Unsteady flame–wall interaction: Impact on CO emission and wall heat flux

Abstract

This study investigates the interaction of a two-dimensional (2D) laminar methane-air premixed flame with a constant temperature wall using Direct Numerical Simulation (DNS). The flame is excited using velocity perturbations at the inlet for a range of forcing frequencies. The GRI 3.0 chemical mechanism is used to perform the simulations. The flame behaviour is first analysed by comparing the results to those observed for a steady 2D laminar flame interacting with the wall and a one-dimensional head-on quenching (HOQ) flame under the same conditions as the 2D flame. This is followed by analysing the wall heat flux and CO emission due to the flame quenching at the wall for different forcing frequencies. At low forcing frequencies, the flame is observed to sweep against a large portion of the wall whereas at high frequencies, the flame has an insignificant response to the incoming velocity perturbations. In both these regimes, side-wall quenching (SWQ) is the dominant mechanism of flame–wall interaction. However, at an intermediate frequency, both HOQ and SWQ occur at different stages of flame–wall interaction. The analysis of the total integrated wall heat flux of forced flames shows a large variation of this quantity compared with a steady state flame, with the highest fluctuations occurring for the flame forced at the intermediate frequency. The radical recombination reactions are found to be significant in the vicinity of the wall, contributing to about 50% of the total HRR at the wall at the quenching instant. The analysis of the species transport budget for CO shows that CO transport near the wall close to the flame tip is dominated by convection and diffusion.

Introduction

Flame–wall interaction (FWI) is commonly present in many combustion systems such as gas turbines and reciprocating engines. This is an important phenomenon due to its influence on the produced emissions and fuel consumption. In particular, with the current trend towards increasing the power density in energy producing systems, FWI plays a more critical role in the new generation of modern engines. As a result, a full understanding of how flames behave near walls and the impact of FWI on the produced emissions and combustor performance is a topic of interest.

FWI can be characterised using the geometrical features of the flame near the wall such as the flame orientation during the interaction. Head-on quenching (HOQ) is one of the well-studied FWI configurations in the literature, e.g. [1], [2], [3], [4], [5], [6], [7]. During HOQ, the flame propagates perpendicular to the wall until the heat loss to the wall stops the flame propagation and finally leads to flame extinction. The minimum flame–wall distance during this process is referred to as the quenching distance (δ_Q). The wall heat flux increases as the flame approaches the wall and commonly reaches its peak once the flame has the minimum distance from the wall [1]. Another common FWI configuration is known as side-wall quenching (SWQ), which is a localised quenching at the wall. This type of quenching can occur when a laminar flame propagates parallel to the wall. However, in most SWQ configurations, the flame normal is not parallel to the wall, even for large distances away from the wall [1], [3], [7], [8], [9], [10]. SWQ typically exhibits larger quenching distances compared with HOQ which in turn reduces the wall heat flux.

Accurate experimental studies of FWI introduce a tremendous challenge because quenching of the flame involves very small time and length scales and therefore it is difficult to measure the parameters of interest accurately [3], [11]. Fully-resolved simulations, such as direct numerical simulation (DNS) can complement experimental and theoretical studies and therefore improve our understanding of the impact of FWI on emissions and wall heat flux.

Early fully resolved simulations of FWI were performed for premixed flames in a HOQ configuration, e.g. [1], [5], [12]. The focus of these studies were primarily on how quenching distance and heat flux vary under different conditions. Two-dimensional (2D) and three-dimensional (3D) DNSs have also been performed to study FWI [1], [13]. For instance, in a 2D DNS study, Poinsot et al. [1] showed that the maximum heat flux and quenching distance for turbulent cases were of the same order of magnitude as that in laminar cases. The impact of horseshoe vortices on FWI and wall heat flux was also investigated in a 3D DNS study with simple chemistry [13]. Near-wall vortices were found to push the flame towards and away from the wall at various locations. Thus, the mean and local wall heat fluxes are affected by the near-wall horseshoe vortices. More recently, Gruber et al. [14] highlighted the importance of flame thickening during FWI. They also found that coherent turbulent structures have an important contribution to the wall heat transfer by pushing the hot products towards the wall. By performing a spectral analysis, they related the wall heat flux to the dominant time and length scales of these coherent turbulent structures.

Fundamental studies like [5], [12] have identified the role of low temperature chemistry near the wall. The highly reactive radical species which are trapped between the flame and the cold wall are subject to low temperature chemistry. The low activation energy, exothermic reactions involving radicals become important near a cold wall. Detailed analysis of radical recombination reactions have been reported for n-heptane flames [15] and hydrogen flames [14]. Owston et al. [15] observed higher heat release rate (HRR) at the wall during quenching. This was attributed to the low temperature chemistry which becomes dominant at the cold wall. Radical recombination reactions for H, O and OH were found to contribute to the increase of the HRR at the wall during quenching. Gruber et al. [14] also found that the zero-activation energy, exothermic radical recombination reactions were responsible for about 70% of the total HRR at the wall for rich, hydrogen/air flames.

Recently, several experimental [9], [10], [16], [17] and computational studies [8], [18] on CO emissions during FWI have been reported. In an experimental study by Mann et al. [16], a HOQ-like configuration was considered. Substantial amount of CO was observed near a cold wall at the quenching instant. This was attributed to the low reaction rates of CO formation and oxidation reactions at low temperatures [16]. In a number of experimental studies at TU Darmstadt, a SWQ burner was used [9], [10]. A detailed analysis of the CO mole concentration versus temperature (CO - T) scatter plots showed that both CO production and oxidation branches were shifted to lower temperatures, in the vicinity of the wall [10]. Kosaka et al. [9] compared CO - T scatter plots obtained from experimental data with 1D freely propagating flame calculations with and without wall heat loss. This was done to understand influence of the wall heat transfer on CO emissions. The wall heat loss was modelled by adding an enthalpy defect term in the energy equation. CO oxidation was predicted with reasonable accuracy by the non-adiabatic 1D flame calculations. However, CO formation, which could not be deduced using 1D calculations, was inferred to be influenced by heat transfer in multiple dimensions. Furthermore, studies have investigated the CO transport mechanisms to understand the near-wall CO behaviour [8], [18]. It is found that the CO produced by the flame in a region which is not subject to a large heat loss will diffuse towards the wall, being the primary reason for a high concentration of CO near the wall [8]. The dominance of diffusion as a mechanism for CO transport was also highlighted by Jiang et al. [18].

The flow unsteadiness introduces a range of time scales into the problem, which in competition with chemical time scales can influence the amount of CO emissions near the wall. The unsteadiness can also affect the flame configuration during its interaction with the wall, which again can impact the CO emissions as well as the wall heat flux. Some recent experimental studies, discussed earlier, have considered unsteadiness and provided insight into the near-wall CO behaviour and the wall heat flux during FWI [9], [10]. The CO - T scatter plots were the primary focus of these studies. However, as suggested by Jainski et al. [10], a full understanding of the near wall flame behaviour including the CO transport mechanism requires DNS with detailed chemistry.

3D DNS of turbulent FWI features a high computational cost. Previous DNS studies in this context have used either simple chemistry [13], [19] or detailed chemistry for hydrogen [14]. For complex fuels like methane, detailed chemistry imposes a much larger computational cost. Therefore, the most recent DNS study on CO emission due to FWI for methane/air mixtures is limited to 2D steady state [8]. A laminar flame subjected to velocity perturbation can be used as an alternative approach to study the transient effects of FWI. This allows us to examine the flame dynamics and its impact on CO emission and heat flux for a given forcing frequency. As a result, FWI can be investigated parametrically for isolated flow time scales. This approach is commonly used in the context of thermoacoustic instability in which an acoustically forced laminar flame is used to obtain the transfer function for describing the flame response to the incoming flow perturbations, e.g. [20], [21]. The obtained insight can be then used to develop low order models for more complex systems.

The aim of this paper is therefore to carry out such study in the context of gas turbines. To do so, a preheated mixture of methane/air flame subject to velocity perturbations at the inlet is studied using direct numerical simulation and the flame dynamics and emission characteristics are analysed. Methane is chosen as the fuel because it is the main component of natural gas commonly used in stationary gas turbines for power generation. Another important feature of this study is using a preheated mixture, relevant to the conditions at the combustor inlet in these gas turbines.

This paper is structured as follows. Section 2 discusses the computational domain, numerical methods and the simulation parameters. The results and discussion section is divided into 6 sub-sections. Sections 3.1 and 3.2 present studies on 1D HOQ and 2D steady configurations, respectively. The flame dynamics, HRR and wall heat flux characteristics of forced flames are explained in Section 3.3 using three representative cases. The effect of radical recombination reaction on the near-wall heat release rate is detailed in Section 3.4 and the near-wall CO emissions are explored in Sections 3.5 and 3.6. While Section 3.5 intends to provide an understanding of the near-wall CO behaviour, the transport mechanisms controlling the same are studied in Section 3.6.