Take a look at what we’re working on…
Staged gas turbines with two sequential combustion chambers are being developed for power generation for their ability to achieve low emissions within a wide operational range while conserving high thermal efficiency. A particular implementation of the sequential combustion concept is characterized by a reheat combustion stage downstream of a first premixed-type combustor. Hot exhaust gases from the first stage are mixed with fuel in a mixing section, which provides the inlet conditions for the second-stage reheat combustor. Direct numerical simulations (DNS) of flame stabilization regimes in the reheat burner, i.e. the combustor including the mixing section (duct-in-a-duct), at idealized conditions are performed using a detailed hydrogen-air mechanism. In particular, this project investigates the role of autoignition in flame stabilization under different stratified conditions of fuel-oxidizer mixture at the inlet. Results from the simulations show that the combustion occurs both due to auto-ignition as well as flame propagation. The auto-ignition mode occurs at and around the centerline of the combustor while flame propagation is stabilized at the recirculation zones near the corners. A typical realization of the flow field is shown in the figure. Chemical explosive mode analysis has been employed to quantify the contribution of auto-ignition to the combustion rate relative to flame propagation. The insights from this project are being used by Ansaldo Energia (a major gas turbine company), for the design their next generation gas turbine combustor.
Flame stabilization in scramjet combustors used in hypersonic flight is a challenge. At the relevant flight conditions that result in short residence times of the order of 0.1 ms, flame stabilization is often achieved using a cavity flame holder that recirculates hot products of combustion including radicals. In this study, we use massively parallel direct numerical simulations (DNS) to understand the stabilization of a lean ethylene-air premixed flame in the shear layer above a rectangular cavity and the interactions between the flame and the shear layer vorticity. We found that the flame stabilization is significantly affected by the shear flow dynamics between the main flow and the cavity flow, while the recirculation flow pattern within the cavity is dictated by its aspect ratio. Results also show that the recirculation zone transports necessary hot radicals such as OH from the downstream region to the stabilization point. The vorticity which is present mainly on the reactant side of the flame near the stabilization point, gets advected into the products downstream. This leads to an enhanced heat transfer from the flame to the wall, and affects the CO and OH oxidation. The finding suggests that a better cooling system needs to be provided in the design of a scramjet combustor to prevent thermal damages. The data from these simulations will be used to develop reduced order models necessary for engineering simulations. These insights are applicable to cavity-stabilized configurations for gas turbines as well
Chemical kinetic (CK) is a vital part of combustion system modeling. It provides reaction mechanisms and rate constants extracted from theoretical approaches. However, CK models are only valid in a specific range of temperatures and pressures and eventually fail at extreme conditions. Reactive molecular dynamics (MD) can incorporate the complexities at extreme conditions and provide the entire mechanism as well as the parameterization required for the combustion systems. MD is often performed at two levels: reaxff-based MD (RMD) and ab-initio MD (DFT-MD). RMD has lower compute costs, but it is less accurate than DFT-MD. We perform CK modeling of the hydrogen-oxygen system using RMD and DFT-MD to extract the complete reaction mechanism and the statistically computed rate constants from atomistic simulations. Arrhenius parameters are then fitted for each elementary reaction using the computed kinetic rates at different temperatures, and a detailed CK model is generated for hydrogen oxidation. The parametrization obtained from both the MD approaches are evaluated by computing the flame speed of a one-dimensional premixed in continuum scale simulations and compared against standard results, showing good agreement.
Designing scalable CFD codes on massively parallel computers is a challenge. This is mainly due to the large number of communications between processing elements (PEs) and their synchronization, leading to idling of PEs. Indeed, communication will likely be the bottleneck in the scalability of codes on Exascale machines. Our recent work on asynchronous computing for PDEs based on finite-differences has shown that it is possible to relax synchronization between PEs at a mathematical level. Computations then proceed regardless of the status of communication, reducing the idle time of PEs and improving the scalability. However, accuracy of the schemes is greatly affected. We have proposed asynchrony-tolerant (AT) schemes to address this issue. In this work, we study the effect of asynchrony on the solution of fluid flow problems using standard and AT schemes. We show that asynchrony creates additional scales with low energy content. The specific wavenumbers affected can be shown to be due to two distinct effects: the randomness in the arrival of messages and the corresponding switching between schemes. Understanding these errors allow us to effectively control them, rendering the method’s feasibility in solving turbulent flows at realistic conditions on future computing systems.
Deep learning has been extensively used for modeling and analyzing fluid turbulence. One such application is the use of super-resolution (SR) algorithms to reconstruct small-scale structures from large-scale ones for turbulent flows. However, current SR algorithms require high-resolution reference data for training, which is not practical for most fluid flow scenarios. To address this issue, a physics-guided model has been developed using generative adversarial networks to reconstruct small-scale structures relevant to homogeneous isotropic turbulence. The quality of the reconstruction was assessed using spectra, structure functions, and probability density functions of velocity components and a passive scalar. The results showed that the outputs were in statistical agreement with the ground truth, which was excluded from the training, and that the trained network generalized well across Reynolds numbers. This work lays the foundation for reconstructing small-scale structures from large-eddy simulation data without the need for high-resolution reference data.
With the increasing availability of data, machine/deep learning methods are becoming an important tool for modeling, analysis and prediction of several phenomena. In this work, we develop a new anomaly detection method to identify anomalous/extreme events in scientific phenomena, which are often described by multi-scale, multi-variate, smoothly varying data. The method leverages the statistical signature of anomalies hidden in the higher order statistical moments. For multi-variate data, this translates to an examination of the higher order joint moments and their association with anomalies or extreme events. Specifically, we use the direction of the principal vectors obtained from the decomposition of the fourth order joint moment tensor, in characterizing the occurrence of anomalous events. We use this method to identify the inception of auto-ignition kernels in turbulent premixed combustion and the extreme intermittent events in isotropic turbulence.