Flame Wall Research Articles

Effect of geometrical dimensions on flame spread rate over U‐shaped polyurethane facade and theoretical analysis

AbstractPolyurethane is widely used in building facades as a good energy‐saving and environment‐friendly organic material, and the U‐shaped structure is a common facade form as it could improve indoor lighting and ventilation. Through numerical simulation and theoretical analysis, this study found the flame spread rate of polyurethane foam over a U‐shaped facade was affected by the structural dimensions due to three main factors: lateral air entrainment restriction, bottom air entrainment restriction, and flame fusion between the back wall flame and side wall flame. The average flame spread rate was negative linearly correlated with the back wall width and positive exponentially correlated with the side wall length. This was due to that the side walls caused lateral air entrainment restriction, reduced the heat loss, enhanced the thermal feedback, generated an upward induced airflow close to the back wall, elongated the flame, and also caused the flame fusion between side walls and the back wall, thus accelerated the flame spread process compared to the flat structures.

Analysis of the Evolution of the Surface Density Function During Premixed V-Shaped Flame–Wall Interaction in a Turbulent Channel Flow at Reτ = 395

ABSTRACT The flame–turbulence interaction and statistical behavior of the surface density function (SDF; i.e. magnitude of the reaction progress variable gradient) in the vicinity of the wall for a stoichiometric methane-air flame are investigated using a three-dimensional direct numerical simulation of a turbulent premixed V-flame interacting with an isothermal inert wall in a fully developed turbulent channel flow at a friction Reynolds number . The results show that the mean SDF significantly decreases in the viscous sublayer in comparison to the corresponding values for the same reaction progress variable in the unstretched laminar flame. Moreover, the mean values of SDF for a given value of reaction progress variable decrease in the downstream direction with the progress of flame quenching in all zones of turbulent boundary layer. The effective normal strain rate (), which acts to reduce the SDF as it increases, is much higher in the viscous sublayer than in the other layers. In the viscous sublayer, the contribution of the gradient of displacement speed in the flame-normal direction () to has been shown to dominate the fluid-dynamic normal strain rate (). This tendency is qualitatively similar to the previous findings for a V-flame interacting with an isothermal inert wall at . However, the maximum mean value of at is approximately twice of that at , which causes a sharper drop in the SDF in the viscous sublayer at higher .

The impact of Clay as wall material on Thermoelectric power generation and flame–wall interaction

The use of clay as the combustor wall material for direct heat-to-electricity conversion using a Thermoelectric generator (TEG) was evaluated in a side-wall configuration burner. The influence of different fuel blends (Methane, Biogas and Biogas + H2) and flame work conditions in TEG power and efficiency are discussed. Flame–wall interaction (FWI) of the laminar side-wall quenching (SWQ) premixed V-flame was investigated. Heat release rate (HRR) and quenching distances were analysed with the chemiluminescence of excited OH* and CH* radicals. The gases velocity field was studied with Particle Image Velocimetry (PIV). For comparison, a galvanized steel (GS) plate was used in the same working conditions. We found that the efficiency (ηTEG) and electrical power (PTEG) losses between materials increases with flame temperature. Quenching distances (dq) were larger in the flame clay wall interaction. Flame quenching was governed by heat losses independently of the wall but the impact of convection was more significant when using the clay wall.

Characterization of a Newly Designed Test Bench for Investigations of Flame–Wall Interaction

Abstract For the thermal design of combustion chambers and turbine blades in jet engines, a detailed knowledge of the combustion and of the heat loads to the walls is necessary. In general, high operating temperatures and reduced combustor size are striven for in order to increase engine efficiency and reduce weight. Consequently, the components are exposed to temperatures above the melting point of the materials and there is a growing risk of incomplete combustion within the combustion chambers. To study these effects, we setup a new test bench for fundamental investigation of chemical near-wall reactions at atmospheric pressure. First results of gaseous, nonpremixed near-wall CH4/air and H2/air flames are presented. Optical methods such as two-line laser-induced fluorescence thermometry and OH* chemiluminescence were applied. Further, the heat release to the wall was determined by means of inverse heat conduction calculation using the data of implemented thermocouples.

Buoyant turbulent diffusion flame heights of free-, wall- and corner air entrainment conditions: Experiments and global model based on mirror approach

This paper investigates buoyant turbulent diffusion flame height of gaseous fuel combustion in various air entrainment conditions: free-, wall- and corner flames. Experiments are carried out using five square burners with side lengths of 0.3 m, 0.25 m, 0.2 m, 0.15 m and 0.1 m, and five rectangular burners with side dimensions of 0.15 m × 0.3 m, 0.125 m × 0.25 m, 0.1 m × 0.2 m, 0.075 m × 0.15 m, 0.05 m × 0.1 m, to produce corner-, wall- and free flames. The heat release rates (HRR) are in the range of 6.62–26.46 kW for the rectangular burners and 2.21–35.28 kW for the square burners to have 172 experimental conditions comprehensively. It is found that for the same HRR and burner size, the flame height (ℓf) of corner restriction is larger than the flame height of wall flame, and ℓf of wall flame is larger than that of free flame. For a given HRR and air entrainment condition, the flame height decreases with the burner dimension increase. Results also show that the effect of wall or corner on the air entrainment can be represented by the “mirror-approach” with an acceptable deviation (−12.5%~13.5%). At the same time, for free and wall flames, there is a transition of dimensionless flame heights from two-dimensional air entrainment to three-dimensional air entrainment. For corner flame, most of the data fall in the two-dimensional air entrainment condition. Combined with comprehensive data of previous reports together, flame height formulas and critical non-dimensional heat release rates for the transition points are proposed for these three types of air entrainment conditions globally by using the perimeter as the characteristic length scale based on “mirror-approach”.

A phenomenological model for the description of unburned hydrocarbons emission in ultra-lean engines

The adoption of lean-burn concepts for internal combustion engines working with a homogenous air/fuel charge is under development as a path to simultaneously improve thermal efficiency, fuel consumption, nitric oxides, and carbon monoxide emissions. This technology may lead to a relevant emission of unburned hydrocarbons (uHC) compared to a stoichiometric engine. The uHC sources are various and the relative importance varies according to fuel characteristics, engine operating point, and some geometrical details of the combustion chamber. This concern becomes even more relevant in the case of engines supplied with natural gas since the methane has a global warming potential much greater than the other major pollutant emissions. In this work, a simulation model describing the main mechanisms for uHC formation is proposed. The model describes uHC production from crevices and flame wall quenching, also considering the post-oxidation. The uHC model is implemented in commercial software (GT-Power) under the form of “user routine”. It is validated with reference to two large bore engines, whose bores are 31 and 46 cm (engines named accordingly W31 and W46). Both engines are fueled with natural gas and operated with lean mixtures (λ > 2), but with different ignition modalities (pre-chamber device or dual fuel mode). The engines under study are preliminarily schematized in the 1D simulation tool. The consistency of 1D engine schematizations is verified against the experimental data of BMEP, air flow rate, and turbocharger rotational speed over a load sweep. Then, the uHC model is validated against the engine-out measurements. The averaged uHC predictions highlight an average error of 7% and 10 % for W31 and W46 engines, respectively. The uHC model reliability is evidenced by the lack of need for a case-dependent adjustment of its tuning constants, also in presence of relevant variations of both engine load and ring pack design.

Active Subspace Variation and Modeling Uncertainty in a Supersonic Flame Simulation

The active subspace method is applied to Burrows–Kurkov supersonic wall-jet flame to quantify modeling uncertainties associated with chemical kinetics, turbulence combustion models, and boundary conditions, as well as to investigate the evolution of key physiochemical process throughout the supersonic flame. One-dimensional active subspaces of the uncertain parameter space are identified for both flame liftoff length and flame–wall distance. Through the analyses of active subspaces, it shows that chemical kinetics plays a dominant role for flame initialization and early development, whereas turbulent transport becomes more important as the flame develops. The univariate response surfaces constructed in the active subspaces are then used to perform uncertainty quantification. For the flame liftoff length, the uncertainty introduced by chemical kinetics is slightly larger than that introduced by turbulence combustion models and significantly larger than boundary induced uncertainty; whereas the uncertainty in the predicted flame–wall distance is dominated by turbulence combustion modeling, and the contribution from boundary and chemical kinetics is trivial. Further comparison between two turbulence models with structural compressibility correction shows that the active subspace is insensitive to turbulence models at the same level of closure. The corrected shear stress tensor and corrected standard yield the same key physiochemical process and similar uncertainties in the predictions.

Research on combustion performance optimization in scramjet combustor with strut/wall combined fuel injection scheme

A strut/wall combined fuel injection scheme was experimentally investigated to improve the combustion performance in a flush-wall scramjet combustor fueled with liquid kerosene. A thin strut was adopted as fuel injector and flame holder, and wall injectors were also placed on side wall of the combustor, downstream of the strut. With the injection of pure oxygen through the injectors in the trailing edge of the strut, a stable flame was achieved in the combustor, and the flame interaction characteristics were experimentally investigated. Experiments were conducted at the inlet Mach number of 2.0 with different strut/wall fuel distribution schemes. Flame was divided into two parts, namely the core flame and the wall flame, and interaction mechanism between the two parts was experimentally investigated. Experiment results show that core flame is enhanced with the increasing of fuel equivalence ratio. Core flame propagates to side wall of the combustor to ignite wall fuel, forming a wall flame. Wall flame leads to a thermal chocking downstream of strut to enhance core flame. Combustor performance and flame characteristics are influenced by fuel distribution schemes. The location of wall fuel injectors is also optimized to improve combustion efficiency in this paper. These results in this paper are valuable for the future optimization of combustion organization in scramjet combustor with multi-stages fuel injection scheme.

Modelling combustion in spark ignition engines with special emphasis on near wall flame quenching

A flame front is quenched when approaching a cold wall due to excessive heat loss. Accurate computation of combustion rate in such situations requires accounting for near wall flame quenching. Combustion models, developed without considering wall effects, cannot be used for wall bounded combustion modelling, as it leads to wall flame acceleration problem. In this work, a new model was developed to estimate the near wall combustion rate, accommodating quenching effects. The developed correlation was then applied to predict the combustion in two spark ignition engines in combination with the famous Bray–Moss–Libby (BML) combustion model. BML model normally fails when applied to wall bounded combustion due to flame wall acceleration. Results show that the proposed quenching correlation has significantly improved the performance of BML model in wall bounded combustion. As a second step, in order to further enhance the performance, the BML model was modified with the use of Kolmogorov–Petrovski–Piskunov analysis and fractal theory. In which, a new dynamic formulation is proposed to evaluate the mean flame wrinkling scale, there by accounting for spatial inhomogeneity of turbulence. Results indicate that the combination of the quenching correlation and the modified BML model has been successful in eliminating wall flame acceleration problem, while accurately predicting in-cylinder pressure rise, mass burn rates and heat release rates.

On explosion characteristics of premixed syngas/air mixtures with different hydrogen volume fractions and ignition positions

This study analyzes the effect of both ignition position and hydrogen volume fraction on the flame propagation structure and overpressure wave. The interpretations are based on the propagation dynamics of the premixed syngas/air explosions at different ignition positions with different hydrogen volume fractions investigated at a constant equivalence ratio of 1 in a semi-open rectangular combustion chamber experimentally. The initial temperature T0 and pressure p0 were 298 K and 0.1 MPa, respectively. The experimental results are consistent with those obtained using the proposed analytical theory, indicating that the proposed theory predicts the early flame acceleration stage and the flame front motion well under some limited conditions. Flame–wall interaction and combustion instability decrease the theoretical flame front position than the experimental value and increase the wall contact time than the experimental value in some cases. At the same ignition location, both the overpressure growth rate and the maximum overpressure increase with increasing hydrogen volume fraction. When ignited at 0 mm off the closed end, the increased volume fraction of hydrogen would cause the initial location of the tulip flame to approach the opening end of the duct, but at an earlier time. The closer the ignition position to the opening end is, the more intense the overpressure and flame leading edge oscillation are, and the period of overpressure is similar to that of the flame oscillation. Finally, the ignition position and hydrogen volume fraction have obvious effects on overpressure, the structure and development of the flame front.

Experimental characterization and modeling of boundary conditions and flame spread dynamics observed in the UL-94V test

Abstract To characterize the UL-94 vertical burning test (UL-94 V), temperature, oxygen content, and heat flux of the burner flame were measured using an R-type thermocouple, oxygen analyzer, and water-cooled Schmidt–Boelter gage, respectively. To account for sample geometry, mock samples from Kaowool PM insulation board were used during the burner measurements. Temperature and heat flux measurements were used to express the heat flux of the burner as a linear, piecewise function of burner flame temperature and a constant convective heat transfer coefficient. UL-94 V tests were conducted using 0.27 cm thick poly(methyl methacrylate) (PMMA) samples with and without insulated sides to investigate edge effects. All UL-94 V tests were recorded on video using a 900-nm narrow-band filter to focus on emissions from soot for tracking flame length over time. In some of these tests, a heat flux gage was embedded into the sample to measure PMMA flame heat flux to verify a previously developed laminar wall flame heat feedback submodel. PMMA flame heat fluxes of insulated samples confirmed the wall flame submodel, while non-insulated samples had significantly greater heat fluxes; the flame submodel was scaled accordingly. The burner and wall flame heat flux submodels were combined with an independently developed pyrolysis model of PMMA to conduct two-dimensional, time-resolved simulations of the UL-94 V test. Burner flame oxygen content was measured to be about 5 vol% and was found to enhance PMMA decomposition. Inclusion of the oxygen effect was essential to predict ignition. The simulation predicted flame length and spread on insulated UL-94 V samples reasonably well, validating the submodels. However, the simulation underpredicted flame length on the non-insulated samples; discrepancies were attributed to burning and spread on the edges which were not modeled explicitly. The developed model can be applied to non-charring polymeric solids with no to moderate amount of dripping. Its applicability for charring and intumescent materials will be explored in the future.

Anchoring mechanisms of a holder-stabilized premixed flame in a preheated mesoscale combustor

A very lean-premixed, laminar methane–air flame is demonstrated, experimentally, to be stable in a mesoscale combustor with a flame holder. Unlike the anchoring location of the flame tip, the anchoring location of the flame root is practically independent of the equivalence ratio, inlet velocity, and thermal conductivity of the solid wall material. When the mixture becomes leaner, both the flame root and tip can adaptively shift toward the locations with higher temperatures, and additionally, the anchoring temperature of the flame root is higher. Subsequently, by means of the three-dimensional computational mechanics, their anchoring mechanisms are thoroughly analyzed in terms of the flow recirculation, stretch effect, preferential diffusion, and conjugate heat exchange. A recirculation zone or a low-velocity zone formed behind/near the flame holder and combustion chamber wall can assist the balance between the flow velocity and the flame speed for the flame anchoring, and the flame root can adaptively shift to a zone of lower local velocity. The stretch effect is not responsible for the flame root anchoring, but this effect stabilizes the flame tip by increasing the local flame speed near the flame tip. Preferential diffusion significantly promotes the local equivalence ratio near the anchoring location of the flame root, thereby facilitating the stability of this flame root, though it stabilizes the flame tip only slightly. Furthermore, the conjugate heat exchange plays an important role in preheating fuel/air and intensifying combustion, which influences the stabilization of both the flame root and tip. The shorter distance between the flame tip and the combustion chamber wall results in a stronger flame–wall coupling. These results indicate that the anchoring mechanisms for the flame root and tip differ.

Simultaneous 1D hybrid fs/ps rotational CARS, phosphor thermometry, and CH* imaging to study transient near-wall heat transfer processes

Near-wall transient heat transfer and flame–wall interaction (FWI) are topics of great importance in the development of downsized internal combustion (IC) engines and gas turbine technology. In this work we perform measurements using 1D hybrid fs/ps rotational CARS (HRCARS), thermographic phosphors (TGP) and CH* imaging in an optically-accessible chamber designed to study transient near-wall heat transfer processes relevant to IC engine operation. HRCARS provides single-shot gas-phase temperatures (40 µm spatial resolution and up to 3 mm wall-normal distances), while thermographic phosphors measures wall temperature and CH* measures the flame front position. These simultaneous measurements are used to resolve thermal boundary layer (TBL) development and associated gaseous heat loss for three important processes of gas–wall interactions: (1) an unburned-gas polytropic compression process, (2) FWI, and (3) post-flame and gas expansion processes. During a mild polytropic compression process, measurements emphasize that even a relatively small wall heat flux (≤5 kW/m2) yields an appreciable temperature stratification through a developing TBL. During FWI, thermal gradients induced by the flame are resolved within the TBL. Gases closest to the wall (y<0.2 mm) continue to experience thermal loading from polytropic compression until the flame is within ∼1.4 mm from the wall. Immediately afterwards, the wall first senses the flame as the wall temperature begins to increase. During FWI, gas temperatures up to 1150 K impinge on the wall, producing peak wall heat fluxes (620 kW/m2) and the wall temperature increases (ΔTwall=14 K). Gaseous heat loss in the post-flame gas occurs rapidly at the wall, yielding a TBL of colder gases extending from the wall as wall heat flux slowly decreases. HRCARS further captures the rapid cooling of gases in the TBL and core-gas during the mild expansion and exhaust process.

Experimental study on pulsation frequency of free-, wall- and corner buoyant turbulent diffusion flames

This work studies the flame pulsation frequency f of free-, wall- and corner buoyant turbulent diffusion flames. Five square gaseous burners with the burner side length of 10, 15, 20, 25 and 30 cm are used to produce the three types of flames. In addition, five rectangular gaseous burners with dimensions of 30 cm × 15 cm, 25 cm × 12.5 cm, 20 cm × 10 cm, 15 cm × 7.5 cm, 10 cm × 5 cm (aspect ratio of 2:1 with half surface area of the square burners) are also used to produce wall flames. Propane is used as fuel. Results show that heat release rates have no significant effect on f. The flame pulsation frequency of free flames is larger than that of wall flames, which is further larger than that of corner flames. For all types of flames, the flame pulsation frequency decreases with the increasing of the burner size. The classic 1/2 power law correlation, i.e., f~(g/D)0.5 for flame pulsation frequency of free flame can well describe the data of all types of flames (square, rectangular; free flame, wall flame and corner flame) on the basis of the “mirror” approach, and the perimeter-equivalent diameter is used as the appropriate length scale.

Modeling heavy-duty diesel engines using tabulated kinetics in a wide range of operating conditions

Fast and high-fidelity combustion models including detailed kinetics and turbulence chemistry interaction are necessary to support design and development of heavy-duty diesel engines. In this work, the authors intend to present and validate tabulated flamelet progress variable model based on tabulation of laminar diffusion flamelets for different scalar dissipation rate, whose predictability highly depends on the description of fuel–air mixing process in which engine mesh layout plays an important role. To this end, two grids were compared and assessed: in both grids, cells were aligned on the spray direction with such region being enlarged in the second one, where the near-nozzle and near-wall mesh resolution were also improved, which is expected to better account for both spray dynamics and flame–wall interaction dominating the combustion process in diesel engines. Flame structure, in-cylinder pressure, apparent heat release rate, and emissions for different relevant operating points were compared and analyzed to identify the most suitable mesh. Afterwards, simulations were carried out in a heavy-duty engine considering 20 operating points, allowing to comprehensively verify the validity of tabulated flamelet progress variable model. The results demonstrated that the proposed approach was capable to accurately predict in-cylinder pressure evolution and NO x formation across a wide engine map.

Analysis of Wall–flame Interaction in Laminar Non-premixed Combustion

ABSTRACT The study is aimed at demonstrating a methodology for the time-scale characterization of the chemistry-wall-heat-transfer interaction. The driving chemical time-scale is estimated by means of the tangential stretching rate, and a proper thermal timescale for the temperature-time variation due to wall heat flux is presented. A thermal Damköhler number, Dath , is proposed as the ratio of the two. The methodology is applied on a prototypical laminar methane-oxygen diffusion flame impinging on an isothermal cold wall. Non-adiabatic effects are described qualitatively and a CSP-TSR analysis is performed to obtain topological information and physical insights. The thermal Damköhler number field is computed and discussed to highlight the interplay between chemical and diffusive processes and to a-priori assess the accuracy of the steady laminar flamelet assumption under non-adiabatic conditions.

High-frequency wall heat flux measurement during wall impingement of a diffusion flame

The efficiency of internal combustion engines is limited by heat losses to the wall of the combustion chamber. A precise characterization of wall heat flux is therefore needed to optimize engine parameters. However, the existing measurements of wall heat fluxes have significant limitations; time resolution is often higher than the timescales of the physical phenomena of flame–wall interaction. Furthermore, few studies have investigated diesel flame conditions (as opposed to propagation flames). In this study, the heat flux generated by a diffusion flame impinging on a wall was measured with thin-junction thermocouple, with a time resolution of the whole acquisition chain better than 0.1 ms. The effects of variations in ambient gas temperature, injection pressure and injector–wall distance were investigated. Diesel spray impingement on the wall is shown to cause strong gas–wall thermal exchange, with convection coefficients of 6–12 kW/m2/K. Those results suggest the necessity of close-wall aerodynamic measurements to link macroscopic characteristics of the spray (injection pressure, impingement geometry) to turbulence values.

On the influence of wall distance and geometry for high-pressure n-dodecane spray flames in a constant-volume chamber

To isolate the effect of flame–wall interaction from representative operating conditions of an internal combustion engine, experiments were performed in a constant-volume pre-burn vessel. Three different wall geometries were studied at distances of 32.8, 38.2, and 46.2 mm from a single-hole 0.09-mm orifice diameter fuel injector. A flat wall provides a simplified case of flame–wall interaction. To mimic the division of a jet into two regions by the piston bowl rim in an engine, a two-dimensional confined wall is used. A third, axisymmetric confined wall geometry allows a second simplified comparison to numerical simulations in a Reynolds-averaged Navier–Stokes framework. As a limiting situation for a free jet, the distance from the injector orifice to the end wall of the chamber is 95 mm. Thermocouples installed in the end wall provided insights into local heat losses for reference cases without a wall insert. The test conditions were according to the Engine Combustion Network Spray A guidelines with an ambient temperature of 900 K and an ambient density of 22.8 kg/m3 with 15% O2. Flame structures were studied using high-speed OH* chemiluminescence with integrated single-shot OH PLIF and combined with pressure-based apparent heat release data to infer combustion progress and spray behavior. Soot was studied in a qualitative manner using high-speed natural luminosity imaging with integrated high-speed laser-induced incandescence. Overall, increased mixing upon interaction with the surfaces is observed to increase early heat release rate and to significantly reduce soot, with the nearest wall distance showing most effect. The flat wall gives rise to the most significant effects in all cases.

QFM: quenching flamelet-generated manifold for modelling of flame–wall interactions

This work introduces a new method to improve the accuracy of the flamelet-generated manifold (FGM) method under conditions of flame–wall interactions (FWI). Special attention is given to the prediction of the pollutant CO.In existing FGM methods, in order to account for heat loss, usually flamelets with constant enthalpy are utilised. These constant enthalpy flamelets used to generate the manifold, do not include the effects of wall heat loss on the manifold composition, resulting in simulation inaccuracies in the near-wall region, where large enthalpy gradients are present. To address this issue, the idea to utilise 1D head on quenching (HOQ) flamelets for tabulated chemistry is adopted and applied here in the context of the FGM method. The HOQ qualitatively resembles the general phenomena of FWI. However, the rates of wall heat loss and the accompanied effects on the chemical species composition may quantitatively differ between various FWI configurations. In addition, the magnitude of heat transfer rate may vary in space and time in general configurations. Therefore, in this work, a method is introduced to generate a 3D manifold, based on multiple HOQ-like flamelets, that includes the variation of the rate of heat loss as an extra table dimension. This dimension is parametrised by a second reaction progress variable for which a transport equation is solved next to the equations for enthalpy and the first progress variable.The new developed method, referred to as Quenching Flamelet-generated Manifold (QFM), is described in this work. Further, the method is validated against detailed chemistry simulations of a two-dimensional premixed laminar side-wall quenching of a methane-air flame. A comparison is presented, analysing the performance obtained using the existing 2D FGM method, a 2D QFM that is based on a single HOQ flamelet which does not account for a varying rate of wall heat loss and a 3D QFM, which does. Finally, it is shown that the 3D QFM tabulated chemistry simulation yields a very good level of accuracy and that the accuracy for prediction of CO concentrations near the wall is improved tremendously.

Flame propagation and flashback characteristics in a kerosene fueled supersonic combustor equipped with strut/wall combined fuel injectors

Flame propagation characteristics in a liquid kerosene fueled supersonic combustor were observed in this paper. The strut/wall combined fuel injection scheme was used to achieve the multi-staged fuel injection, and the thin strut was also adopted as the flame holder. A series of experiments and numerical simulations were carried out in the condition of Ma=2.8. In order to represent the flame characteristics during the flame propagation process, high-speed photography and pressure measurement were taken to record the data. Based on the basic date, the variation of flame area and Mach number distributions during the experimental process were analyzed, and the flowing field characteristics were also discussed according to the numerical results. Results indicated that the core flame would diffuse to the primary flow with the increasing of equivalence ratio, enlarging the flame width. The flame flashback phenomenon of both core flame and wall flame, induced by the flame propagation, was detected and analyzed. Then, the interaction between combustion and flowing in the supersonic combustor was investigated, based on which, the mechanism of flame flashback phenomenon was also revealed. Numerical results showed that the generation of recirculation and low-speed region in boundary layer was the inducing mechanism of flame flashback. With the investigations in this paper, a depth understanding of the flame propagation process in supersonic airflow was achieved, based on which, a further optimization of combustion performance could be conducted.