Momentum and heat flux in a swirl-stabilized combustor

The use of a fine-wire compensated thermocouple probe and two-color laser anemometry to measure both heat and momentum fluxes in the axial and azimuthal directions is assessed for a complex flow, swirl-stabilized laboratory combustor. Thermocouple probe perturbation and time constant variation are evaluated and the former is found to be significant in the central region of the recirculation zone. To minimize the effect of perturbation, the configuration of the probe is varied. In the recirculating region, the mean temperature is uniform with peak, instantaneous temperatures approximating the maximum adiabatic flame temperature. The maximum rms temperature approaches 250°C and occurs off-axis, just downstream of the recirculation zone. The heat flux on the centerline is negative and becomes more negative with downstream distance. Correspondingly, the temperature time series and the probability density function of the fluctuating temperature indicate that neither is cool dilution air present near the centerline nor are hot fluid particles present near the wall. This is attributed to the preferential transport of hot, low velocity fluid particles towards the centerline and cool dense gas, towards the wall by the swirl.


Introduction
Turbulent, recirculating (i.e., complex) and elliptic flows are found in many practical engineering systems, such as gas turbines, furnaces, and boilers. An increased understanding of these flows requires detailed, time-resolved measurements of flow properties. The principal objective of the present study was to measure simultaneously time-resolved temperature and velocity fluctuations and, from the computed statistical quantities, provide a qualitative and quantitative description of the flow with respect to the mixing processes and turbulent transport in a model laboratory, swirl-stabilized combustor. An electronically compensated fine-wire thermocoupte probe was used to obtain the timeresolved temperature and a two-color laser anemometer was used to measure the two components of velocity.
Fine-wire thermocouple probes have been used *Presently at Energy and Environmental Research Corporation, Irvine, California, 92718. for more than two decades, i,'z,3 and have been recently combined with single-component laser anemometry to provide simultaneous measurements of both temperature and velocity in relatively simple flows. 4'5 The principal objections to fine-wire thermometry are durability, frequency response, and flow perturbation, each of which is exacerbated in a complex flow. In previous work, durability and frequency response, 6 and probe perturbation 7 have been assessed to establish the utility and limits of applicability of the use of fine-wire thermometry in a swirl-stabilized, complex flow combustor. The present study addresses the interpretation of the data set with respect to the mixing processes and turbulent transport.

Combustor Geometry
The model laboratory complex flow combustor ( Fig. 1) has an aerodynamically controlled, swirl- A set of swirl vanes (57 mm O.D.) is concentrically located within the combustor tube around a 19 mm O.D. centrally positioned fuel delivery tube. Dilution and swirl air are metered separately. The dilution air is introduced through flow straighteners in the outer annulus and the swirl air passes through 60 ~ swirl vanes. For a swirl-to-dilution ratio of unity, the value used in this study, the swirl number obtained by integrating across the swirl vanes is 0.8 while that obtained by integrating the total inlet mass flux is 0.3.
Propane is introduced at the end of the central fuel delivery tube through a cone-annular gas injector sized to emulate the directional momentum flux of a hollow-cone liquid spray nozzle. Data discussed herein corresponds to an overall equivalence ratio of 0.1 with a bulk reference velocity of 15 m/s. The combustor was operated at atmospheric pressure. Radial profiles were measured at four axial locations downstream of the exit plane: x = 2.0, 7.0, 14.0, and 24.0 cm where x is measured relative to the exit plane and the radial position, r, is measured relative to the centerline of the combustor.

Temperature Measurements
For temperature measurements in combusting flows, the sensor should be as small as possible to maximize uncompensated frequency response but must be large enough to survive the hot, oxidizing conditions of the flameJ '3 The latter criterion sets a lower limit on the wire diameter for the present case of 25 Ixm. Although 25 p,m diameter wires survived in many regions of the flow, larger diameters were required in the central region within and immediately downstream of the recirculation zone.
The thermocoupte junction was formed by overlapping and spotwelding a platinum and platinum/ 10% rhodium wire. The small diameter wires were then gas welded to larger support wires (of 250 txm or 500 i~m diameter) of the same material. The support wires were cemented in a 0.159 cm or 0.318 cm O.D. alumina tube which in turn was placed in Inconel tubing of various diameters and shapes. Various probe configurations, consisting of small and large straight and curved probes, were evaluated. Measurements are reported for the probe that minimized perturbation at a specific measurement location.7, 9 The probe was inserted into the flow through the exhaust plane of the combustor and held in a mechanical traverse which moved with the laser anemometer optical table. The thermocouple junction was positioned one mm downstream of the laser anemometer probe volume.
The thermocouple frequency response necessary for accurate time-resolved measurements in the wake region can be estimated from the physical dimensions of the combustor and the bulk mean velocity. If the length scale of the energy-containing structure is taken to be one-half the combustor radius, the corresponding frequency for a mean velocity of 15 m/s is about 750 Hz. In order to accurately measure the root-mean-square temperature, the sensor frequency response must extend to at least two or three times the frequency of the energycontaining structures or to approximately 2 kHz. The uncompensated frequency response of the 25 Ixm sensor at this mean velocity is about 15 Hz. Thus, the frequency response must be increased by a factor of 130 and is accomplished using electronic compensation.
The frequency response required for accurate measurement of the temperature near and in the recirculation zone is more difficult to estimate but certainly depends on the flame zone thickness and its angle relative to the probe. Ion probe measurements made in and at the edge of the recirculation zone indicate that the minimum flame zone thickness corresponds to a time duration of about 1 ms while the average duration is 5 to 10 ms. (The average is relatively high due to the oblique passage of flame zones past the probe and the low velocities in the recirculation zone.) Thus, in and at the edge of the recirculation zone, a compensated frequency response of 2 kHz should be adequate for the accurate measurement of time-resolved temperature.
The compensation method used in this study is similar to the one used by Lockwood and Moneib 3 and is described in detail by Seiler. 9 The basic function of the compensator is to perform the operation [1 + ~r(d/dt)]E to increase the low (~15 Hz) frequency response of the sensor. In that expression, ff is the mean time constant which is determined experimentally and in-situ at each measurement location, and E is the thermocouple voltage.
Briefly, the time constant is determined by electrically heating the thermocouple and, when the heating current is turned off, using the computer and appropriate software to monitor the tempera-ture decay. First an ensemble average is obtained for nine individual realizations of the thermocouple decay. Second, the averaged temperature decay is linearized by taking the natural logarithm of the ensemble averaged decay. Finally, the slope of the resultant curve, which corresponds to the time constant, is determined by the method of least squares. This process is repeated a minimum of six times so that at least 54 individual realizations of the thermocouple decay are used to obtain the mean time constant at each position. Repeated determinations of the mean time constant obtained in this manner vary by about -+ 10%. Also time constants obtained in this manner using AC or DC heating currents differ by no more than -+ 10%.

Velocity Measurements
Velocity measurements are made using a two-color laser anemometry (LA) system. The beam from a 200 mW Argon-ion laser (LEXEL Model 75) is collimated and passes through a prism to separate the various wavelengths. The blue beam (488 nm) and the green beam (514 nm) are the most intense and are each polarized and then split into two beams of equal intensity spaced 50 mm apart. A 40 MHz frequency shift (TSI Model 915 Bragg Cell) is applied to each pair of beams to eliminate directional ambiguity that otherwise results from the turbulent, recirculating flow.
The four beams are focused at a common point within the combustor. A set of perpendicular interference fringes, spaced at 2.6 I~m for the green beams and 2.5 Ixm for the blue beams, are oriented to yield the axial (u) and azimuthal (w) velocity components respectively.
Receiving optics consist of a 120 mm lens focused onto a 0.25 mm diameter photomultiplier tube aperture (via an appropriate dichromate filter to selectively pass either the blue or green light). These optics are placed at an angle of 20 ~ off direct forward scatter which results in a probe volume of 3 0.022 mm and a cross-sectional area perpendicular to the axis of measurement of 0.10 mm 2. However, due to the requirement imposed by the processing electronics that both axial and azimuthal velocity components be obtained simultaneously, the effective probe cross-section is much less (approximately 2 0.03 mm ). The transmitting and receiving optics are mounted on an optical bench capable of placing (accurate to -+0.3 mm) the measurement volume at points throughout the stationary combustor test section.
The air and fuel jet flows are seeded independently but to the same levels of concentration with 1 I~m alumina particles. A liquid suspension atomization seeding technique (Ikioka et. al) ~ is employed. Signal validation is obtained using two counter processors (Macrodyne Model 2098).

Data Acquisition
For simultaneous temperature and velocity data acquisition, a sample and hold op amp (AD 583) is used in conjunction with special electronics built to interface the output of the two channel digital counter processor channels with a DEC PDP 11/ 23 computer. The interface determines whether the two (u and w) velocity events occur within 50 I~S of each other which corresponds to a spatial resolution of 0.75 mm. 8 If so, the data are stored and multiplexed into the computer by means of a parallel interface. Once a u realization and a w realization are determined to have occurred simultaneously, the sample and hold is activated by the interface and the thermocouple signal corresponding to the simultaneous velocities is held. The thermocouple signal is then sampled, digitized and stored with the u,w pair along with the event time (t) relative to the initiation of the run cycle. The raw data sets consisting of u,w,t,T are stored for data reduction and analysis. A detailed discussion of the data acquisition system is available. 9A1

Uncertainty
Uncertainty in the velocity data has several sources including statistical convergence due to the finite amount of data, bias error due to particle-averaging versus time-averaging, digital resolution, repeatability of flow conditions, and probe volume positioning accuracy. Seiler 9 and Brum u have evaluated these uncertainties and shown that the error in the mean and rms velocity are a few percent. Velocity bias is avoided by using a low seeding rate and thereby operating the counter processors in the unsaturated mode with sample times more than an order of magnitude above the flow correlation time. lz (The absence of velocity bias is verified by analysis of the time-marked archived data base in uniform time steps of differing intervals.) The collection of 5000 data triplets (u,w,T) requires 30 to 90 minutes. (The time to collect 5000 axial velocity samples is about half that required to collect the data triplets.) A doubling of the sample rate, accomplished by increasing the particle generation rate, reduces the data collection time for 5000 samples by 45 percent for both the axial velocity alone as well as for the data triplets, thereby providing a verification that the data rate is proportional to seeding rate and not a result of misalignment of the laser sampling volumes.
The uncertainty in the temperature data depends as well on repeatability of flow conditions, positioning accuracy of the probe, and finite number of samples. Based on independent realizations of at least 5000 samples, the uncertainties in the mean and rms temperature, due to lack of statistical convergence and not including probe perturbation and time constant effects, are found to be ---10% and -+10 to -+30% respectively. The effect of probe perturbation and variation in time constant are discussed separately in the following section.

Results
First, a brief discussion of the probe perturbation study is presented where the perturbation caused by the use of probes of various shapes is reviewed. Second, the effect of time constant variation is discussed followed by a presentation of the temperature time series and the probability density function (PDF) of the temperature. The section concludes with the statistical properties of the combined velocity and temperature field. A tabulated data base, including measured inlet conditions, is available. 13

Probe Perturbation
The thermocouple probe, while inexpensive and relatively easy to fabricate, must be placed in the flow. Unfortunately, the presence of a physical probe in a recirculating or elliptic flow (cf., Reference 14) can cause local and global flow field perturbation.
Probe perturbation was assessed by separately placing various probe configurations in the flow and determining statistical properties of the veloci~ field in the presence and absence of the probe. 7'9 The probe perturbation effects are both small scale (local) and large scale (global). The former is associated with the size of the probe, especially near the probe tip, while the latter is related as well to the overall shape of the probe. Based on this perturbation study, probe configurations were selected to minimize the perturbation. In the central region, the perturbation is both local and global, and a curved probe is required to minimize the perturbation. In the outer regions, perturbation is limited to local effects and a straight probe is acceptable. In both cases, it is necessary to make the probe tip as small as possible.
Using probe sizes and shapes that minimize the perturbation, the perturbation as measured by laser anemometry was found to be significant but within reasonable limits for both the mean and rms velocities except within the recirculation zone itself. Perturbation effects outside the recirculation zone are less than 10% at all axial locations for r/R >-0.2, less than 20% at r/R = 0.1, and 30% on the centerline. At the edge of the recirculation zone, perturbation is less than 20%. Perturbation in the recirculation zone, however, led to differences exceeding 100%. As a result, the acquisition of timeresolved temperature measurements is precluded within the recirculation zone.

Time Constant Sensitivity
Another source of uncertainty is the amount of electronic compensation required to extend the frequency response of the fine-wire thermocouple probe. The correct amount of compensation depends on the physical properties of the gas and the instantaneous velocity and is therefore a fluctuating quantity. Thus, even when the sensor is accurately compensated for mean flow conditions, it will at times be over-or under-compensated. Variations in the time constant will affect statistical quantities associated with the temperature field (e.g., root-meansquare temperature) and introduce a phase shift in the temperature signal which will affect the value of the heat flux.
Time constant sensitivity was assessed by deliberately over-or under-compensating the sensor relative to the value obtained with the sensor compensated for the mean flow conditions. 6 The effect on mean temperature is not significant. For example, the measured value of the mean temperature varies by less than 10% for a 50% variation in the time constant. The root-mean-square temperature is also relatively insensitive to variation in the time constant. A 10% variation in time constant has less than a 5% effect on the measured root-meansquare temperature which suggests that over-and under-compensation affects that portion of the temperature power spectrum that contributes a relatively small amount to the root-mean-square temperature. However, a 10% under-compensation in the time constant yields a 20% variation in the measured axial heat flux and a 50% variation in the azimuthal heat flux. In contrast, over-compensation has relatively little effect on the value of the heat fluxes. This observation is consistent with that of Yanagi and Mimura. 4 It is clear that both phase shift and variations in the amplitude response that are correlated with the temperature and velocity affect the value of the heat fluxes, but the difference between the effect of over-and under-compensation is difficult to explain.

Temperature Time Series
Representative samples of the compensated temperature signal are presented in Fig. 2 at different radial positions for the 14 cm axial station. A peakto-peak temperature of about 500 ~ C is found at the centerline (Fig. 2a)  "energy" of the temperature signal at a frequency of 90 Hz. This corresponds to the frequency of longitudinal oscillation of the recirculation zone observed in high speed photography. 15 Near the wall (Fig. 2c), the temperature signal is dominated by high amplitude, short duration, temperature spikes which correspond to a positive skewness. The average rate of occurrence of hot particles crossing a threshold of 400~ is 25 Hz. Again, the crossings are not periodic. The minimum temperature corresponds to about 40 ~ C. The data obtained at the 24 cm station are similar, indicating that the sheath of dilution air near the wall of the combustor remains intact through the length of the combustor.

Probability Density Function of Temperature
Representative probability density functions (PDF) are presented in Fig. 3 for the axial location, x = 14.0 cm. The data are normalized by the maximum adiabatic temperature (Tad = 1996 ~ C). The PDF's show positive skewness at locations well displaced from this centerline (r/R >-0.6). This was obtained at all four axial stations and indicates that cool, unreacted gas predominates in this region of the flow throughout the length of the combustor. In addition, the PDF's are narrow in this outer region of the flow which corresponds to low rms temperatures and indicates that only a limited amount of fluid from the hot recirculation zone mixes with the cool dilution air. The PDF's are very broad at the intermediate radial locations of r/R = 0.4 and r/R = 0.3 with peak-to-peak temperatures on the order of 1100 ~ C, indicating the alternating presence of cool and hot and mixed fluid parcels. Maximum rms temperatures are also found at these ratemperature, which occurred less than 1% of time, are attributed to uncertainties associated with the effects of thermocouple radiation loss and catalysis, and time constant specification.
The temperature time series at the mid-radius position (Fig. 2b) indicates that the temperature signal has positive skewness due to the presence of relative short duration, high temperature peaks. The average rate of occurrence of hot particles (defined as the average rate at which the temperature signal crosses a threshold temperature) is 42 Hz for a threshold of 1000 ~ C. The crossings are not periodic but appear to be distributed randomly in time, and are evidence that a periodic structure is not associated with the passage of the high temperature fluid particles. Power spectral measurements of the temperature signal show a relative increase in the

Characteristics of the Velocity and Temperature Field
Velocity data obtained without the thermocouple probe in place are shown in Fig. 4a. The mean and rms temperature, the axial velocity, and the normalized axial heat flux are shown in Fig. 4b. The velocity data indicate that the statistical features of the flow field remain essentially unchanged from those found in the absence of a probe (Fig. 4a). The mean temperature, at all axial locations, has a maximum near or on the centerline. The maximum  (Fig. 2a) precludes the former. Hence, the negative sign corresponds to high temperature, low velocity fluid particles which, most likely, originate in the recirculation zone. The growth of the region of negative axial heat flux in the central region as the axial distance increases downstream is associated with the normal mixing and mean azimuthal velocity which preferentially transports low temperature, high density fluid away from the central region and high temperature, low density fluid toward the centerline. The low values of the axial heat flux at the midradius position corresponds to the fact that fluid is well mixed on this region. At the dilution air interface, the axial heat flux becomes negative because of the occasional presence of high temperature, low velocity particles from the central region of the flow. Evidence of these particles can be seen in Fig. 2c where fluid particles with temperatures as high as 800~ occur.
The mean azimuthal velocity at the 2.0 cm station (Fig. 4c) is relatively low in the recirculation region, with the peak mean azimuthal velocity, as expected, in line with the outer circumference of the swirl vanes. This sharp peak is quickly suppressed (x = 7.0 cm) by the interaction with the non-swirl dilution air. The azimuthal heat flux at the x = 14.0 and 24.0 cm stations exhibits relatively sharp, positive peaks at radial positions (r/R = 0.2 and 0.4 respectively) corresponding to the maximum values of the mean and rms azimuthal velocity.

Summary and Conclusions
Compensated fine-wire thermocouples can be used to measure time-resolved temperature along with axial and azimuthal velocity in a complex flow, swirlstabilized combustor but precautions are required, especially with respect to probe perturbation and survivability. Different probe configurations must be used in the various regions of the flow to minimize probe perturbation. The sensor size required to survive in the flow depends on the mean temperature and velocity intensity and must be increased in high temperature, highly turbulent regions. In the recirculation zone, the time-resolved measurements of temperature are impractical due to severe perturbation of the probe.
The mean temperature is nearly constant in the recirculation zone with instantaneous peaks approximating the maximum adiabatic flame temperature. Further downstream, mixing of hot products with unreacted swirl and dilution air takes place and a gradient in the mean temperature develops in the central region. Axial heat fluxes in the central region are negative, reflecting a large population of high temperature, low velocity fluid from the recirculation region. The azimuthal velocity is shcwn to increase this population by inducing a transport of low density, high/temperature fluid to the axis.
Please state the diameter of the thermocouple wire, the data rate of particles and their type and size. Did these particles influence the time constant of the thermocouple? With 15 Ixm wire thermocouples in premised flames and with both AlzO3 and T~O2 particles, particle size is an SMD of around 2 }xm, we found significant effects 'after 10 minutes with less than 60 particles/sec measured by the laser velocimeter control volume which had dimensions of around 1 mm • 0.15 mm.
Authors" Reply. For measurements in the wake region, the thermocouple junction was formed by spot welding 25 m diameter Pt-Pt/10%Rh wire. As a result, the length scale of the junction was about 1.4 times the wire diameter. Due to lack of probe durability in the recirculation zone, 125 m diameter Pt-Pt/10%Rh wire was required to measure mean temperature.
No evidence for seed particle deposition was found in the present experiment. Repeated measurement of the time constant at a reference position in the wake region of the flow, for example, resulted in measured variations of the time constant of less than +10%. These variations were random and a general trend of the variation, as would be expected if particle buildup had occurred, was not observed. Fi-nally, particle buildup was not evident in visual inspections of the wires. The absence of particle deposition in the present case is attributed to (1) a low particle seeding rate (10 to 1 particles/sec), and (2) suppressed local temperatures (as a result of the lean conditions) which are conducive to maintaining the alumina particles in a refactory state.