by incorporating mass conservation constraints directly into the loss function during training. This ensures that the total mass and the elemental mass are conserved, a critical requirement for reliable downstream integration with CFD solvers. Proof-of-concept studies are performed with physics-constrained neuralODE (PC-NODE) approach for homogeneous autoignition of hydrogen-air mixture over a range of composition and thermodynamic conditions. Our results demonstrate that this enhancement not only improves the physical consistency with respect to mass conservation criteria but also ensures better robustness. Lastly, a posteriori studies are performed wherein the trained PC-NODE model is coupled with a 3D CFD solver for computing the chemical source terms. PC-NODE is shown to be more accurate relative to the purely data-driven neuralODE approach. Moreover, PC-NODE also exhibits robustness and generalizability to unseen initial conditions from within (interpolative capability) as well as outside (extrapolative capability) the training regime.

physics.comp-ph↗

arXiv2023

Forward Gradients for Data-Driven CFD Wall Modeling

Computational Fluid Dynamics (CFD) is used in the design and optimization of gas turbines and many other industrial/ scientific applications. However, the practical use is often limited by the high computational cost, and the accurate resolution of near-wall flow is a significant contributor to this cost. Machine learning (ML) and other data-driven methods can complement existing wall models. Nevertheless, training these models is bottlenecked by the large computational effort and memory footprint demanded by back-propagation. Recent work has presented alternatives for computing gradients of neural networks where a separate forward and backward sweep is not needed and storage of intermediate results between sweeps is not required because an unbiased estimator for the gradient is computed in a single forward sweep. In this paper, we discuss the application of this approach for training a subgrid wall model that could potentially be used as a surrogate in wall-bounded flow CFD simulations to reduce the computational overhead while preserving predictive accuracy.

physics.flu-dyn↗

arXiv2023

Effect of horizontal magnetic field on Küppers-Lortz instability

We investigate the effect of an external horizontal magnetic field on the Küppers-Lortz instability (KLI) in rotating Rayleigh-Bénard convection of Boussinesq fluids using weakly nonlinear theory along with linear theory. By KLI, we mean the instability where the two-dimensional roll solutions of the system occurring at the onset of convection becomes unstable against the perturbations by rolls oriented at different angle with the previous one as the rotation rate exceeds a critical value. The governing parameters, namely, the Prandtl number ($\mathrm{Pr}$), Taylor number ($\mathrm{Ta}$) and Chandrasekhar number ($\mathrm{Q}$) are varied in the ranges $0.8 \leq \mathrm{Pr} < \infty$, $0 < \mathrm{Ta} \leq 10^4$ and $0 \leq \mathrm{Q} \leq 10^4$ respectively by considering the vanishingly small magnetic Prandtl number limit. In the $\mathrm{Pr}\rightarrow \infty$ limit, magnetic field is found to inhibit the KLI by enhancing the critical Taylor number ($\mathrm{Ta}_c$) for its onset. On the other hand, for finite Prandtl number fluids, KLI is favored for lower $\mathrm{Q}$, and it is inhibited for higher $\mathrm{Q}$. Interestingly, in the finite Prandtl number range both KLI and small angle instability are manifested depending on the Prandtl number. No small angle instability is observed for $\mathrm{Pr} \geq 50$ and the rotation induced KLI is inhibited predominantly by the magnetic field. While, for $\mathrm{Pr} < 50$, along with the Küppers-Lortz instability, small angle instability is also observed. However, in this case, KLI is favored for lower $\mathrm{Q}$, while it is inhibited for higher $\mathrm{Q}$.

physics.flu-dyn↗

arXiv2023

Impact of phase lag on synchronization in frustrated Kuramoto model with higher-order interactions

The study of first order transition (explosive synchronization) in an ensemble (network) of coupled oscillators has been the topic of paramount interest among the researchers for more than onedecade. Several frameworks have been proposed to induce explosive synchronization in a network and it has been reported that phase frustration in a network usually suppresses first order transition in the presence of pairwise interactions among the oscillators. However, on the contrary, by considering networks of phase frustrated coupled oscillators in the presence of higher order interactions (upto 2-simplexes) we show here under certain conditions, phase frustration can promote explosive synchronization in a network. A reduced order model of the network in the thermodynamic limit is derived using the Ott-Antonsen ansatz to explain this surprising result. Analytical treatment of the reduced order model including bifurcation analysis explains the apparent counter intuitive result quite clearly.

nlin.AO↗

arXiv2023

Perfect synchronization in complex networks with higher order interactions

We propose a framework for achieving perfect synchronization in complex networks of Sakaguchi-Kuramoto oscillators in presence of higher order interactions (simplicial complexes) at a targeted point in the parameter space. It is achieved by using an analytically derived frequency set from the governing equations. The frequency set not only provides stable perfect synchronization in the network at a desired point, but also proves to be very effective in achieving high level of synchronization around it compared to the choice of any other frequency sets (Uniform, Normal etc.). The proposed framework has been verified using scale-free, random and small world networks. In all the cases, stable perfect synchronization is achieved at a targeted point for wide ranges of the coupling parameters and phase-frustration. Both first and second order transitions to synchronizations are observed in the system depending on the type of the network and phase frustration. The stability of perfect synchronization state is checked using the low dimensional reduction approach. The robustness of the perfect synchronization state obtained in the system using the derived frequency set is checked by introducing a Gaussian noise around it.

physics.soc-ph↗

arXiv2022

Rate-induced tipping can trigger plankton blooms

Plankton blooms are complex nonlinear phenomena whose occurrence can be described by the two-timescale (fast-slow) phytoplankton-zooplankton model intrpduced by Truscott and Brindley 1994. In their work, they observed that a sufficiently fast rise of the water temperature causes a critical transition from a low phytoplankton concentration to a single outburst: a so-called plankton bloom. However, the dynamical mechanism responsible for the observed transition has not been identified to the present day. Using techniques from geometric singular perturbation theory, we uncover the formerly overlooked rate-sensitive quasithreshold which is given by special trajectories called canards. The transition from low to high concentrations occurs when this rate-sensitive quasithreshold moves past the current state of the plankton system at some narrow critical range of warming rates. In this way, we identify rate-induced tipping as the underlying dynamical mechanism. Our findings explain the previously reported transitions to a single plankton bloom, and allow us to predict a new type of transition to a sequence of blooms for higher rates of warming. This could provide a possible mechanism of the observed increased frequency of harmful algal blooms.

nlin.PS↗

arXiv2021

ChemNODE: A Neural Ordinary Differential Equations Approach for Chemical Kinetics Solvers

Solving for detailed chemical kinetics remains one of the major bottlenecks for computational fluid dynamics simulations of reacting flows using a finite-rate-chemistry approach. This has motivated the use of fully connected artificial neural networks to predict stiff chemical source terms as functions of the thermochemical state of the combustion system. However, due to the nonlinearities and multi-scale nature of combustion, the predicted solution often diverges from the true solution when these deep learning models are coupled with a computational fluid dynamics solver. This is because these approaches minimize the error during training without guaranteeing successful integration with ordinary differential equation solvers. In the present work, a novel neural ordinary differential equations approach to modeling chemical kinetics, termed as ChemNODE, is developed. In this deep learning framework, the chemical source terms predicted by the neural networks are integrated during training, and by computing the required derivatives, the neural network weights are adjusted accordingly to minimize the difference between the predicted and ground-truth solution. A proof-of-concept study is performed with ChemNODE for homogeneous autoignition of hydrogen-air mixture over a range of composition and thermodynamic conditions. It is shown that ChemNODE accurately captures the correct physical behavior and reproduces the results obtained using the full chemical kinetic mechanism at a fraction of the computational cost.

cs.CE↗

arXiv2021

Identifying symmetries and predicting cluster synchronization in complex networks

Symmetries in a network connectivity regulate how the graph's functioning organizes into clustered states. Classical methods for tracing the symmetry group of a network require very high computational costs, and therefore they are of hard, or even impossible, execution for large sized graphs. We here unveil that there is a direct connection between the elements of the eigen-vector centrality and the clusters of a network. This gives a fresh framework for cluster analysis in undirected and connected graphs, whose computational cost is linear in $N$. We show that the cluster identification is in perfect agreement with symmetry based analyses, and it allows predicting the sequence of synchronized clusters which form before the eventual occurrence of global synchronization.

nlin.AO↗

arXiv2021

Amplification of explosive width in complex networks

We present an adaptive coupling strategy to induce hysteresis/explosive synchronization (ES) in complex networks of phase oscillators Sakaguchi-Kuramoto model). The coupling strategy ensures explosive synchronization with significant explosive width enhancement. Results show the robustness of the strategy and the strategy can diminish (by inducing enhanced hysteresis loop) the contrarian impact of phase frustration in the network, irrespective of network structure or frequency distributions. Additionally, we design a set of frequency for the oscillators which eventually ensure complete in-phase synchronization behavior among these oscillators (with enhanced explosive width) in the case of adaptive-coupling scheme. Based on a mean-field analysis, we develop a semi-analytical formalism, which can accurately predict the backward transition of synchronization order parameter.

nlin.AO↗

arXiv2021

Application of an automated machine learning-genetic algorithm (AutoML-GA) coupled with computational fluid dynamics simulations for rapid engine design optimization

In recent years, the use of machine learning-based surrogate models for computational fluid dynamics (CFD) simulations has emerged as a promising technique for reducing the computational cost associated with engine design optimization. However, such methods still suffer from drawbacks. One main disadvantage of is that the default machine learning (ML) hyperparameters are often severely suboptimal for a given problem. This has often been addressed by manually trying out different hyperparameter settings, but this solution is ineffective in a high-dimensional hyperparameter space. Besides this problem, the amount of data needed for training is also not known a priori. In response to these issues that need to be addressed, the present work describes and validates an automated active learning approach, AutoML-GA, for surrogate-based optimization of internal combustion engines. In this approach, a Bayesian optimization technique is used to find the best machine learning hyperparameters based on an initial dataset obtained from a small number of CFD simulations. Subsequently, a genetic algorithm is employed to locate the design optimum on the ML surrogate surface. In the vicinity of the design optimum, the solution is refined by repeatedly running CFD simulations at the projected optimum and adding the newly obtained data to the training dataset. It is demonstrated that AutoML-GA leads to a better optimum with a lower number of CFD simulations, compared to the use of default hyperparameters. The proposed framework offers the advantage of being a more hands-off approach that can be readily utilized by researchers and engineers in industry who do not have extensive machine learning expertise.

cs.LG↗

arXiv2021

A novel machine learning-based optimization algorithm (ActivO) for accelerating simulation-driven engine design

A novel design optimization approach (ActivO) that employs an ensemble of machine learning algorithms is presented. The proposed approach is a surrogate-based scheme, where the predictions of a weak leaner and a strong learner are utilized within an active learning loop. The weak learner is used to identify promising regions within the design space to explore, while the strong learner is used to determine the exact location of the optimum within promising regions. For each design iteration, exploration is done by randomly selecting evaluation points within regions where the weak learner-predicted fitness is high. The global optimum obtained by using the strong learner as a surrogate is also evaluated to enable rapid convergence once the most promising region has been identified. First, the performance of ActivO was compared against five other optimizers on a cosine mixture function with 25 local optima and one global optimum. In the second problem, the objective was to minimize indicated specific fuel consumption of a compression-ignition internal combustion (IC) engine while adhering to desired constraints associated with in-cylinder pressure and emissions. Here, the efficacy of the proposed approach is compared to that of a genetic algorithm, which is widely used within the internal combustion engine community for engine optimization, showing that ActivO reduces the number of function evaluations needed to reach the global optimum, and thereby time-to-design by 80%. Furthermore, the optimization of engine design parameters leads to savings of around 1.9% in energy consumption, while maintaining operability and acceptable pollutant emissions.

cs.LG↗

arXiv2020

Overstable rotating convection in presence of vertical magnetic field

We present the results of our investigation on nonlinear overstable rotating magnetoconvection (RMC) in presence of vertical external magnetic field. We focus on the dynamics appearing near the onset of convection by varying the system control parameters, namely, the Taylor number ($\mathrm{Ta}$), the Chandrasekhar number ($\mathrm{Q}$) and the Prandtl number ($\mathrm{Pr}$) in the ranges $750\leq\mathrm{Ta}\leq10^6$, $0 < \mathrm{Q} \leq 10^3$ and $0 < \mathrm{Pr} \leq 0.5$. Three dimensional (3D) direct numerical simulations (DNS) of the governing equations and low-dimensional modeling of the system are performed for this purpose. Extensive DNS in the specified parameter space shows two qualitatively different onsets depending on $\mathrm{Ta}$, $\mathrm{Q}$ and $\mathrm{Pr}$. In the first one, bistability appears at the onset, where both subcritical and supercritical convection coexist, while only supercritical convection is observed in the second one. Analysis of the low-dimensional model reveals that a supercritical Hopf bifurcation is responsible for the supercritical onset and a subcritical pitchfork bifurcation is responsible for the subcritical onset. It is also observed that appearance of subcritical convection at the onset has strong dependence on all three control parameters $\mathrm{Ta}$, $\mathrm{Q}$ and $\mathrm{Pr}$. The scenario of subcritical convection is found to disappear as $\mathrm{Pr}$ is increased for fixed $\mathrm{Ta}$ and $\mathrm{Q}$. However, most striking findings of the investigation is that the increment in $\mathrm{Ta}$ for fixed $\mathrm{Q}$ and $\mathrm{Pr}$ opposes the subcritical convection, whereas the increment in $\mathrm{Q}$ for fixed $\mathrm{Ta}$ and $\mathrm{Pr}$ favors it. This is in sharp contrast with the earlier results reported in RMC.

physics.flu-dyn↗

arXiv2020

Transitions in overstable rotating magnetoconvection

The classical Rayleigh-Bénard convection (RBC) system is known to exhibit either subcritical or supercritical transition to convection in the presence or absence of rotation and/or magnetic field. However, the simultaneous exhibition of subcritical and supercritical branches of convection in plane layer RBC depending on the initial conditions, has not been reported so far. Here, we report the phenomenon of simultaneous occurrence of subcritical and supercritical branches of convection in overstable RBC of electrically conducting low Prandtl number fluids (liquid metals) in the presence of an external uniform horizontal magnetic field and rotation about the vertical axis. Extensive three dimensional (3D) direct numerical simulations (DNS) and low dimensional modeling of the system, performed in the ranges $750 \leq \mathrm{Ta} \leq 3000$ and $0 < \mathrm{Q} \leq 1000$ of the Taylor number ($\mathrm{Ta}$, strength of the Coriolis force) and the Chandrasekhar number ($\mathrm{Q}$, strength of the Lorenz force) respectively, establish the phenomenon convincingly. Detailed bifurcation analysis of a simple three dimensional model derived from the DNS data reveals that a supercritical Hopf bifurcation and a subcritical pitchfork bifurcation of the conduction state are responsible for this. The effect of Prandtl number on these transitions is also explored in detail.

physics.flu-dyn↗

arXiv2019

Optimizing synchronization in multiplex networks of phase oscillators

We present an analytical scheme to achieve optimal synchronization in multiplex networks of frustrated and non-frustrated phase oscillators. We derive a multiplex synchrony alignment function (MSAF) for that purpose, the expression of which consists of structural as well as dynamical information of the layers of the multiplex network. Analyzing the MSAF, a set of frequencies (optimal frequencies) is determined to achieve optimal synchronization in the network. Further, using the scheme, we show that perfect synchronization can be achieved in a layer of the multiplex network for given coupling strength and phase frustration parameters. The analytical scheme presented here has been tested for heterogeneous multiplex networks of frustrated and non-frustrated Kuramoto dynamics.

nlin.AO↗

arXiv2019

Efficient bifurcation and parameterization of multi-dimensional combustion manifolds using deep mixture of experts: an a priori study

This work describes and validates an approach for autonomously bifurcating turbulent combustion manifolds to divide regression tasks amongst specialized artificial neural networks (ANNs). This approach relies on the mixture of experts (MoE) framework, where each neural network is trained to be specialized in a given portion of the input space. The assignment of different input regions to the experts is determined by a gating network, which is a neural network classifier. In some previous studies, it has been demonstrated that bifurcation of a complex combustion manifold and fitting different ANNs for each part leads to better fits or faster inference speeds. However, the manner of bifurcation in these studies was based on heuristic approaches or clustering techniques. In contrast, the proposed technique enables automatic bifurcation using non-linear planes in high-dimensional turbulent combustion manifolds that are often associated with complex behavior due to different dominating physics in various zones. The proposed concept is validated using 4-dimensional (4D) and 5D flamelet tables, showing that the errors obtained with a given network size, or conversely the network size required to achieve a given accuracy, is considerably reduced. The effect of the number of experts on inference speed is also investigated, showing that by increasing the number of experts from 1 to 8, the inference time can be approximately reduced by a factor of two. Moreover, it is shown that the MoE approach divides the input manifold in a physically intuitive manner, suggesting that the MoE framework can elucidate high-dimensional datasets in a physically meaningful way.

physics.flu-dyn↗

arXiv2019

Emergent dynamics in delayed attractive-repulsively coupled networks

We investigate different emergent dynamics namely oscillation quenching and revival of oscillation in a global network of identical oscillators coupled with diffusive (positive) delay coupling as it is perturbed by symmetry breaking localized repulsive delayed interaction. Starting from the oscillatory states (OS) we systematically identify three types of transition phenomena in the parameter space: (1) The system may reach inhomogeneous steady states (IHSS) from the homogeneous steady state (HSS) sometimes called as the transition from amplitude death (AD) to oscillation death (OD) state i.e. OS-AD-OD scenario, (2) Revival of oscillation (OS) from the AD state (OS-AD-OS) and (3) Emergence of OD state from oscillatory state (OS) without passing through AD i.e. OS-OD. The dynamics of each node in the network is assumed to be governed either by identical limit cycle Stuart-Landau system or by chaotic Rossler system. Based on clustering behavior observed in oscillatory network we derive a reduced low-dimensional model of the large network. Using the reduced model, we investigate the effect of time delay on these transitions and demarcate OS, AD and OD regimes in the parameter space. We also explore and characterize the bifurcation transitions present in both systems. The generic behavior of the low dimensional model and full network are found to match satisfactorily.

nlin.AO↗

arXiv2018

Transitions near the onset of low Prandtl-number rotating magnetoconvection

We investigate the onset of convective instability and subsequent transitions near it in Rayleigh-Bénard convection (RBC) of electrically conducting low Prandtl-number (Pr) fluids in the simultaneous presence of rotation about the vertical axis and external uniform horizontal magnetic field with free-slip boundary conditions. Both three dimensional direct numerical simulations (DNS) and low dimensional modeling of the system have been performed for the investigation. DNS results show a plethora of patterns near the onset of convection for weak magnetic field. Most interestingly, the Nusselt number (Nu) computed from the DNS data for relatively stronger magnetic field shows a sharp jump at the onset, indicating substantial enhancement of the convective heat transfer. The origin of these flow patterns, as well as the reason for a notable increase of heat transfer near the onset of convection, are then explained using low dimensional models constructed from DNS data.

physics.flu-dyn↗