Research on the Ignition Process and Flame Stabilization of a Combination of Step and Strut: Experimental and Numerical Study

Abstract

A combined application of step and strut was put forward to achieve reliable ignition and flame stabilization. In this work, the ignition process and temperature distribution have been tested, and a new reduction approach applied to jet fuel oxidation mechanism was developed to present a flow map via tracking C and H reaction paths, then the minor and major reactions were verified according to relative occurrence probabilities. With the half decrease of mechanism size, bias occurred and was controlled within 1.8%. This reduction method had such characteristics as universality, intuition, and quantification, due to its inherent simplification theory. This simulation of ignition process was always consistent with experimental results, which depicted kernel generation, flamelet breakup and flame propagation. Also, the influence of inlet temperature on outlet temperature and component distribution was performed, the biases of experimental and numerical results were within 5%. Chemical characteristics of Kerosene/air premixed combustible had changed and side reactions occurred to jet fuel above 900 K, which led to a converse effect on flame spreading. The side reactions aggravated the increasing coproducts of CO and CH4, which caused the decrease of volumetric heat production.

1. Introduction

With the improvement of gas turbine technology, combustion technology is constantly improving. For high-performance combustion, it needs to meet the requirements of reliable ignition and flame stabilization under various inlet conditions and a wide operating range [1]. Bluff body is frequently utilized within advanced engine afterburners and ram combustion chambers for flame stabilization in high speed airflows [2]. The step and strut are two typical model for ignition and flame stabilization, their basic performances are conducted such as ignition and extinction limits. It is also a great challenge to explore the thermal characteristics including ignition process and flame stability due to the high speed of flow and intense turbulence in the combustor [3,4,5]. Consequently, the detailed numerical simulation of their combustion continued to be out of reach [6,7,8]. To better understand the mechanism of ignition and flame stability, it is desperately needed to develop a high-precision and low-cost simulation method, which is adapted to wide range variation [9,10,11].

The most prevalent proposal to circumvent this problem is the use of a surrogate fuel. A surrogate fuel should be a mixture of a limited number of components capable of emulating some physical and/or chemical property of this target real fuel [12,13]. A surrogate consists of a mix of hydrocarbons n-decane 80% and 1,2,4-trimethybenzene 20% by weight, which is called the Aachen surrogate, and it is selected for consideration as a possible surrogate of kerosene [14]. The hydrogen to carbon (H/C) ratio for this surrogate is 2.0 this is close to the H/C ratio of 1.92 for JP-8. Additionally, the air temperature at autoignition agreed with experimental data at low values of the strain rate, which can well reproduce autoignitionand extinction characteristics of JP-8.

To better model and simulate chemical reaction process, there have been extensive number of kinetic model established for surrogate. A single detailed kinetic mechanism for the oxidation including 7920 reactions of n-decane and n-heptane has been written [15]. And several models have been built for ignition/autoignition, for example, one was for ignition involving 506 species and 3684 reactions [16]; the other was for autoignition involving 600 reactions and 67 species [17]. Pyrolysis model of n-decane was compiled and validated including 144 species and 1021 reversible reactions. Also, a large amount of models were built to predict formation as well as thermoacoustic instability [18,19]. All of these models mentioned above described kinetic of reaction well, but the elementary chemical kinetic mechanism or full chemistry was usually too detailed, computationally expensive and stiff resulting in a great challenge in combustion modelling [20,21]. A starting point for most of the reduction methods was a detailed mechanism, e.g., GRI-mech 3.0 [22], which has been studied extensively and validated experimentally serving thus as a base for different smaller mechanisms.

Several reduction methods was derived based on its merits and restrictions. Sensitivity analysis investigated the static properties of a mechanism [23,24] and species-target reduction [25,26]; this method identified species coupled together, it required an iterative procedure, and the selection of threshold values was arbitrary. Rate-Controlled Constrained Equilibrium (RCCE) [27] was combined with Computational Singular Perturbation (CSP) to identify and eliminate only the elementary reactions that were not important for any species through the use of important index [28]; CSP method [29] fully considered the time dependency of the Jacobian matrix and accurately identified the fast modes, although the refinement procedure for the time-dependent Jacobian demands heavy computation time. Consequently, for practical purposes in generating static reduced mechanisms, CPS was often used together with the constant Jacobian assumption [30]. The intrinsic low-dimensional manifold (ILDM) method that was similar in concept to CPS method was developed due to its generality and simple conception; while the ILDM was tabulated as function of a small number of variables and used in applications.

Besides the aforementioned methods, a straightforward and high-precision method of small-scale mechanism is lacking, and it is hard to calculate the thermal characteristics under complicated aerodynamic conditions, such as ignition process, outlet temperature and components distribution. In this work, a new simplification approach was developed to reduce chemical reactions of RP-3 jet fuel by tracking Carbon (C) and Hydrogen (H) reaction paths. The reaction paths with various occurrence probabilities determined minor and major reactions. This approach was generalized to adapt any elementary reactions and it illustrated quantitative and qualitative relationships among the coupled species [31]. Additionally, the accuracy and precision of mechanism have been validated over a wide range of initial gas temperatures and pressures. Also, the experimental results including ignition process, outlet temperature and components distribution were compared with numerical results over a wide range of inlet aerodynamic conditions. The detailed characteristics of ignition process were depicted even predicted for further range. And the influence of chemical kinetics and gas dynamic on outlet temperature and components was performed.

2. Methodology

2.1. Experimental Set-Up

To investigate the performance of designated combination of step and struts, premixed flow of jet-A and vitiate gas avoided the influence of two-phase. An experimental set-up was designed and presented in Figure 1, which was used to supply various inlet conditions. The inlet conditions of burner covered a large range, such as inlet temperatures (600–1200 K) and inlet velocities (50–200 m/s). This set-up involved air supply system, fuel supply system, preheating system like combustor 1 and combustor 2, and combined burner with various struts. Thermocouples and flue gas analyser were to monitor temperature and components of hot gas that was inputted into burner. The two combustors provide a basis for wide-speed incoming flow with wide-range temperature variation, and reduce the consumption of combustion resources under different working conditions through valve adjustment.

The burner involved a 100 mm vertical by 150 mm horizontal cross-section inlet and quartz windows for sidewalls, allowing full optical access to the flame shown in Figure 2. The step structure was ameliorated from conventional truncated cone and optimized with 30 mm-height step for pilot ignition. And the strut were designed with rear section widths (30 mm) and slant angles (30 deg), which was well validated with low pressure loss and good thermodynamic characteristics.

2.2. The Approach of Mechanism Reduction

This new approach was presented in Figure 3. A detailed mechanism was set as original mechanism named Mechanism_original.

Firstly, it was divided to several parts based on elements, reactions including C and H were extracted, namely, Mechanism_initial. Secondly, Mechanism_initial was simulated using CANTERA that was a suite of object-oriented software tools for problems involving chemical kinetics, thermodynamics, and/or transport processes [32]; and minor and major reactions were verified intuitively according to relative occurrence probabilities. Here reduced Mechanism_C, H was obtained via tracking C&H reaction paths. At the same time, the reduction method of Sensitivity Analysis was also conducted according to the sensitivity of intermediate species, reduced Mechanism_sensitivity with similar size was established. And then the two reduced mechanisms and Mechanism_initial were simulated using CHEMKIN to make comparison. The reduced mechanism with less bias was applied for further simplification via enlarging filtration range. Finally, with the decrease of mechanism size, the error analysis was performed to evaluate the bias brought by discounting reactions.

2.3. Simulation Method Using CFD

And this numerical simulation was conducted using Fluent 19.2, the main target was to investigate its aerodynamic performance in a reactive flow. A realizable k-e model was selected as a turbulence model. The near wall was processed using a standard wall function method. Pressure and velocity coupling was achieved using the SIMPLE (Semi-Implicit Method for Pressure Linked Equations) algorithm. Mechanism_further was used as reaction mechanism. The simulation model was simplified shown in Figure 2, the meshes were refined locally at the combination of step and strut, where the minimum was 0.5 mm and the maximum was 0.7 mm. Also, it was compared with smaller size (0.4 mm–0.6 mm) mesh to verify mesh independence. The detailed boundary conditions are shown in

Table 1. Boundary conditions.

Boundary Conditions	Boundary Location	Parameters
Velocity inlet	Inlet	V = 50, 100, 150, 200 m/s; T = 320 K
Outflow	Outlet
Wall	Solid wall and liquid boundary	Stationary wall; no slip; no heat flux

.

3. Results and Discussion

3.1. The Process of Mechanism Reduction

3.1.1. Selection of Original Mechanism

To describe the reduction approach distinctly, a skeletal mechanism for n-decane oxidation including 141 reactions shown in Table 1 which was developed by Chang et al. [33], had been taken as the original. And the thermodynamic data of the species refered to the detailed chemical mechanisms developed by Lawrence Livermore National Laboratory.

3.1.2. Preliminary Mechanism Reduction

Emission of NO_X was not taken into consideration, a preliminarily reduced mechanism involving 128 reactions (deducting 129th–141st reactions from the original) was obtained, named Mechanism_initial (Mech_Ini). Numerical simulations using Mechanism_initial were conducted with stoichiometric mixture of n-decane/air. The diagram of chemical transformations tracking C&H according to this initial mechanism was shown in Figure 4 and Figure 5. The weight of lines and the value beside the lines represented the relative occurrence probabilities of reactions. It was quantitative and qualitative to demonstrate the coupling relationship of species. C&H reaction paths were straightforward to identify and then eliminate the unimportant elementary reactions that contribute negligibly to the production of every species. It well avoided the problem of discounting unimportant species that was possibly kept in a group strongly couple to it. This approach reproduced reaction diagram that demonstrated main and intermediate reactions successfully. The major paths of consuming fuel were thermal decomposition leading to H abstraction. The following notation was used here and in what follows: RH = C_nH_2n+2, R = C_nH_2n+1, Q = C_nH_2n. Once reacting, C₁₀H₂₁ was the indispensable and coupled strongly with several long-chain hydrocarbons, which determined 1st–7th reactions cannot be neglected. It included:

(1)

Oxidation of alkane and alkyl radical: RH + O₂ = R + HO₂, R + O₂ = QOO;

(2)

H-atom abstraction from alkane: RH + H = R + H₂, RH + OH = R + H₂O;

(3)

Isomerization of the alkyl radical: (ROO)₁ = (ROO)₂.

The products of the decomposition of alkyls higher than C₃ also underwent thermal decompositions fast and formed C₃- and C₂- species, with ethylene (C₂H₄) and propene (C₃H₆) being the important stable ones. C₂H₅ (or C₂H₄) and C₃H₆ were the only two resultants from C₁₀H₂₁, therefore, 8th–12th reactions were remained. Here summarized, 1st–12th were decomposition process from long chain structure to short chain structure, which constituted main reactions. The main reduction was performed to C₂-C₃ skeletal mechanism and H₂/CO/C₁ detailed mechanism. The species occurred in Figure 4 and Figure 5 were coupled strongly and connected tightly.

The reactions related to these species were with various importance, it was basic reference to eliminate some reactions which occurred with probability smaller than 0.02 were removed firstly. And some reactions with the species not existing in the reaction diagram were discounted, such as species CH₂CO. Thus, a more simplified reaction mechanism was put forward named Mechanism_C, H (Mech_CH) including 87 reactions (deducting 41 reactions: 15th, 23rd, 33rd–37th, 39th, 50th, 53rd–61st, 68th, 71st, 72nd, 86th, 92nd–100th, 104th, 105th, 107th, 110th,111st, 119th, 120th, 122nd, 123rd, 125th).

For comparison, the initial mechanism was also simplified using the reduction method of Sensitivity analysis. Sensitivity analysis of intermediate species, such as C2H4, C2H5, and C3H6, was conducted via CHEMKIN; reactions which satisfied the inequation blow were removed:

(Reaction − Sensitivity)_i/(Reaction − Sensitivity)_max < 0.01

where (Reaction-Sensitivity)_i was sensitivity of each reaction, and (Reaction-Sensitivity)_max was the biggest value of all. Then reduced mechanism by sensitivity analysis was established, named Mechanism_sensitivity (Mech_Sen, deducting 10th, 12nd, 20th, 24th, 27th, 29th, 30th, 33rd, 34th, 37th–39th, 44th, 46th–48th, 54th, 57th–65th, 69th–71st, 74th–76th, 88th, 90th, 91st, 98th, 99th, 107th–110th, 113rd–116th, 120th, 121st, 123rd), whose size (including 84 reactions) was similar to Mechanism_C, H.

3.1.3. Comparison and Validation

The two reduced mechanisms (Mechanism_C, H, Mechanism_sensitivity) and initial one (Mechanism_initial) were applied to simulate closed, homogenous and constrain-volume reaction via CHEMKIN. The evaluation indexes were critical factors to assess combustion efficiency and performance, such as heat production, temperature ratio, pressure ratio, H₂O, CO₂, and CO. Moreover, the objective was also to validate the reduction method by tracking C&H reaction paths.

As shown in Figure 6, the results of evaluation indexes, such as heat production, temperature ratio, pressure ratio, and H₂O, CO, and CO₂, illustrated the similar trend; when temperature increased, the biases decreased. However, the simulation results of CO₂ and CO using Mechanism_sensitivity went absolutely against using Mechanism_initial and Mechanism_C, H.

As was shown in Figure 4, the species of HCO was on the key position, especially for the production of CO and CO₂. Reactions from 62th to 75th were related tightly to species of CO/HCO. The simplification of Sensitivity Analysis method removed this part then led to misfit the initial mechanism. While results of Mechanism_initial and Mechanism_C, H went with same trend and appeared little difference as to each evolution index, this demonstrated that the approach via tracking C&H reaction paths was a validated reduction approach.

3.1.4. Further Simplification of Mechanism and Error Analysis

Further simplification of mechanism (Mechanism_further) was performed through enlarging the discounting limit from 0.2 to 0.3, and 9 reactions were discounted (deducting 51st, 52nd, 85th, 87th, 113rd, 116th–118th, 128th). Mechanism_further (Mech_Fur) involving 78 reactions was also applied to simulate closed, homogenous and constrain-volume reaction with stoichiometric mixture of n-decane/air via CHEMKIN. The simulation results of evaluation indexes, such as heat production, temperature ratio, pressure ratio, H₂O, CO₂, and CO, were shown in Figure 7.

Each evaluation index of Mechanism_further went the same trend with Mechanism_initial and Mechanism_C, H, which verified the method of further simplification method was available. Due to discounted elementary reactions, biases occurred, and it inevitably did. Error analysis would be imperative to measure whether further simplification need to be continued. As to the biases of Mechanism_C, H and Mechanism_further, the results of Mechanism_initial application was as standard.

$${\mathrm{Error} \mathrm{analysis}}_{\left({\mathrm{Mech}}_{\_\mathrm{CH}}\right)}=\frac{\left({ \mathrm{Index}}_{\left({\mathrm{Mech}}_{\mathrm{Ini}}\right)}-{ \mathrm{Index}}_{\left(\mathrm{Mech}{\_}_{\mathrm{CH}}\right)}\right)}{{\mathrm{Index}}_{\left({\mathrm{Mech}}_{\mathrm{Ini}}\right)}}$$

(1)

$${\mathrm{Error} \mathrm{analysis}}_{\left(\mathrm{Mech}\_\mathrm{Fur}\right)}=\frac{\left({\mathrm{Index}}_{\left(\mathrm{Mech}\_\mathrm{Ini}\right)}-{ \mathrm{Index}}_{\left(\mathrm{Mech}\_\mathrm{Fur}\right) }\right)}{{\mathrm{Index}}_{\left(\mathrm{Mech}\_\mathrm{Ini}\right)}}$$

(2)

As was presented in Figure 8, absolute maximum biases of heat production, temperature ratio and pressure ratio, were less than 0.6%; the maximum bias of resultants, such as H₂O, reached 1.8%. In addition, Positive or negative number presented relative size. Ultimate pressure and temperature, as well as production of H₂O of the two reduced mechanisms were greater than of initial mechanism. The main way to produce heat was transformation between CO and CO₂; the production of CO and CO₂ calculated by the two reduced mechanisms was smaller than by original mechanism, hence, heat production was little decreasing with the reduced reactions.

Except for heat production, the others kept in the same trend: biases declined with the growing temperature. Here explained, the second bias growth of heat production with increasing temperature was due to the occurrence of side reactions. Overall, the biases of the reduced mechanisms were limited to a reasonable and acceptable range. A reduced mechanism with high precision and fidelity was developed using reduction approach of tracking C&H reaction paths. According to the acceptable bias in practical engineering, this approach was performed to reduce detailed mechanism to a smaller size. The former detailed mechanisms were with hundreds or thousands of redundant elementary reactions and species. While, the smaller size of reduced mechanism was applied to engineering practice to predict the reaction resultants more concise, especially for the numerical simulation with huge computational cost. For example, calculation fluid dynamics (CFD) was commonly used for simulate combustion process, the supersized mechanism was an additional burden that extended beyond the capacity of calculation in most cases.

3.2. Experimental and Numerical Ignition Process

3.2.1. The Validation of Simulation Method

The development process of ignition was conducted using CFD and high speed camera including kernel generation, growth and propagation. Figure 9 showed the ignition process within 6 ms and the self-luminous flame structure was photographed in the observation window with a spatial resolution slightly lower than 5 million pixels and a shooting frame rate of 1000 frames per second. It presented the consistence of simulation and experimental results, at the same time, it was also a good verification of simulation accuracy. The results clearly revealed the detailed development processes. As shown in Figure 8, in the first millisecond, the spark discharged, the flame kernel generated and it surrounded the igniter. Due to the aerodynamic conditions, flame kernel was bent out of shape. In the second millisecond, flame kernel was in an irregular shape attached on the step where the velocity zone was low. It was conductive to flame stabilization and propagation. In the third and fourth milliseconds, flame grew rapidly and ‘climbed’ along the back face of strut. It also included the breakup of flame surface, then the break-up flamelet propagated to further zone and ignited the unburned mixed gas. In the fifth and sixth milliseconds, flame was fully developed and filled the whole burner. As shown, the motion trail of flame was regular and it was related to aerodynamic conditions.

To better understand the development process of ignition, flow pattern and flame structure were presented in Figure 10, the influence of aerodynamics was complicated. Firstly, once the flame kernel generated, its windward surface was deformed by the inlet flow. Then distortion, folding and curling occurred to aggravate the heat release rate, at the same time, severe deformation would cause energy dissipation even blow-out. Secondly, there was a recirculation zone behind the radial stabilizer, and it was irregularly distributed. As shown, heat and mass transfer happened in the recirculation zone, which provided momentum transport for the radial flame propagation. Thirdly, little recirculation zones of pilot zone were different from main recirculation zone but connected with it. Based on the little recirculation zones, heat and mass transfer occurred more easily, which was conducted to flame propagation and stabilization in pilot zone.

3.2.2. The Predictions of Ignition Process

Figure 11 showed the influence of velocity (50 m/s, 100 m/s and 150 m/s) on ignition process using numerical simulation. As shown, firstly, once the spark discharged and flame kernel generated, the size of flame kernel was decreasing with the growth of velocity, and the deformation of flame surface was increasing. It was due to the smaller flux at lower velocity, the one-time discharged spark energy could ignited more combustible gas and generated bigger flame kernel. Secondly, the time of complete flame propagation was longer with the growth of velocity. It was related to propagation path, the energy of flame kernel at 50 m/s enough to meet the requirements of igniting surroundings, making flame kernel to propagate along multiple directions and fill the complete burner as quickly as possible; however, flame kernels at 150 m/s and 200 m/s generated and were in irregular shapes attached on the step where the velocity zone was low, then flame kernel grew rapidly and ‘climbed’ along the back face of strut to realize filling the whole burner. Thirdly, in the propagation process, the flame might get broken up and the broken flame fell off to ignite further unburned gas, and the separated burning zone reconnected to complete flame propagation.

3.3. Outlet Temperature Distribution

Figure 12 shows the outlet temperature distribution of the combination of step and strut, the simulation results are calculated using the numerical method introduced in Section 2.3, and the experimental results of outlet temperature are tested with B type thermocouples, including the inlet temperatures of 750 K, 900 K and 1050 K, respectively. As shown, the outlet temperature distributions of simulations were higher than it of experiments, the bias of simulations and experiments were within 5%. It illustrated that the novel numerical method worked effectively and achieved high precision. Also, the growth rate of outlet temperature was decreasing with the increase of inlet temperature. Chemical characteristics of Kerosene/air premixed combustible had changed at high temperature. A mixture of n-decane and 1,2,4-trimethybenzene (8:2, by weight), called the Aachen surrogate, was selected for consideration as a possible surrogate of kerosene. Side reactions occurred to n-decane above 900 K, which led to a converse effect on flame spreading.

This result shows a unified phenomenon that the outlet temperature of central region was lower than the other regions. Once the recirculation zone ignited, the flame spread along the expansion structure, and the volumetric enthalpy of combustion is diluted, which causes local temperature to decrease. And it shows that the high-temperature region is concentrated in the toroidal surface, it proves that the rationality of combined design.

3.4. Outlet Components Distribution

Figure 13 shows the outlet components distribution of the combination of step and strut, the simulation results are also calculated using the numerical method introduced in Section 2.3, and the experimental results are tested with flue gas analyser, including the inlet temperatures of CO₂, CO, CH₄ and O₂, respectively. The bias of simulations and experiments were within 3%. As shown, the component of O₂ was decreasing with the increase of inlet temperature, the reactions occurred in oxygen-lean condition, especially in inlet temperature the 1050 K. The main way to produce heat was transformation between CO and CO₂; the production of CO and CO₂ existed a turning point at 900 K in the growth of inlet temperature. Extra-high temperature oxidation aggravated the side reactions, the formation of CO₂ was impaired and the coproducts of CO and CH₄ were promoted. This trend proved that the growth rate of outlet temperature was decreasing with the increase of inlet temperature.

4. Conclusions

In this work, modelling, simulations and experiments were performed including mechanism reduction, ignition process and flame stabilization. The following conclusions were drawn:

(1)

A simplified mechanism was developed and verified using a reduction approach according to tracking C&H reaction paths, especially by tracking key elementary reactions more clearly such as correlate heat release. Compared with the sensitivity analysis method, the element tracing method based on probability density distribution proposed in this paper has stronger adaptability and higher accuracy. Mathematical equilibrium model and calculation were developed based on Gibbs principle of minimum free energy. And this simplified mechanism with half the size, but got similar accuracy and fidelity with its error controlled within 1.8%.

(2)

This simplified mechanism was well adapted to ignition simulation and prediction under complicated aerodynamic conditions, and its simulation was always consistent with experimental results. Base on the validated simulation method, flame details including kernel generation, flamelet breakup and flame propagation, were depicted and analysed; and the ignition process was predicted under the wider variation of velocity. Also, it provides possibilities for predicting ignition performance prediction beyond the designed boundary conditon.

(3)

The influence of inlet temperature on outlet temperature and component distribution was performed, the bias of experimental and numerical results was within 5%. The higher temperature accelerated side reactions, which caused the increasing coproducts of CO and CH₄, also led to the decrease of volumetric heat production.