1 Introduction

Accurate and simultaneous measurement of multiple scalars within an engine combustion field is key to a better understanding of the fundamental flow physics and more importantly to provide instructions on engine design aiming for high efficiency and low emission. Modern combustion diagnostics can be roughly divided into two categories: intrusive and non-intrusive methods. The intrusive method typically uses probes, thermocouples, or gas sensors to diagnose combustion flow properties, such as velocity, temperature, species concentrations and so on. Apparent drawbacks of intrusive method include inevitable interferences to flowfield and limited response rate. In contrast, the latter method allows one to effectively circumvent those shortcomings of intrusive methods and realize measurements with relatively large test range, small flow disturbance, high spatial and temporal resolution, fast response and so on [1]. Many kinds of laser-based and optic methods have been proven to be reliable and useful tools for probing combustion mechanism, such as Rayleigh scattering [2,3,4], spontaneous Raman scattering (SRS) [5,6,7], laser-induced fluorescence [8,9,10], chemiluminescence [11,12,13,14], particle image velocimetry [15,16,17] and so on.

Among all these techniques, Raman scattering is highly desirable due to its capability of measuring multiple scalars simultaneously and instantaneously in complex flows. It has been widely used in the study of turbulent jet flame [6, 18], internal combustion, supersonic combustion [19] and so on. Compared with Rayleigh scattering and laser-induced fluorescence technique, the signal of spontaneous Raman scattering is rather weak, making it vulnerable to all kinds of interference [20]. However, the use of spectrometer with image intensifier and careful tuning of the optical path can effectively improve the signal-to-noise ratio (SNR) and realize the measurement of both temperature and major species concentrations in the combustion field [21, 22]. A number of Raman-based investigations in model combustors operated at elevated pressure and preheated air have been reported [22,23,24]. Wehr et al. performed 1D Raman scattering measurements in a gas turbine model combustor fueled by nature gas with operating pressure up to 10 bar [25]. The overall combustor geometry was relatively simple, consisting of a rectangular cross section mounted with optical windows on the sides. Strong turbulence–chemistry interaction and large deviation from chemical equilibrium was found in the study. Hicks et al. conducted research work of point SRS measurements in a Jet-A fueled axial flow gas turbine combustor with pressure up to 55 bar [26]. The temperature within the combustor chamber was derived according to the ratio of stokes to anti-stokes Raman signals. Tedder et al. further measured temperature and major species concentrations in a model combustor produced by a 9-point lean direct injector fueled with JP-8 [27]. Through spatially traversing the laser beam and collection optics, a 2D mapping of the average temperature distribution on both meridian and lateral cross sections was obtained.

To apply SRS measurements in aeroengine combustion, one needs to overcome several practical challenges. First, a certain amount of laser pulse energy should be delivered to ensure a sufficiently high SNR. However, the undesired optical breakdown of air and/or dust particles sets an upper limit on the pulse energy that can be used, especially under high pressure and liquid kerosene combustion conditions [28]. Second, exclusion of stray light from reflection of combustor shell, optical window and all other sources needs dedicated optical path design since it is within a confined and complicated geometry with limited size. Third, kerosene combustion often brings unburned carbon deposition on optical window glass, which will seriously affect the collection efficiency of Raman signal. Therefore, it is necessary to realize quick replacement of the optical glass and ensure the good seal of the combustor. When it comes to the reverse-flow combustor studied in this work, aforementioned challenges could be even more severe, partially due to the S-shaped main flow pattern.

Reverse-flow combustor has been widely used in moderate-thrust aeroengines which are typically equipped on unmanned aerial vehicle and helicopters. Compared with axial flow combustor, the reverse-flow combustor has a unique S-shape overall flow pattern. Currently, the understanding about the internal flow field is quite limited. On the other hand, though Raman technique has been successfully applied on open flame and combustion chamber with simple geometry, its application in reverse-flow aeroengine combustor has not been reported yet, to the authors’ best knowledge. Therefore, this work attempts to explore and demonstrate the feasibility of SRS measurement on this complicated geometry model combustor. In this work, therefore, we performed the SRS measurement within a space-confined, double-layered, and reverse-flow model combustor fueled with liquid Kerosene. The primary objective and contribution was to demonstrate and validate the feasibility of spontaneous Raman measurement of multiple scalars, i.e., temperature and major species concentrations in such challenging environment. In the rest of this paper, Sect. 2 will introduce experimental setup of model combustor, laser delivery system, and Raman signal collection. The temperature and species concentration variation and dependence on pressure and probe location will be discussed in Sect. 3, and Sect. 4 concludes this work.

2 Experimental setup

The experimental setup in this work consists of three main parts, namely, reverse-flow combustion test rig, laser deliver system, and the Raman signal collection system. Due to the special geometry of reverse-flow combustor, practical concerns about the system design and implementation are also discussed in detail.

2.1 Reverse-flow combustion test rig

The central part of the combustion test rig was a single sector reverse-flow model combustor, as illustrated in Fig. 1. An obvious feature of the reverse-flow combustor differing from axis flow combustor was that compressed air flows into the reaction zone from the rear end of combustion chamber, while the burnt gas first flows forward and then backward to form a circulation reflux. Such a reverse-flow configuration allows a shorter axial length of the combustor to be used and it quite suitable for propulsion system of relatively small aircraft. The combustor mainly consisted of outer shell and inner liner, which are schematically shown in Fig. 1. As shown in Fig. 1a, the outer shell in this work can be separated into the left and right casing parts, and an upper cover, which were designed like this to allow an easy replacement or mounting of the damaged components. The outer casing wall was chosen to be 5 mm stainless steel, making it resistant to a moderately elevated operating pressure. The left casing had an air inlet flange, a thermocouple and a pressure sensor interface to monitor the air inlet temperature and pressure of the combustion chamber. The upper cover had two optical windows designed for laser entering and an ignition nozzle interface. The laser window was about 100 mm higher than the upper cover to reduce the potential laser damage of the optical glass by departing further away from the focal point of laser beam. Two circular observation windows with diameter of 35 mm were arranged on both sides of combustor outer shell to allow Raman signal collection and simultaneous video monitoring of the combustion status within the combustor.

Fig. 1
figure 1

a Overall view of the reverse-flow combustor, b illustrate of flame chamber and probe point locations, c illustrate of three thermocouples used for monitoring outlet temperature of the flame chamber

Full size image

Similar to the outer shell, both sides of inner liner were also equipped with circular optical windows with diameter of 35 mm, as shown in Fig. 1b. As kerosene combustion can lead to unburned hydrocarbon deposition on the glass surfaces, the flange structure of the circular observation windows made it easy to remove and replace the window glasses. The upper and lower liners of the flame chamber were drilled with mixing and cooling holes. The maximum diameter was 5 mm, and the minimum was around 0.5 mm. Part of the fresh air entered the flame chamber from the swirler and the primary holes, while the other part entered from the cooling and the mixing holes, which interacted with the flow field of the combustor together, making flowfield sustainable and stable. A coordinate system was established as shown in Fig. 1b. The center of swirler was taken as the origin of coordinates, the primary direction of the airflow inside the combustor was the X axis, and the vertical upward direction was the Z axis. Raman scattering signals were measured for 10 spatial points by adjusting the position of probe laser beam, and their spatial position coordinates are shown in Table 1. The measurement contains only 5 points at meridian plane to ensure both spatial resolution, spectral resolution and signal intensity. For the resolution concern, the probe laser line needs to be thin, otherwise the spatial resolution cannot be well defined, and the spectral overlapping for different species would be even worse. According to SRS theory, the Raman signal intensity is rather weak, even with laser energy of ~ 100 mJ. We needed to adopt the pixel binning and created the so-called super-pixel to enhance spectral intensity. However, more data points and real line measurement is possible for clean environment and open flames, as already demonstrated in many literatures. There were three thermocouples placed at the outlet of combustion chamber to monitor the burnt gas temperature, as shown in Fig. 1c. Primary goal of these thermocouples was to allow an approximate but direct monitoring of combustion state within the chamber, avoiding overheating of the chamber, and potentially severe mechanical damage.

Table 1 Spatial coordinates of the measurement points
Full size table

The corresponding air source system was able to feed continuous airflow at 0.6 kg/s mass flow rate with pressure up to 1.3 MPa and temperature of 500 K. In this work, we measured the Raman scattering signals of reverse-flow combustor under varying pressure, air mass flow rate and fuel/air ratio conditions, as summarized in Table 2. A normal pressure of one atm and an elevated pressure of five atm were examined, and the mass flow rate was correspondingly changed. Three different fuel/air ratio cases were studied to examine the feasibility of Raman measurement in different fuel/air conditions. The hot exhaust gas flew into a large volume tank with cold water sprayed to cool the gas and then ventilated.

Table 2 Experimental conditions
Full size table

As can be seen from Fig. 1 and discussed above, the geometry of the model combustor was relatively complicated, and it introduced many practical challenges to the laser measurements. Unlike some studies simplified the combustion chamber into a single layer structure [29], however, we would like to stick with such shell-in-shell configuration due to three considerations. First, the double-shell structure can best simulate the real combustion environment within the engine, including both the geometric and aerodynamic boundary condition. Second, the exploration of SRS measurement in such complicated situation promotes the potential applications in more realistic aeroengine in the future. Third, the data acquired from this work can provide some insights for the practical combustor design.

2.2 Laser delivery system

The excitation laser was a frequency doubled solid state Nd:Yag laser with maximum available pulse energy up to 1.4 J/pulse at 10 Hz repetition rate. The overall view of the laser measurement system is depicted in Fig. 2, with the combustor and signal collection system also integrated. The initial beam diameter of the laser was ~ 13 mm and original pulse duration was 9 ns. The laser beam diameter was expanded to 18 mm using a telescope lens group (L1 and L2) with a convex lens (f = 150 mm) and a concave lens (f = − 200 mm) [30]. The beam expansion was needed here to reduce the potential damage of optics when illuminated by high-energy laser pulse. Moreover, in order to avoid frequent optical breakdown in the focal region, which may cause permanent damage to the CCD camera and the image intensifier and the increase of allowable energy that can be used, the laser pulses were stretched by a two-leg optical ring stretcher, as shown in Fig. 2. It is noted that only one pulse stretcher was depicted in Fig. 2 for simplicity. The resulted laser pulse duration was expanded from 9 ns to about 100 ns [31]. Using two 5:5 beam splitters (BS1) and corresponding high reflectivity mirrors, the peak power reduced to about 25% of the original pulse. After that, an additional 1:9 beam splitter (BS2) was used to divide the laser energy one more time, with 10% energy used for real-time energy monitoring by a Coherent energy meter and 90% used for Raman measurement. A second telescope lens group (L3 and L4) was used to collimate the laser beam one more time and reduced the beam diameter to 9 mm.

Fig. 2
figure 2

Schematic diagram of Raman measurement system of a single sector reverse-flow model combustor

Full size image

The probe laser beam was refocused by a convex lens (L5 with f = 500 mm) and then delivered into the test section of the combustion chamber from top through the main holes on the inner liner. No additional aperture was drilled on the flame chamber in this work. The nominal laser beam diameter at the probe point was 0.5 mm. This choice of probe beam diameter was determined by a compromise of two contradictory factors, dielectric breakdown at the focal point and at the optical window of the top cover. Moreover, the reverse-flow configuration of combustor does not allow the bottom of the combustion chamber to be made transparent, therefore obvious laser ablation can be observed when high pulse energy used during experiment. Considering all these limiting factors, the actual pulse energy delivered to the probe point was around ~ 100 mJ, as indicated by the energy meter.

2.3 Signal collection system

The right part of Fig. 2 illustrates the Raman signal collection optics. To ensure the highest collection efficiency, the Raman emission was collected and collimated at 90° by an achromat (L6), and then focused using a second achromat (L7) to the entrance slit of the spectrometer [18]. The focal length of the first achromat (L6) was chosen as f = 75 mm, 100 mm, and 125 mm, respectively, depending on the Y coordinate of the probe point, while the focal length of the second achromat (L7) was f = 150 mm. The switch of different focal length lens of the first achromatic was to accommodate the different probe locations and maintained a minimum movement of other devices. For example, a probe point with Y = 0 would require the focal length of L6 being 100 mm, which was also the distance from the point to the observation window. When the focal length of the first collecting lens was set, the actual f-numbers of the lenses were about 1.4, 2.0, and 2.4, respectively, according to the limited size of the observation window. A homemade black paper tube was used between each collection optics to minimize stray light interference into the detector. To reject elastically scattered laser light, three notch filters centered at 532 nm, with two OD6 and one OD2, were placed at different places of the signal acquisition optical path. Such separated arrangement of filters was aimed to avoid possible reflection between consecutively aligned filters. Princeton Isoplane Spectrograph SCT320 spectrometer equipped with intensified CCD camera was used to recording the signal. The entrance slit width was 0.5 mm, and a pixel binning strategy of 25 pixels along laser propagation direction was adopted. With all these optics and corresponding setup, a typical spatial resolution of ~ 0.5 mm was finally reached.

This experimental setup introduced above has been aligned, redesigned, and tested for several times before a relatively stable configuration was realized. The key problem was due to a compromise between maintaining a near-real combustion setup and rather low signal level of Raman measurements. Nevertheless, through finely tuning these devices, we were able to obtain quantitative scalar measurements within the model combustor, as will be discussed in the following.

3 Results and discussion

In this section, the accuracy of the Raman system was first validated through controlled experiment and comparison against thermocouple measurement, and then the measurement results of temperature and major species will be discussed in detail.

3.1 Validation of Raman system

The accuracy of the SRS method was validated by comparing temperature measured by SRS and thermocouple using controlled experiment under known environment. When the cooling tank was not mounted and at 500 mm downstream of the combustor exit, a thermocouple and probe laser were closely arranged with ~ 2 mm apart from each other in the flow direction, as shown in Fig. 3. Both preheated airflow case and open-outlet combustion case were examined. For the latter case, three different fuel/air ratios of 0.014, 0.021, and 0.028 were used. In spectrum data processing, the pure airflow with atmospheric thermodynamic state was used for calibration of the Raman system optical coefficients. To reduce the interference of the background luminosity from the flame, the background light was collected immediately after Raman measurement at the same combustion condition but without laser present. The group of background spectrum was averaged and then subtracted from the Raman spectra. Further background was removed through a 4th-order adaptive baseline fitting algorithm. The signal left was then accumulated for vibrational Raman peaks of major species. In this work, we used the total number density of the major species and the known pressure condition to derive the temperature. There are several methods which can be treated as standard approaches for calculating temperature and species concentration using Raman spectrum. The readers are referred to literatures [20, 22, 32] for more information.

Fig. 3
figure 3

Arrangement of SRS and thermocouple at 500 mm downstream of the combustor to measure temperature synchronously

Full size image

The validation results are shown in Fig. 4. Figure 4a shows the temperature variation in 10 s of preheated air at 450 K measured by both thermocouple and SRS. It is noteworthy that the result of both thermocouple and SRS is about 390 K, caused by heat loss along a pipe of 2.5 m. Besides, dynamic variation can be observed from Raman measurement, while the temperature of the thermocouple remains at the same level. Such difference was attributed to different response time of the two methos. The stretched pulse width of the probe laser was about 100 ns and thus, the SRS can be considered as an instantaneous measurement. However, the response time of the thermocouple is one second, so the result measured by thermocouple is the mean temperature for one second. Nevertheless, the temperature thermocouple measured was on the same level as the mean temperature of the Raman measurements. In Fig. 4b, the averaged temperature measured by SRS was compared with that of the thermocouple for both non-reacting and reacting cases. The black dashed line indicates an ideal match, and the symbols show the Raman results. The error bar was estimated according to standard deviation of the dynamic variation in ten seconds. As can be seen that the two measuring result are very close. The best fit occurs in the case with preheated air, and the agreement of hot exhaust gas is not so good as preheated air case but generally acceptable. The maximum deviation of mean temperature was about 12% for the fuel/air ratio of 0.028 case. Such deviation was mainly because of the impure Raman spectrum of liquid fuel combustion environment.

Fig. 4
figure 4

a Instantaneous temperature variation of preheated air measured by SRS and thermocouple, b comparison of mean temperature measured by SRS and thermocouple for both preheated air and flame cases, c species concentrations of air measured by SRS in calibration experiments

Full size image

As for the species concentration, we measured the composition of the preheated air to verify the accuracy of SRS measurement method. The measured concentrations of seven major species are shown in Fig. 4c. It can be seen that only N2, O2, and H2O are found existing since it corresponds to the air composition that were measured, while concentrations of other species such as CO2 and CO were negligible. The mean molar fractions of major species are marked by the black dotted line in Fig. 4c, and those three sensible species are 0.775, 0.195 and 0.055 respectively, which agreed well to air temperature and humidity conditions. According to the air condition when performing the validation experiment, the temperature was 22 °C and humidity was 90%. Such air condition implies a molar fraction of N2, O2, and H2O as 0.77, 0.20 and 0.03. The molar fraction of the air would stay unchanged when it was heated to a higher temperature. According to the air condition during the experiment, the relative error of N2 concentration was 0.65%. The measurement uncertainty of spontaneous Raman technique comes from both theoretical and practical aspects, such as overlapping of spectral intervals of different species, temperature dependence of Raman cross sections, background luminosity and so on. For much detailed and thorough uncertainty analysis of Raman measurement, readers are referred to several literatures [5, 20, 22]. Many of those error sources are commonly encountered, here we explain three special sources which can interfere our Raman measurements. (1) In both reactive and non-reactive case, strong reflection of laser from the inner walls of the combustor would add wide-band noise into the Raman spectrum and reduce the SNR. (2) In reactive flow, fuelled with kerosene, the unvaporized fuel droplets or unburned hydrocarbon increase the possibility of laser-induced breakdown or incandescence, which severely reduces the spectrum purity. (3) In reactive flow, the strong background luminosity could interfere the spectrum and further reduce the SNR.

According to Fig. 4 and analysis above, we believe that our SRS measurement system is capable to yield a reasonable temperature and concentration results. It should be noted that the internal combustion field of the reverse-flow combustor is even harsher, and the accuracy of measurement would deteriorate. This is due to many Raman related effects. For instance, the temperature within the combustion chamber could be even higher, the signal collection efficiency is smaller due to confined space, and the laser energy can be used is even more limited. Moreover, the flow condition in the main reaction zone where our Raman measurement performed is much more complicated with strong turbulence-chemistry interaction and deviates from thermochemistry equilibrium, making the potential unburned hydrocarbons contaminate the Raman spectra. However, the use of SRS can consistently provide a direct and quantitatively measure of the combustion states within the double-shelled chamber, while some other intrusive methods is not capable to complete such work. Therefore, we stuck with the SRS method in this work and tried to present valid measurement results of inner reacting flow field, as will be discussed in the next section.

3.2 Temperature and species results

As stated above, the SRS signal was collected by a Princeton Isoplane Spectrometer with a single acquisition time of 100 ns and a sampling frequency of 10 Hz. The total length of a complete acquisition was 10 s, equivalent to 100 instantaneous snapshots of data during a test. Figure 5 shows the Raman spectra of the reverse-flow combustor in both non-reacting and reacting states, which reveals the dynamic variation of the SRS signal. The energy entering the measurement area was about 100 mJ to ensure a good signal-to-noise ratio. There may be occasional data loss at some moment due to camera overexposure caused by dust spark. These cases are caused by the ignition of dust, soot, or fuel droplets. For such useless data instants, we simply exclude them in data analysis. In addition, we note that such camera overexposure occurred more often in higher fuel/air ratio combustion case, because of the existence of unvaporized fuel droplets and unburned hydrocarbons.

Fig. 5
figure 5

a Raman spectra at 20 moments in non-reacting airflow, b Raman spectra at 20 moments in reacting flow

Full size image

Figure 6 shows the temporal variation of major species and derived temperature for fuel/air ratio of 0.028 and pressure of 1 atm. From the Fig. 6a we can see the concentrations of all the observed species fluctuated with time. Take N2 as an example, the highest concentration is close to 70% while the lowest is close to 40%, which results from the fluctuation of the flow field parameters to be measured in combustor. Figure 6b shows the dynamic variation of temperature in the flow field of the primary combustion zone. It can be seen that the temperature is around 850 K, and there is significant fluctuations indicating an intense turbulence in primary zone of combustor. For cases with other operating conditions, the measured temperature also varies with time.

Fig. 6
figure 6

Temporal variation of a species and b temperature for fuel/air ratio of 0.028 and pressure of 1 atm

Full size image

In addition to the dynamic variation of the temperature and species, Figs. 7, 8, 9 examine the effects of spatial resolution, varying pressure, and fuel/air ratio in terms of mean temperature. Figure 7 shows the variation trend of the mean temperature of five measured points on the meridian cross section at one atm, with three fuel/air ratio (0.014, 0.021, and 0.028). The plotted error bar was estimated based on the standard deviation during the test. Each point in Fig. 7 represents the average temperature of approximately 100 moments within the time interval of 10 s. As can be seen from Fig. 7, the corresponding average temperature of measuring point Z = 0 was about 900–1000 K at different fuel/air ratio. The maximum average temperature at Z = 0 was about 200 K higher than that at Z = 4 mm. This can be attributed to a fact that the measuring point is located near the main combustion hole, and a larger amount of fresh cool air entered the combustor and mixed with the hot burning gas near the swirler, reducing the temperature of the flow. On the contrary, at the position Z = 0, the distance to main combustion hole is further, and the influence of mixing air flow can be negligible, so that the temperature at the position Z = 0 is the highest. From Fig. 7, we can also see that when the fuel/air ratio is 0.028, the temperature of all measuring points was higher than 0.021 and 0.014, and the corresponding 0.021 was also slightly higher than 0.014, which indicates that the higher fuel/air ratio, and the higher temperature.

Fig. 7
figure 7

Mean temperature variation at different measuring points in the meridian plane under pressure of 1 atm

Full size image
Fig. 8
figure 8

Mean temperature variation with fuel/air ratio of flowfield in the main combustion zone

Full size image
Fig. 9
figure 9

Mean temperature variation at the measuring point 1 on the meridian plane under different pressure

Full size image

Figure 8 shows the average temperature of two measuring points on the meridian plane as a function of the fuel/air ratio. For kerosene combustion, the stoichiometric fuel/air ratio is about 0.068. The three fuel/air ratios used in this experiment have been shown in the Table 2. Therefore, the experiments in this work were all overall fuel lean. However, a part of the air entering the combustor was used for mixing and cooling, the actual amount of air involved in main combustion zone was much less. As a result, the three states in the primary combustion zone are close to fuel-rich, stoichiometric ratio, and fuel-lean states, according to the initial combustor design. Considering the real experimental condition and relating effects, the practical equivalence ratio could be different from the design. The spatial coordinates of the two measuring points were (43.9, − 2.0) and (43.9, − 0.0), respectively. As shown in Fig. 8, the average temperature was higher for higher fuel/air ratio at each measuring point. The average temperature at different fuel/air ratios can be different, which was also reflected by observation in the video monitoring during the test. When fuel/air ratio was 0.028, the luminous flame occupied the entire observation window, and the flame glowed red light. While fuel/air ratio of 0.014, the liquid fuel atomization was poor and flame could not fill the whole viewing window, and the swirling flame often exhibits a large area of blue color.

Figure 9 compares the dependence of mean temperature at the position of measuring point 1 on the meridian plane under different pressure (p = 1 atm and 5 atm). The average temperature of combustion under high pressure is higher than that under normal pressure, which can be increased by about 100–200 K. These results comply with visual and auditory observation. During the experiment, it was found that much more intensive fuel combustion occurred at five atm, demonstrated by the overexposed glowing radiation and roaring noise from the combustion chamber. Quantitatively, we attributed such temperature rise to a better mixing. With a five times higher the operating pressure, both the airflow rate and fuel flow rate are also 5 times higher than that of atmospheric condition. In this case, the swirler worked in a much better condition, and liquid fuel atomization promoted making spray cone penetrate deeper in the combustor. All these contributed to a better fuel/air mixing and a better combustion condition.

To illustrate comprehensively the dependence of measured temperature on spatial locations, pressure, and equivalence ratio, Fig. 10 shows the mean temperature variation at five probe locations on the meridian plane under different fuel/gas ratios. Some observations can be found from the Fig. 10, in good agreement with the previous analysis shown in Figs. 7, 8, 9. On the one hand, the average temperature was unevenly distributed in space. In general, it was the highest at the central point and relatively low at other points. Again, the reason is that the change of temperature in space reflects the structure of flow field in reflux combustion chamber. Because the measuring point is near the main combustion hole, a large amount of low temperature air enters the combustion chamber and mixes with the high-temperature burned gas near the cyclone, thus reducing the temperature of the airflow. The middle point with Z = 0 is far from the main combustion hole, and the penetration distance of the mixed air flow is limited, so the temperature at Z = 0 is the highest. Moreover, this point was designed to be within the recirculation zone of burned gas for flameholding purpose. On the other hand, the mean temperature of flame combustion is higher at high fuel/gas ratio than at low fuel/air ratio. This result is consistent with the phenomenon seen by video monitoring during the test. A special case is that when the fuel/air ratio was 0.014 at 5 atm, the average temperature decreased more significantly than that at 1 atm. We attributed the reason to possible malfunctioning of the atomization nozzle after a long time of operation. Soot generated by fuel pyrolysis could somehow block the nozzle. In this work, the measurement at one atm was carried out first. Therefore, at five atm, insufficient atomization and incomplete combustion may occasionally lead to smaller high-temperature area. Third, we can also see that the performance of average temperature is also different under different pressure conditions, which agrees with the results in Fig. 9. In general, the average temperature measured at 5 atm was higher than that at 1 atm.

Fig. 10
figure 10

Mean temperature variation at different measuring points in the meridian plane under different pressure and fuel/air ratio

Full size image

Besides the measurement on the meridian plane, \*MERGEFORMAT Fig. 11 illustrates the temporal variations of temperature and N2 mole fraction at probe point 1 of lateral cross section at different fuel/air ratio under operating pressure of 1 atm. As can be seen, the temperature and mole fraction varies dynamically with time, similar to results shown in \*MERGEFORMAT Fig. 6. However, the measured temperature level on the lateral points is apparently lower than on the points of meridian plane. For instance, the average temperature of 10 s with fuel/air ratio of 0.028 in \*MERGEFORMAT Fig. 11 was about 629 K, which is about 200 K less than the corresponding point 2 on meridian point shown in \*MERGEFORMAT Fig. 7. Such difference is attributed to the spatial location difference and the corresponding states of flow condition in the combustor. The point 1 on lateral plane is on the edge of the main reaction zone, while the point 2 on meridian is in the center part. Similar observations can be found from other corresponding points.

Fig. 11
figure 11

Variation of a temperature and b mole fraction of N2 for different fuel/air ratios under operation pressure of 1 atm

Full size image

Until this point, we have presented the temperature and concentration variation at different spatial locations, fuel/air ratio, and operating pressure. Considering we were measuring the main reaction zone state, the temperature we measured was relatively low, compared with a common impression of a hydrocarbon combustion field. To explain this observation and provide a side proof of our measurement, numerical simulation of the combustor was carried out for case 3 and case 5, with inlet air temperature of 500 K and fuel/air ratio of 0.028. The temperature distribution results are shown in the Fig. 12. A high-temperature zone is formed in the area near the nozzle of the flame chamber. The maximum temperature of the main combustion zone can reach about 2250 K at 1 atm, and about 2350 K at 5 atm. By comparison, it can be found that the temperature of primary combustion zone at 5 atm is higher than that of the low-pressure condition, and the range of the high temperature zone is relatively larger, which agrees well with our measurement. Meanwhile, under high-pressure conditions, the stronger the ability of head air flow to penetrate downstream, the greater the impact on the combustion area. With the increase of pressure, the airflow entering from the mixing hole has a greater influence on the flow field and outlet temperature. By comparing the two working conditions, it can be seen that due to the continuous flow of cool air into the combustor from the cooling hole, a low temperature zone near vertically penetrated from the wall surface into the main combustion zone is formed, which provide a good protective effect on the wall surface of the combustor. Finally, as shown from the diagram, the temperature of measuring area is about 800 K, and roughly agrees with the actual data obtained by SRS technology, providing a side explanation of our measured temperature.

Fig. 12
figure 12

RANS-averaged temperature distribution of the combustor under 1 and 5 atm obtained by numerical simulation

Full size image

It is realized that a more complete and thorough understanding of reaction zone flow field within the reverse-flow combustor needs to the measuring of more positions. Actually we are conducting further study on this model combustor, such as expanding the observation window on the sides, redesign of the top part of the flame chamber, and better control of experimental conditions.

4 Conclusions

In this experiment, we measured thermochemistry state at multiple points in the flow field within the main combustion zone of a reverse-flow model combustor using spontaneous Raman technique under different pressures and fuel/air ratios. Results of temperature and major species concentrations of multi-point varying in the space and with time were obtained. Based on systematic and comprehensive analysis of the existing experimental data, we conclude this work as follows:

  1. 1.

    Strong turbulence inside the reverse-flow combustor led to dynamic variations of as temperature and species concentrations. The maximum deviation of temperature was about ± 150 K.

  2. 2.

    The higher the fuel/air ratio, the higher the combustion temperature. When the fuel/air ratio was 0.028, the mean temperature can reach 950 K, while for fuel/air ration of 0.014, the mean temperature was about 100 K lower.

  3. 3.

    Pressure has a great influence on the flow field measured in the combustor. At five atm, the reaction of combustion was much more intensive and the flame temperature measured is also higher.

  4. 4.

    The temperature inside the combustor varied with the spatial position of the measured points. The temperature was the highest in the center of the combustor, and the temperature decreases when it was approaching the edge of the flame chamber due to the penetration of cooling air.

This work has made a preliminary exploration on the measurement of flow field within the reverse-flow combustor. There are still several aspects to be improved, such as the interference of unburned hydrocarbons Raman spectral in the flame, the weak Raman signal caused by small observation window, and the strong background resulted from the reflected light on the casing internal face. Further research work is under study to solve these issues.