Figure 39: Side view of Free Layers Triple Flame structure

In some cases, the observed flame did not exhibit a triple flame structure but instead exhibited a random shape (Figure 40). All flames that were successfully ignited though, had a sideways bell-shape with a curved front.

Figure 40: Free Layers random structure

Another notable characteristic of each flame was that they did not occupy the entire duct from front to back but rather spread mainly along either the front or back wall of the duct. Generally, the flame would ignite along the centerline but propagate towards the airfoil along a side wall of the duct. Figure 41 shows a top view of one of the flames.
As shown in the figure, the flame is propagating along the back wall of the duct and does not occupy the entire duct. This result suggests that the experiments performed to date with the free layers apparatus cannot be readily modeled with a 2-D combustion model. It should be noted that top views of the flames observed in the prior NASA Layers apparatus showed that these flames did in fact occupy the entire duct and were thus nearly two-dimensional.

Figure 41: Top view of flame propagating towards airfoil with side walls and duct centerline denoted.

The cause of the flames not spreading along the centerline as would be expected is most likely a fueling issue. The fuel seems to be distributed along the sides of the duct rather than the centerline. This is discussed in the future suggestions in Chapter 6.

5.2.6 Flame Spread Rates

Obtaining spread rates for flames propagating through a non-uniform, free layer fuel mixture was one of the main objectives of this thesis. Prior to the present work, studies had been performed to determine the properties of flame spread through layered mixtures. However, the characteristics of flame spread through a free layered mixtures (Figure 7) were unknown. The flame spread results for free layers are summarized below.

A total of 26 flame spread tests were conducted. Of those tests, 17 were successful in igniting a flame. From these results, spread rates for 10 of these cases were acquired. A spread rate is obtained by tracking the image sequence taken during each run. For a single run, the video was broken into frames. The time between each frame was 1/30 s. Each frame is then de-interlaced, resulting in video fields with a ?t of 1/60 s between each field. Figure 42 shows a typical video field sequence. In this sequence, the airfoil is on the left side of each image. The flow through the duct is from left to right. The igniter is positioned downstream, or to the left, in the images.

Figure 42: Video field sequence of flame spread. ?t=1/60 s between each image.

After the frames are de-interlaced, the number of pixels that the flame traveled between each field is noted. A scale factor recorded before each test is used to convert number of pixels into number of centimeters traversed between each field. Since the time elapsed between each field is known, the flame spread rate can be found. The data is then plotted in a spreadsheet to calculate the spread rate. Figure 43 shows a typical flame position vs. time plot. This specific plot was obtained by tracking the image sequence in Figure 42.

Figure 43: Flame spread rate plot corresponding to Figure 42.

A trend line is fit through the data points. If the data is linear, the slope of the equation fit to the points is the flame spread rate. The position vs. time data plotted in Figure 43 yielded a spread rate of 148.31 cm/s. It should be noted, however, that the flame spread rate calculated in this manner is the flame velocity in lab coordinates. Accounting for the air flowing toward the flame, which in this case is 37 cm/s at the inlet, the actual flame speed, in flame-fixed coordinates, is nearly 185 cm/s. While this is an estimate because the velocity of the air flow along the centerline behind the wing is slower than at the inlet, this result is similar to the results obtained in prior studies (180 cm/s). (Kaptein and Hermance, 1976; Feng, et al., 1975, Ishida, 1988 and Miller et al, 2002).

In some cases, the position vs. time data did not have a constant slope during the entire flame spread duration. In these cases, an approximation was made to calculate a flame spread rate. The non-linear line was broken into sections that were fairly linear. The slope of each individual section is found, and the average of those slopes is considered the spread rate.

Table 6 is a summary of flame spread rates found to date.

Table 6: Summary of flame spread rate results.

6 Summary and Conclusions

6.1 Summary of Work to Date

In summary, the main objectives of this study were to design an apparatus to setup a free-layer stream of a fuel-air mixture and to analyze characteristics of a flame propagating in such conditions. This research began with an initial design concept of such an apparatus to be used experimentally. The initial design consisted of a flow duct with a porous, fuel-emitting airfoil along the centerline of the duct. A NACA 0012 airfoil cross-section was chosen for its low-profile, symmetric shape which produced low drag and had a large surface area to emit fuel.

The development of a non-reacting CFD model of the proposed system followed. The CFD model was used to characterize the proposed geometry and to ensure that a flammable mixture could exist in the desired operating conditions. The model showed that it is possible to setup a flammable free-layered mixture within the duct but that the flammable region is typically only approximately 1 cm thick (Figure 13, Figure 16). This result was used to determine optimal ignition locations. Buoyancy effects were also examined using the CFD model. The model predicted that, at lower speeds, the fuel plume sinks rather than rises, due to the molecular weight of the fuel, ethanol in this case, being heavier than air (Figure 12).

While not a main goal in this thesis, stabilizing a stationary flame is a longer-term goal of the NASA Layers research study. Accordingly, modeling was done to predict a flame shape and stabilized location. The flow stabilization calculations were performed using a correlation for laminar flame speed in uniform mixtures from ethanol data along with the exported CFD results. The prediction showed a sideways bell-shaped flame, with unstable areas just above and below the centerline. The stationary location was roughly 8 cm past the tail of the airfoil (Figure 21).

The apparatus that was designed and built was a 78 cm x 10cm x 10cm flow duct. A porous bronze airfoil was mounted 10 cm from the inlet of the duct along the centerline. Ethanol was fed into the airfoil through side fueling ports. A stream of air was forced through the duct parallel to the airfoil to set up the free-layered mixture. The duct was designed with the ability to converge and diverge, a feature that would aid in obtaining a stationary flame in the duct.

Once the duct was built, cold flow testing was performed to characterize the fluid flow within the duct and to compare the measured duct characteristics with CFD model predictions. Cold flow testing included hot wire velocity measurements, temperature measurements and flow visualization. The results showed that the duct produces steady laminar flow and that the CFD modeling accurately predicts the temperature and velocity profiles measured in the duct.

After cold flow testing was completed ignition and combustion tests were performed. These tests started with fuel vapor profile visualization using a Rainbow Schlieren system, which agreed qualitatively with the earlier CFD modeling results and showed the fuel plume sinking (Figure 38). A total of 26 combustion tests were performed, producing 17 propagating flames. Under certain conditions, a triple flame structure formed, with a curved front, wins above and below the centerline, and a trailing flame “tail” along the centerline (Figure 39). The general structure of the all flames, though, incorporated sideways bell shape (Figure 40). The average flame spread rate obtained from tracking results from 10 different tests was 164 cm/s, with the fastest rate being 217 cm/s and the slowest being 134 cm/s (all in lab coordinates).

6.2 Conclusions and Suggestions

The experiments presented in this thesis have shown that it is possible to establish a free-layered fuel-air mixture. Most importantly, it is possible to successfully ignite a flame in such conditions. The average spread rate of 164 cm/s is comparable to results found in prior research, relative to laminar flame speeds for uniform mixtures.

However, there were some inconsistencies with the flame spread rate results. These results appear to be due to flaws in the experimental apparatus. The following will document some concerns regarding the current apparatus and include some suggestions for improvement.

One problem observed with the current apparatus is with the airfoil temperature. To date, the temperature of the airfoil was measured in two ways. The first measurement consisted of a thermocouple mounted in the side of the airfoil, roughly 1 inch deep. This thermocouple measures the internal temperature of the airfoil. The second temperature measurement is a manual scan across the airfoil with a separate thermocouple. The surface temperature is the most important because it gives a more accurate measurement of the fuel temperature, which is evaporating at the surface of the airfoil. Mounting a thermocouple permanently on the surface of the airfoil will make taking measurements easier, as well as providing an input into a temperature controller, which will be discussed next.

The next issue also pertains to the airfoil temperature. The procedure described in this thesis included manual control of the heaters. However, keeping a constant airfoil temperature proved to be difficult and tedious. Every small change in the flow rate of the fuel affected the temperature of the airfoil, resulting in more adjustments. Likewise, changes in flow velocity led to further manual adjustments of the heaters. Installing a temperature controller to automatically adjust the heaters and keep the surface temperature of the airfoil constant should make setup easier and reduce prep time.

The airfoil design itself appears to be the cause of the problem of non-uniform fuel distribution throughout the surface of the airfoil, which is ultimately observed experimentally via non-uniform surface temperature. As described in Chapter 4, the airfoil was machined with an internal fuel cavity. The initial design concept was developed to rely on capillary action to pull the fuel through the porous surface. However, the results showed that when the fuel was fed into the airfoil, the fuel would collect in a pool. Moreover, and in many cases, depending on airfoil temperature, the gravitational pressure gradient of the fuel would cause the fuel to drip through the bottom of the airfoil. If the temperature and flow rate were balanced perfectly, the fuel would diffuse through the top and bottom surfaces of the airfoil with no dripping. This condition was met for each test, though after long setup times. To attempt to correct this, sand was packed into the airfoil in an attempt to disperse the fuel throughout the cavity. However, only one test was performed to date with the sand in place. If possible, a new airfoil should be constructed.

To facilitate the possible redesign of the airfoil, new CFD modeling of the internal airfoil fluid mechanics is underway at NASA Glenn Research Center (Miller, Personal Communication 2003). Specifically, the effects of porosity, gravity and geometry on surface fuel delivery are being examined. The results show that, indeed, the bottom surface of the airfoil has a much higher fuel delivery rate than the top surface. Moreover, the results suggest that the overall fuel delivery rate may have been much lower in the experiments than the CFD modeling predictions presented in Chapter 3 of this thesis, which assumed constant surface ethanol mole fraction. The new CFD results suggest that the stratified fuel/air layer might have been much leaner than expected. This result would explain the ignition difficulties observed experimentally.

Another noticeable cause for concern was that the fuel, being fed into the sides of the airfoil, collected more near those as opposed to the center. This effect was most likely the cause of the flames spreading only along the walls (Figure 41). A new airfoil could again be porous bronze, but with no open cavity in the center. Instead, a fuel manifold could be installed into the airfoil. Such a design would cause the fuel to be distributed more evenly about the top and bottom surfaces as well as the full width of the airfoil.

As mentioned, the flames seemed to ignite closer to the walls of the duct rather than along the centerline. A possible fix for this would be a redesign of the duct. The duct is currently roughly square, due to the fact that the airfoil was designed prior to the research conducted in this thesis. If the airfoil and duct were designed to be more rectangular than square with a width much longer than its height, the wall interference would likely be less evident.

6.3 Future Work

The NASA Layers research project is ongoing at NASA Glenn Research Center and Rowan University. Now that the free layers apparatus has been designed and built, an extensive series of combustion and ignition tests can now be performed. With the minor modifications described above, the apparatus should be able to produce repeatable flame spread through a free stratified layer.

In conjunction with the experimental program, a numerical flame spread model will be developed to simulate the flame spread experiments. Such a model will be key in determining the mechanisms responsible for the elevated flame spread rates observed for stratified layers with respect to purely premixed flames. The mechanisms for elevated flame spread rate are not completely understood, but likely include aerodynamic and Lewis number effects.

In addition to the free layers apparatus, work continues on the “floor” layers apparatus described in Chapter 1. Experiments are being conducted and the numerical model is being refined to determine the flame structure and assess the effect of aerodynamic pressure on the flame spread rate.

Appendix A - Mass Diffusivity Calculation

The tubing diameter used to deliver ethanol to the airfoil was not arbitrarily picked. To ensure the proper flowrate (according to FLUENT results) was being fed into the airfoil, a calculation was done to determine the needed tube diameter size. This is explained with the equations below:

where: r = tube radius [m]

Q = flowrate of fuel [m³/s]

M = viscosity of ethanol [N-s/m²]

l = length of tubing [m]

P = change in pressure in tubing [Pa]

The flowrate of ethanol is obtained from the FLUENT model. Viscosity is taken from properties tables at room temperature. The length of the tubing was determined simply by experiment placement. Change in pressure was obtained from the following equation:

where: с= density of ethanol [kg/m³]

g = gravity [m/s²]

h = height of tubing [m]

P_atm = atmospheric pressure [Pa]

Appendix E - Hotwire Calibration

The TSI hotwire (Model 1210) used for taking flow velocity measurements needed to be calibrated before being used. This was accomplished by placing the hotwire at the outlet of a converging nozzle through which a known mass flow rate was flowing. The mass flow rate settings and corresponding hotwire readings were entered into an automatic spreadsheet which calculated the hotwire calibration curve.

The main concern with this technique was that the velocity profile of the air at the outlet of the nozzle was assumed to be flat. This, of course, is incorrect. The actual profile is dome-shaped with the highest velocity along the centerline. To correct for this velocity profile, FLUENT was used to model the outlet of the nozzle. Eight different cases were run at different flow velocities. The results were then output to a spreadsheet. The data in the spreadsheet was plotted for each case. The outlet profile for a 14.18 cm/s inlet velocity is shown in Figure 46.

Figure 46: Outlet velocity profile for nozzle used to calibrate hotwire.

For the eight cases ran, the ratio of the centerline velocity for each to the average outlet velocity for each case was plotted against the average outlet velocity. A trendline was fit to the resulting curve. This plot is shown in Figure 47.

Figure 47: Plot of velocity results of each nozzle modeling case ran to be used as a correction factor in final hotwire calibration.

The equation obtained from the trendline was used as a correction factor in the hotwire calibration spreadsheet. The final calibration for the hotwire is shown in Figure 48.

Figure 48: Final Calibration for hotwire relating hotwire voltage output to flow velocity.

Appendix F - Rotameter Calibration Curve

Figure 49: Rotameter calibration curve for ethanol