3.1.1 Contour Plots

3.1.1.1 Mole Fraction Contours

Contours of ethanol mole fraction give an understanding of fuel concentration throughout system, with the ultimate goal of determining the optimum location for locating the flame igniter. Figure 10 shows the mole fraction contour plot for the 40 cm/s, 0-g, 323 K airfoil temperature operating condition. This plot shows the fuel concentration coming off the airfoil.

Figure 10: Ethanol mole fraction contour plot for 40 cm/s, 0-g, 323 K.

The inlet air stream is flowing from left to right. The large area surrounding the airfoil and plume in the plot is where the fuel concentration is zero. The highest concentration shown is on the surface of the airfoil, as expected. The thickness of the plume varied with flow velocities. At lower velocities (10 cm/s and below), the plume was thicker (Figure 11), while with cases run from 20 cm/s and up, the plume was thinner, around 1 cm thick. This was an important result when positioning the igniter during the experiments.

The mole fraction contours were also used to show the effects of buoyancy in the model. From the FLUENT results, it was shown that the plume of fuel sinks under normal gravity conditions. One may think that because the airfoil and fuel are being heated, the fuel vapor coming off the airfoil would tend to rise. However, the molecular weight of ethanol (46 kg/kmol) is heavier than that of air (29 kg/kmol), which ultimately causes the plume of fuel to sink. This effect is noticed more so in the slower cases (< 30 cm/s) where buoyancy has a greater affect on the flow. This effect is shown in Figure 12.

Figure 11: Mole fraction contour plot for 10 cm/s, 0-g, 323 K.

Figure 12: Mole fraction contour plot for 10 cm/s, 1-g, 323 K, showing buoyancy effects.

3.1.1.2 Equivalence Ratio Contours

One of the main objectives for performing the CFD modeling is to find an optimal ignition location. Using the mole fraction predictions from FLUENT, it is possible to convert to equivalence ratio (See Appendix C) and create a custom field function to produce contour plots that are not default options within FLUENT. Equivalence ratio plots show the entire flammable region of the mixture and provide a general understanding of the optimal location in the duct to ignite a flame. Figure 13 shows a contour plot of equivalence ratio for the 40 cm/s, 0-g, 323 K case. The flammability limits for ethanol/air range from an equivalence ratio of approximately 0.5 (lean) to 2.5 (rich). As shown in Figure 13, the actual flammable region within the duct is fairly thin; however it is shown to extend past the third igniter port in the actual duct (~58 cm from the trailing edge of the airfoil). This will be more evident in the X-Y plots below.

Figure 13: Equivalence ratio contour plot for 40 cm/s, 0-g, 323 K.

3.1.1.3 Velocity Contours

Velocity contour plots provide information on the overall fluid flow field within the duct. The velocity contours also provide a means of checking the results to make sure model ran correctly, since analytical solutions exist for similar laminar flow situations such as entry length in a rectangular duct and flow over an airfoil.

Figure 14 shows a velocity contour plot for a 40 cm/s, 0-g, 323 K case. From this plot, one can see the inlet velocity (flowing from the left) is entering at 40 cm/s. The velocity goes to zero at the tip of the leading edge of the airfoil, indicating a stagnation point. The flow velocity increases above and below the airfoil, as would also be expected. Other points of interest in the plot include the velocity going to zero at the surface of the airfoil, boundary layers growing on the walls of the duct itself, and the velocity deficit along the centerline, downstream of the airfoil. In cases ran at a slower flow speed, the boundary layers are thicker. These are all indications that the code is working properly.

Figure 14: Velocity contour plot for 40 cm/s, 0-g, 323 K.

3.1.1.4 Temperature Contours

Temperature contours are another means of checking the model for proper results. The temperature contours should be virtually identical to the mole fraction contours because both fuel and heat diffuse similarly. Figure 15 shows a temperature profile for a 40 cm/s, 0-g, 323 K case. When compared to the mole fraction contours for this same case (Figure 10), they appear to mirror each other. When setting up the material properties (See Section 2.4.2), constant property assumptions for the thermal properties such as specific heat and thermal conductivity were made since the flow is non-reacting, incompressible and subsonic. The results are not affected by using constant assumptions compared to using the mixing law option that is also available during property setup.

Figure 15: Temperature Contours for 40 cm/s, 0-g, 323 K case.

3.1.2 X-Y Plots

3.1.2.1 Equivalence Ratio vs. Y-position

Plotting equivalence ratio (see Appendix C for related calculations) vs. y-position at various x-positions quantitatively shows the best location to ignite the fuel. It is assumed that the optimal location to ignite the fuel is where the equivalence ratio is equal to one. As mentioned above, the flammable region is very small, approximately 1 cm thick. Figure 16 shows an X-Y plot of equivalence ratio as a function of the y-position (height) in the duct, for three different x-locations in the duct. These three x-locations, which are shown in the figure inset, correspond to igniter port location in the experimental apparatus (to be discussed in chapter 4).

Figure 16: Equivalence Ratio vs. y-position in duct for 25 cm/s, 1g-Y, 323K case.

This graph is for a 25 cm/s, 1-g (-y direction), 323 K airfoil temperature case. On the graph, the lean flammability limit for ethanol is indicated by the red vertical line (equivalence ratio = 0.5). As shown in the figure, the flammability region within the duct spans only around 1 cm for a velocity of 25 cm/s and airfoil temperature of 323 K. This result does not leave much room for error when placing the igniter at an appropriate location for ignition in a roughly 10 cm high duct. At lower speeds, the plume would get slightly thicker, to around 1.75 cm, still fairly thin with respect to the duct height, and the flammability region would not span as far in the duct. There is not a significant difference between the plume thickness for a 25 cm/s case and a 40 cm/s case. Both plumes are close to 1 cm thick; however, the fuel concentration increases as the velocity increases. While higher speeds do not affect the plume as much, a higher airfoil temperature does. A FLUENT case run at 40 cm/s with an airfoil temperature of 338 K (as opposed to 323 K), was run. The results showed the maximum equivalence ratio (directly on the surface of the airfoil) was over 11, whereas the maximum equivalence ratio at the surface of the airfoil in the 323 K case was around 4. A difference in the maximum values of the line plots was also noted, though not as drastic as the overall maximum value. The 15 K increase in airfoil temperature raised the maximum equivalence ratio at igniter port #1 (~19 cm past the trailing edge of the airfoil), represented by the blue line on Figure 16, from 1.0 to 2.0, and the maximum at igniter port #3 (~75 cm past the trailing edge), represented by the yellow line on Figure 16, from 0.5 to 1.2. Even with the increase in temperature, the modeling results show that there still remains a flammable mixture at each igniter port, however the mixtures at all of these locations are now rich instead of lean.

3.1.2.2 Velocity Profile Line Plots

Velocity profiles viewed along different lines spanning from the top to the bottom of the duct at a given x-location are used to compare FLUENT results with experimental duct characterization results (to be discussed further in Chapter 5). Figure 17 shows a velocity profile along the y-position (height) of the duct at an x-location of 9.5 in. past the leading edge of the airfoil.

Figure 17: FLUENT velocity profile downstream of airfoil

From this plot, one can clearly see a velocity deficit along the centerline. This velocity defect is a direct effect of the airfoil. Other properties of the flow shown in this plot are a velocity increase above and below the centerline, which is expected due to the airfoil, and the boundary layers that are beginning to grow. In Chapter 5, the predicted velocity profiles are compared directly to experimental results using hot wire anemometry.

3.1.3 Surface Integrals

The mass flow rate of ethanol was checked at three different points past the airfoil in order to check conservation of fuel vapor. The conservation of mass requires that the integrated mass flow rate of any species across any line in the y-direction (downstream of the airfoil) should not vary with x-position in the duct. Table 4 shows the integrated mass flow rate results. For each of the nine cases shown, continuity was checked at three different places in the duct. If mass is being conserved, then the flow rate should be equal along each line, as is shown in the table.

Table 4: Ethanol mass flow rate integrals from FLUENT modeling.

40cm/s - 0g	40cm/s - 1gY	40cm/s - 1gX
X-distance	flowrate	X-distance	flowrate	X-distance	Flowrate
4.75	-0.000865879	4.5	-0.000862639	4.5	-0.000903073
11.5	-0.00086689	11.5	-0.000866273	11.5	-0.000907025
20.5	-0.000868406	20.5	-0.000866901	20.5	-0.000907025
20cm/s - 0g	20cm/s - 1gY	20cm/s - 1gX
X-distance	flowrate	X-distance	flowrate	X-distance	flowrate
4.5	-0.000588035	4.5	-0.000588519	4.5	-0.000682088
11.5	-0.000593267	11.5	-0.000592536	11.5	-0.000688725
20.5	-0.000594428	20.5	-0.000593827	20.5	-0.000689283
10cm/s - 0g	10cm/s - 1gY	10cm/s - 1gX
X-distance	flowrate	X-distance	flowrate	X-distance	flowrate
4.5	-0.000410027	4.5	-0.000416635	4.5	-0.000580319
11.5	-0.000415253	11.5	-0.000421993	11.5	-0.000586215
20.5	-0.000415521	20.5	-0.000422088	20.5	-0.000587346
x distance measured from front tip of airfoil in inches flowrate in kg/s-m

3.2 Predicted stationary flame shape/location

Stabilizing a stationary flame makes analyzing shape and structure as well as quantifying the fuel concentration around the flame easier. With the flame “sitting” in one place, more fuel concentration data can be taken and clearer images of the flame can be captured, instead of relying on a single video frame to quantify the properties of the flame at a certain point.

While stabilizing a stationary flame was not a main objective of this research, the new flow duct was designed with the ability to converge or diverge in order to aid in stabilizing a stationary flame. Some predictions were made as to where a stationary flame may stabilize within the duct. These predictions are presented below.

3.2.1 Laminar Flame Speed (Uniform mixtures)

A stationary flame will stabilize wherever the component of the flow velocity normal to the flame front is balanced by the propagating flame speed at a given point. With this theory, modeling results and calculations can be used to find a predicted stationary flame shape and location.

The first step in doing so was to develop a relationship between laminar flame speed (S_u), equivalence ratio (?), and temperature (T). By doing this, it would be possible to use the results obtained from FLUENT to obtain an estimated laminar flame speed value for any given point in the CFD grid. Egolfopoulos and Law provided ethanol data relating flame speed as a function of equivalence ratio at 4 different fuel temperatures (24^th Symposium on Combustion). A 3-D Gaussian curve was fit to this data using SigmaPlot 2000. Gaussian refers to the type of function that SigmaPlot fits to the curve. Lorentzian and Paraboloid curve fits were other options in SigmaPlot. However, the Gaussian gave the best fit with an R² value of .986 compared to .9315 (Paraboloid) and .978 (Lorentzian). The equation that is obtained from the curve fit will be used later to balance the flow velocity. The curve (Figure 18) and relation are shown below.

Figure 18: Curve fit of ethanol data from Egolfopoulos and Law (24^th Symposium on Combustion).

3.2.2 Predicted Stationary Flame Location/Shape

The second step was to see where the flow velocity and flame speeds were equal, leading to a stationary flame. This was not as intuitive as one would think. First, the flame speed has to be balanced by the component of the flow velocity that is normal to the flame front. Of course, the shape of the flame is unknown; therefore, a flame front had to be estimated, using the CFD grid as a guide.

Starting at the centerline of the duct, where the flow velocity is always normal to the flame front (assuming the flame is propagating along the centerline), the first point was found. This point represents where the difference between the laminar flame speed and the flow velocity (normal to the flame front) is zero. From this point, we move up in the y-direction to the next gridline in the GRD mesh. The angle at which a line connecting the centerline point to a random x-point in the same vicinity on the next y-gridline is calculated using simple geometry. This angle simulates the angle a flame front would make if between these two points. This step is repeated for all x-locations ±5 cm from where the centerline point is located, giving many “flame fronts.” The incoming flow velocity is resolved normal to each one of these estimated flame fronts as shown in Figure 19.

Размещено на http://www.allbest.ru/

Figure 19: Diagram of flow velocity components used to balance flame speed.

The U and V components of the flow velocity are obtained from FLUENT, and U_N (the component normal to the “flame front”) can be calculated through trigonometric relations shown below.

Finally, the difference between U_N and S_U (again, calculated using the curve fit and the FLUENT data), is plotted versus x-position. Where this line intersects the x-axis is the point at which the difference between U_N and S_U is zero, and in theory, a stationary flame can exist at that point for the given conditions. There are instances where the line may intersect the x-axis twice. In this case, one must check the flow velocity values around each intersecting point to determine whether or not a point is stable. For example, in a graph with two intersecting points, if the velocity field ahead of the first intersecting point is greater than that point, it would cause the flame at that point to blow downstream and stabilize at the second point. Likewise, if the velocity field behind the second point is greater than that second point, it would cause the flame at that point to blow forward. An example of one such plot is shown in Figure 20.

Figure 20: X-location where stationary flame will stabilize at 0.81cm above centerline

This entire process was repeated for five different y-locations, not including the centerline point. Figure 21 shows the results, giving a predicted stationary flame location and shape.

Figure 21: Graph of predicted stationary flame location and shape for 40 cm/s, 0-g, 323 K case.

The results shown in Fig. 21 correspond to conditions of 40 cm/s, 0-g, 323K case. One should note that there are sections above and below the centerline where this process predicts that a stationary flame will not stabilize due to flame speed being faster that the flow. It is important to note, however, that this was done using the calculated laminar flame speed for a uniform mixture, not layered, as this research is studying. As shown in prior research, flames have been known to travel up to 4.5 times faster in a stratified mixture than a uniform mixture. Even with this being the case, this prediction gives a first, rough estimate of what may be expected in the likes of flame shape and location.

3.3 Modeling Summary

Modeling the mixing properties of the flow with FLUENT proved very useful in design and initial testing of the experimental apparatus. The most significant aspects about the flow characteristics provided by FLUENT were the flammable region and optimal ignition locations. Knowing the proper velocity and airfoil temperature to produce a flammable layer will be useful when attempting to ignite a flame within the duct.

Another important property predicted by the FLUENT code was how buoyancy affects the flow field. As shown above, the plume tends to sink under normal gravity, low speed conditions. However, at higher speeds the air flow has a more pronounced effect on the fuel than buoyancy and forces the fuel toward the centerline. This result is also shown experimentally through smoke wire tests, which will be discussed further in Chapter 5.

Finally, predicting a flame shape and stationary flame location, though done using a laminar flame speed for uniform mixtures, gives an approximate means of what to expect in terms of flame structure characteristics. It also aided in designing the duct, as to locations of windows ports and such, as will be seen in Chapter 4.

4 Development of the Experimental Apparatus

In this chapter, the design and development of the new Free-Layers apparatus is described. The Free Layers hardware consists of an airfoil mounted along the centerline of a flow-duct. In this case, the fuel diffuses through the airfoil with airflow parallel to the airfoil, thus creating a stream of fuel vapor concentrated along the centerline (Figure 7). One main purpose of this configuration is to eliminate the contact between the flame and the cooler floor surface, reducing heat transfer effects and creating a free-stream flammable layer. Running the duct with this free-stream is also a means of simulating a fuel leak in microgravity and provides the possibility to stabilize a stationary flame. Originally, a cylinder was considered to emit the fuel in the duct. However, a cylinder with a large enough surface area to emit enough fuel would shed vortices and cause a significant disturbance in the flow. Figure 22 shows a side view schematic of the entire experimental apparatus to be discussed in detail below.

Размещено на http://www.allbest.ru/

Figure 22: Free Layers apparatus schematic.

4.1 Airfoil Style/Design

There were two different airfoils constructed at NASA Glenn for the Layers project. Both designs incorporated the same general characteristics such as a porous material through which the fuel can diffuse, and internal heaters to help the fuel evaporate quicker. However, the two airfoils differ by their shapes and construction. This is explained below.

4.1.1 NACA 0012

The first airfoil is a NACA 0012 cross-section. This specific cross-section was chosen for its large surface area through which fuel can evaporate. Also, its thin, symmetric shape gives low drag for minimal flow disturbance. The mid-section of the airfoil was machined from a solid piece of porous bronze. The outer surface was cut to NACA 0012 specifications using electrical discharge machining (EDM). A hollow section inside the bronze was cut to create a chamber for the fuel to accumulate. Two other holes were drilled into the sides of the airfoil for installation of the heaters. Aluminum end caps were cut to the same shape and glued on to the ends of the bronze. The end caps have appropriate holes drilled for the heaters, mounting screws, thermocouples, and fuel filler tubes. Figure 23 shows a cross-sectional view of the 3 in. long airfoil.

7.62 cm (3 in.)

Figure 23: NACA 0012 porous bronze airfoil cross-section schematic

4.1.2 Parallel Plate

The second airfoil is a custom design. It features a semi-cylindrical leading edge that joins two porous parallel plates. A triangular trailing edge is attached downstream of the parallel plates. As with the NACA 0012, the parallel plate airfoil has a hollow core that holds the fuel, and necessary holes for the heater, mounts, thermocouples, and fuel filler. The objective for constructing this airfoil was to have the option to switch to a different porosity by using interchangeable porous plates in the design. While this airfoil was available to use at the time of research, the experiments conducted concentrated solely on the NACA 0012 design, which is the shape that was modeled in FLUENT calculations described in Chapters 2 and 3, though the duct was designed to accommodate this airfoil as well..

4.2 Duct Design

While the airfoils are an integral part of the system, the flow-duct itself was subject to many design constraints and was actually designed around the existing airfoils. These airfoils had been designed to run in the original porous-plate gallery. However, as part of this research, converging and diverging ducts have also been designed with the objective of stabilizing a stationary flame downstream of the airfoil. The converging or diverging duct would result in a centerline flow velocity that varies with x-position and potentially result in an x-position at which the flame propagation velocity balances with the convection velocity.

Aside from the converging/diverging aspect, designing a separate flow duct specifically for the airfoil would allow more specific placement for desired instrumentation. This instrumentation includes ports for thermocouples, a hotwire anemometer, smoke wire, igniter at various locations along the top of the duct, and interferometer windows. In total, hotwire/thermocouple ports (1/8in NPT holes) were placed in 10 different locations along the top and sides of the duct, smoke wire holes (.0625 in. drilled holes with press-fit stainless steel tubing) in 8 different locations, and igniter ports (1/8in NPT holes x 2) in 3 locations along the top of the duct. Figure 24 shows a side view schematic of the duct with instrumentation locations.

Another important feature of the duct was the inclusion of screens and honeycomb at the inlet and outlet. The screens provide a pressure drop for the incoming flow and are an added safety feature, acting as a means of extinguishing any flame that might propagate to the inlet or outlet. Honeycomb is used as a flow straightener for the incoming stream of air, since a flow parallel to the airfoil is desired. Some other design goals taken into account were the ability to run both airfoils and a means of running experiments with the duct positioned horizontally as well as vertically (to reduce effects of buoyancy.

Figure 24: Duct side-view schematic showing instrumentation ports

The top, bottom, and two sides of the duct were all Lexan. To support the duct, two end flanges were constructed of aluminum. While providing support, these flanges also provided a means of mounting the duct both horizontally and vertically, as well as mounts for honeycomb and the air inducer. The basic dimensions of the duct are 10.16 cm x 10.8 cm x 78.74 cm (4in x 4.25in x 31in). These overall dimensions were kept the same as the old duct for continuity with the computer models that had previously been completed, as well as the fact that the airfoils had been designed for use in the old duct (which fixed the width of the new configuration). By fixing a pivot point close to the inlet of the duct, the top and bottom are able to slide up and down through slots milled into the sides of the duct, thus allowing the duct cross-sectional area to converge or diverge. The maximum and minimum convergence/divergence angle was designed to be two degrees in both directions. Because of the long length of the duct, an angle of only two degrees actually provides a 2.15 inch increase or decrease in the height of the outlet of the duct, or a change in outlet area by roughly 50% over the inlet area (Figure 25)

Figure 25: Movable duct wall position in maximum converging (top) and diverging (bottom) configurations.

4.3 Instrumentation and features

The completed apparatus is shown in Figure 26. The apparatus consists of a heated, porous airfoil, a Coanda air inducer, a flow duct and a variety of instrumentation. Each of these aspects will be described in the sections below with a detailed summary table following.

Figure 26: Completed free layers apparatus in test configuration.

4.3.1 Instrumentation

Two different thermocouples were used to characterize the duct. A type K exposed end, 0.020 inch sheath thermocouple used for temperature scans was mounted in a 5 cm travel length translation stage above the duct. The thermocouple was bent so that the end was positioned along an isotherm in the flow (across the duct). The bent length was 3 in. to minimize conduction errors. A type T, 0.020 inch sheath thermocouple used to measure the internal temperature of the airfoil was mounted through the side wall of the duct into the thermocouple access hole drilled into the aluminum end cap of the airfoil. The thermocouples were plugged into a thermocouple reader giving a display of the measured temperature. At each scan location as well as for various point along the surface of the airfoil, the high and low readings were averaged and recorded.

Chromel wire, 0.002 inch, was used for smoke wire. It was mounted through stainless steel tubes that were press fit in the walls of the duct. To maintain constant wire tension, one end of the wire was attached to a spring-steel tensioner to account for expansion of the wire when heated and keep it taut. Soldering flux paste was dabbed on the wire to produce the smoke. There were between 10-12 dabs approximately 1mm in size along the smoke wire. The wire was connected to a power supply that heated the wire and the paste, thus producing the smoke lines. A slide projector was used to produce a light sheet to illuminate the smoke lines.

For performing velocity measurements, a TSI Model 1210 hotwire with a 6 inch stem was mounted in an access port in the top of the flow duct. The hotwire was a constant temperature hotwire. A bridge controls the voltage across a wire connected to two posts at the end of the hotwire stem. The voltage changes accordingly with changes in air velocity in order to keep the wire at a constant temperature. Calibration of the hotwire was necessary before use to relate the voltage reading to a velocity. This process is detailed in Appendix E. Measurements were taken upstream of the airfoil at the duct inlet as well as downstream in the wake of the airfoil. The hotwire was mounted in the same translation stage used for the temperature scans. At each point along the scan line, a high and low voltage reading was recorded. These values were then averaged and related to velocity using the correlation chart obtained from the hotwire calibration.

Two separate color cameras were used to take images of the flame. A Panasonic GP-KR222 was mounted as a top view and a COHU Model 2222-2040 was mounted as a side view. Both cameras were positioned to capture images downstream of the airfoil where the flame would be propagating. Digital image data was captured using a VHS recording deck as well as a frame grabber installed in a PC.

To visualize the fuel concentration profiles, a Rainbow Schlieren system was used (Greenberg). This system uses the refraction of collimated light as it passes through the fuel mixture. The angle that the light deflects varies with the fuel concentration gradient. The light then passes through a rainbow filter and into a color camera. Depending on the amount of refraction, the light will pass through various colors of the filter, providing a colored image of the fuel vapor profile. Filter sizes of 900 ?m (total width) x 50 ?m (center width) and 1950 ?m x 50 ?m were used. An overhead schematic of the system is shown in Figure 27. A list of all instrumentation used in this study is included in Table 5.

Figure 27: Overhead diagram of Rainbow Schlieren system used to visualize fuel concentration profile.

Table 5: Summary of duct instrumentation devices

Instrument	Type	Use
Thermocouple	Type K, exposed end, .020 in. sheath	Temperature profile scans
Thermocouple	Type T, .020 in. sheath	Internal airfoil temperature measurements
Smoke wire	Chromel wire, .002in. (soldering flux paste to produce smoke)	Flow visualization
Hotwire Anemometer	TSI Model 1210, 6 in. stem length	Velocity profile scans
Rainbow Schlieren System	Filter sizes: 900 ?m x 50 ?m, 1950 ?m x 50 ?m, Camera lens: 1/30th s shutter 25 mm lens set to F/1.7 Mirror specs: 4" diam. x 450 mm focal length	Fuel concentration visualization
Camera	COHU Model 2222-2040 1/250th s shutter 9 mm lens set to F/2	Side view
Camera	Panasonic GP-KR222, 1/250th s shutter 9 mm lens set to F/2	Top view
Heaters	Watlow Firerod cartridge heater (x4), 0.125 in. diameter x 2 in. long, 100W/120V rating	Heat airfoil
Air inducer	McMaster-Carr 5571K9	Pull airflow through duct
Rotameter	Key instruments model GS8000	Control fuel flow into airfoil
Igniter	Kanthal wire, .0142 in	Ignite flame
Slide Projector	Kodak	Light source for smoke tests
Frame grabber	Epix PIXCI-SV4	Capture and digitize images
Software	XCap v2.2	Capture and digitize images
Camera	COHU B/W Model 6500	Capture smoke images

4.3.2 Airfoil Internal Heaters

A total of four Watlow Firerod cartridge heaters were installed in the airfoil. Two heaters were installed into the leading edge, and two were installed into the trailing edge (Figure 28). The heaters were 0.125 inches in diameter x 2 inches in length. Each heater had a rating of 100W and 120V. The four heaters were wired in parallel and 24W of power was applied on average (approx. 30V and 0.8A). On later tests, the heaters were operated at 44V in order to increase the fuel temperature to increase the gas phase mole fraction of fuel at the airfoil surface. Each of the four heaters had a resistance of 145?, giving the parallel set of heater an equivalence resistance of 36.25?. Figure 28 is a diagram of the airfoil, which shows the locations of the cartridge heaters, and Figure 29 shows the actual airfoil with installed instrumentation.

Figure 28: Top view diagram of airfoil showing internal heater locations.

Figure 29: Airfoil showing installed instrumentation.

4.3.3 Air inducer

A Coanda Air Inducer is used to pull the flow of air through the duct. The inducer itself is attached to a converging section which is attached to the outlet of the duct. The Coanda Air Inducer is shown in Figure 30. A small volume of high pressure stream of air is fed into the inducer which in turn pulls a high volume of low pressure stream of air through the duct. Nitrogen is used for the high pressure stream into the Coanda. As an added safety feature, a separate valve stemming from the Coanda inlet line is placed at the duct inlet. This is the primary means of extinguishing flames that are ignited in the duct. A stream of nitrogen is injected straight into the duct, thereby extinguishing the flame.

Figure 30: Coanda air inducer

4.3.4 Fueling System

The fuel delivery system on the apparatus is gravity-fed. It consists of a funnel reservoir suspended above the duct. The outlet of the funnel leads into a rotameter with an inline needle valve to control flow rate. The rotameter was calibrated for use with ethanol simply by flowing fuel through it, running the outlet into a graduated cylinder, and timing the fuel filling the graduated cylinder. The calibration curve is shown in Appendix F. Two important aspects of the fueling system occur after the outlet of the rotameter. First, the fuel line is split at a T-junction in order to deliver fuel to both sides of the airfoil. This gives a more uniform fuel distribution across the airfoil which was needed after a single fuel tube failed to provide the uniform distribution. Second, the tubing size is stepped down to deliver the required flow rate into the airfoil, as predicted by the FLUENT modeling. Appendix D shows calculations to determine the correct tube size.

4.3.5 Ignition System

Flame ignition is accomplished using a hot wire igniter. The hot wire is fashioned from Kanthal wire (0.0142 in. diameter), which is strung between two igniter posts. The posts are linked to a power supply that applies a voltage to the wire, in turn heating the wire and ultimately igniting a flame in the presumed flammable region that the igniter is placed.

4.4 Testing Conditions

The initial proposed testing conditions include a combination of duct configurations, positions, flow speed variations and the use of different fuels such as ethanol, methanol, and propanol, as well as interchanging airfoils.

As stated above, the duct was designed to incorporate the ability to operate in a straight, converging, or diverging configuration in order to tailor the velocity profile downstream of the airfoil. Another aspect of the design is the ability to position the duct vertically, with the air flowing through the duct perpendicular to the table it is mounted on. The vertical position will likely be necessary to reduce buoyancy effects on the flow.

The research conducted for this thesis concentrated on operating the duct solely in the straight configuration and horizontal position using the NACA 0012 style airfoil. The inlet velocity varied between 25 and 70 cm/s. Ethanol was the only fuel used for this stage of the project. The fuel temperature was controlled by the heaters within the airfoil. The surface temperature of the airfoil was initially maintained at an average temperature of 50єC. The inlet flow stream was at room temperature (~23єC) and the duct was open to atmospheric pressure.

5 Experimental results

5.1 Cold Flow Tests

Before any combustion tests were attempted in the new duct, cold flow tests were performed. These tests included initial duct calibration, velocity scans, temperature scans, and smoke wire tests. The cold flow tests were important in order to characterize the flow through the duct to make sure the flow through the duct was similar to what was predicted by the numerical simulation. The tests were also conducted to ensure that the flow within the duct was steady and laminar. Maintaining strictly laminar flow within the duct is a key aspect to the free layers apparatus to achieve the desired objective of studying flame propagation through a consistent, stratified layer.

While there was no set test matrix for the cold flow testing as was the case in combustion runs described below, the majority of the tests were conducted at a mid-range inlet velocity (approx. 25 cm/s). For several of the cold flow tests, the inlet velocity and airfoil temperature were varied slightly from the baseline conditions.

5.1.1 Duct Calibration

Before any profile scans could take place, the duct needed to be “calibrated.” Calibration of the duct was necessary since the inlet velocity was controlled by inputting a given high pressure stream of nitrogen into the Coanda air inducer and the only means of monitoring the input flow was by monitoring the pressure gauge on the nitrogen bottle. Therefore, the nitrogen pressure value had to be correlated with the duct inlet velocity. To perform the calibration, the inlet velocity was measured using a hotwire anemometer. The probe was placed 0.5 inches from the inlet of the duct along the centerline (2.0 inches from the top of the duct). The pressure into the Coanda air inducer was varied and data were taken in increments of 5 psig from 10 to 120 psig. Figure 31 shows the duct's final calibration curve. This curve relates Coanda inlet pressure to flow velocity. The variation in inlet velocity with Coanda pressure was linear, with the range of 10 to 120 psig resulting in velocities between 7.9 and 68.4 cm/s.

Figure 31: Duct velocity calibration.

5.1.2 Velocity Scans

The first tests in duct characterization, aside from inlet flow calibration, were velocity scans. Velocity measurements were performed using the TSI hotwire anemometer. Data were taken at two locations upstream of the airfoil and one location downstream of the airfoil. Depending on the case, data were either taken every one or two millimeters along the scan line. At x-positions closer to the duct inlet, velocity measurements were taken every 2 millimeters since the velocity profile was expected to be rather flat across the height of the duct at this position. For x-positions downstream of the airfoil where a more detailed profile was expected due to effects on the flow caused by the airfoil, the hot wire data were acquired every millimeter. The profiles obtained from the hotwire velocity scans were then compared to predicted velocity profiles from FLUENT.

Velocity scans near the inlet of the duct showed some interesting results. Figure 32 shows a velocity scan performed 0.5 inches from the inlet at an inlet velocity of 25 cm/s (line with error bars) compared with FLUENT results for the same location (line with triangle markers). The experimental results show a spatial variation of the inlet velocity. The vertical line on the graph represents the average of the measured data points. This line agrees well with the FLUENT results. The variation is an effect of the honeycomb used to straighten the flow. The peaks and valleys correspond to the 0.25 inch cell size of the honeycomb. Aside from the variation, one noticeable difference is the higher velocities along the walls in the experimental results. This is most likely due to open gaps where the honeycomb is not contacting the upper and lower portions of the inlet. Another probability could be small leaks near the top and bottom of the duct that would cause a jetting effect similar to the results of the scan. Again, this was not a cause for much concern due to the fact that the flame is expected to propagate along the centerline of the duct.

Figure 32: Velocity profile 0.5 inches past inlet (experimental and predicted) for 25 cm/s, 323 K case.

Figure 33 shows the experimental results of a velocity scan 3 inches past the inlet (nearly 1 inch upstream of airfoil) at 25 cm/s. As can be seen in this plot, the velocity has leveled off compared to the results shown in Figure 32. The effects of the honeycomb are not seen this far past the inlet.

Figure 33: Velocity profile 3 inches past inlet for 25 cm/s, 323 K case.

In Figure 34, a typical velocity profile downstream of the airfoil obtained experimentally is shown (line with error bars), along with a comparison plot of the modeling results at the same conditions (line with square markers). The figure shows that there is a generally good agreement between the modeling and experimental results. The experimental results show very good symmetry. Another notable aspect of the experimental results is the velocity deficit that is clearly shown along the centerline. As mentioned in the discussion of CFD results in Chapter 3, this is a direct effect of the airfoil.

Figure 34: Velocity profile downstream of airfoil (experimental and predicted) for 25 cm/s, 323 K case.

One main difference between the measured and predicted velocity profiles can be seen in the boundary layers at the top and bottom of the duct, which were measured to be thicker in experiments as compared to modeling predictions. This result is most likely a consequence of 3D effects from the front and back walls of the duct. Recall from Chapters 2 and 3 that the FLUENT simulation was performed by modeling the duct as 2D. Specifically, the model assumes an infinitely wide channel through which the air flows. The width of the actual duct is finite (10 cm) and, in reality, the front and back walls also produce a boundary layer, which is not accounted for in the model. One way to reduce the effects of the boundary layers would be to use a duct with a larger width. The 3D effects, however, should not cause a major problem because the experimental zone of interest for the combustion experiments is concentrated along the centerline of the duct.

5.1.3 Temperature Scans

Another cold flow test performed in order to characterize the duct was a temperature scan behind the heated airfoil. For the temperature scans, data were acquired every millimeter in the y-direction since the scans were performed at x-locations down stream of the airfoil were a high level of detail was necessary. Figure 35 shows both experimental (line with diamond markers) and modeling (line with square markers) results. The graph shows a few differences between the two results. First, the temperature peaks do not match. Though the difference is actually only 1 K, it's still notable. FLUENT, however, assumes the entire surface of the airfoil is a constant temperature. This is not exactly the case for the actual experiment though, because the heaters are installed inside the airfoil and the temperature of the airfoil was measured internally. The most noticeable difference is the asymmetric shape of the experimental results. This result was most likely due to a small leak near the top of the duct, which caused a jet-like stream of air into the duct, altering the profile. In the event of a leak, the duct was completely resealed by adding sealing putty along the joints of the walls before the combustion tests began.

Figure 35: Temperature profile downstream of airfoil (experimental and predicted) for 25 cm/s, 323 K case.

5.1.4

5.1.5 Smoke Wire

The final cold flow tests that were done were smoke wire tests. These tests were important because they provided a means of visualizing the flow and ensuring the quality of the flow was smooth and laminar as desired. A turbulent flow would lead to mixing the fuel and surrounding air, thus altering the desired layered profile. For these tests, the duct was on its side with a light sheet perpendicular to the airfoil and a camera mounted overtop of the duct. A mirror was placed on the side of the duct opposite the light source to reflect the light back to the airfoil. A schematic of the setup is shown in Figure 36 and a sample of one of the smoke wire images is show below in Figure 37.

Figure 36: Top-view schematic of smoke wire testing setup

Figure 37: Smoke wire test at 37 cm/s

In Figure 37, the location of the airfoil is indicated by the scaling line which is drawn along the centerline of the airfoil from its leading to trailing edge. The flow direction in Figure 37 is from left to right. The dark triangular-shaped area below the airfoil is a shadow produced by the airfoil where the reflected light could not reach. Note the smooth flow around the airfoil and how the smoke lines converge coming off the tail of the airfoil, as would be expected. This image was taken at a flow of 37 cm/s. The tests were done up to 70 cm/s and were all laminar. Asymmetry caused by buoyancy effects was seen a speeds of 25 cm/s and slower. However, at higher speeds, the flow velocity overcame the effects of buoyancy resulting in a high degree of symmetry about the duct centerline as indicated by Figure 37.

5.2 Combustion Tests

5.2.1 Test Matrix

Initial plans for the combustion experiments were to perform a combustion test matrix that was identical to the simulation matrix detailed in Chapter 3. However, achieving repeatable ignition at these conditions turned out to be more difficult than expected. Therefore, operating conditions such as heater temperature, flow velocity, and fuel flow rate were all varied in order to achieve ignition. The majority of the combustion tests were run at an inlet velocity of 40 cm/s and an average external heater temperature of 55єC. The fuel flow rate needed to gain consistent ignition, as predicted by FLUENT, was 4 mL/min.

5.2.2 Experimental Procedure

The following procedure was used to perform the ignition tests. First, the voltage supply was turned on to supply current to the airfoil heaters. Next, the igniter was set to the desired height. The internal and average surface temperature of the airfoil was recorded. The internal temperature was measured with the thermocouple mounted through the side of the airfoil. The surface temperature was measured at three different points along the width of the airfoil by manually placing a thermocouple on the airfoil's surface. These measurements were averaged and recorded. When the airfoil reached the desired temperature of 50єC, the fuel flow valve was opened thereby supplying fuel to the airfoil. It was important to keep checking on the temperature reading after the fuel valve was opened because the evaporation of the fuel through the airfoil led to a drop in temperature by approximately 10єC. When the surface of the airfoil is wetted, the air flow through the duct is turned on. At this point, adjustments of fuel flow and/or heater temperature may be necessary to keep the airfoil wet, but not dripping, after airflow is turned on. After the adjustments are made, the honeycomb was placed in the inlet flange and the screen was clipped into place. A closed valve supplying nitrogen to extinguish the flame is positioned near the front of the inlet. Making sure the igniter switch was turned off, the igniter power supply is turned on and set to 10 V. Once these preparations are complete, the lights were turned off and the video recording started. When recording began, the igniter switch was turned on. The switch was turned off immediately after ignition occurred. If the ignition was successful, the nitrogen valve near the inlet of the duct was opened to extinguish the flame.

5.2.3 Fuel Vapor Profile

Before attempting to ignite the fuel/air mixture, the fuel vapor profile downstream of the airfoil was visualized to qualitatively assess the thickness and symmetry of the free, stratified fuel/air layer. To achieve this, a Rainbow Schlieren system was used. As detailed in Chapter 4, this system uses the deflection of a light ray as it passes through the fuel mixture to visualize the mixture. Figure 38 shows an image produced by the Schlieren system. The airfoil is to the left of the image as denoted by the outline.

Figure 38: Rainbow Schlieren test at 40 cm/s

This image was captured for an inlet velocity of 40 cm/s and airfoil surface temperature an average of 55єC. It shows the fuel concentration gradient that is present in the flow. This is why the centerline coming off the tail of the airfoil is the same shade at the top and bottom of the image, because there is no gradient in these zones. While these fuel concentration results were not quantified, the images provided a qualitative understanding of how the actual fuel was behaving in the flow. The flow appears laminar based on the well-defined plume. The Schlieren results also show that the plume appears to sink. As previously shown, the FLUENT model predicted this as well. This result is because the fuel (ethanol) has a higher molecular weight than air. Under these conditions, the plume sank roughly .011 mm for every millimeter in the x-direction. This was helpful, along with FLUENT results, to determine optimal igniter position.

5.2.4 Flame Ignition

Igniting a flame was successful in the first few attempts. However, subsequent ignitions proved to be difficult. The conditions from the first test were matched. Many different igniter positions were tried. These positions were just below the centerline where the plume was sinking, along the centerline, as well as above the centerline. The hotwire igniter was even positioned diagonally top to bottom, but to no avail. After many failed attempts, ignition with a lighted match was attempted. A sealed hotwire tap in one of the side walls was opened, and a lit matched was placed in the flow. This was tried a few time, and successfully only once. The flame ignited when the match was close to a side wall, in an area where the igniter wire did not reach.

While troubleshooting the ignition problems, the thought occurred that the first ignition was done with a single fuel filler tube installed. However, the improvements made on the apparatus shortly after the first ignition, namely the second (smaller) fuel filler line, resulted in insufficient fuel supply into the airfoil. This result was discovered by calibrating the flow meter for use with ethanol. After the calibration, it was found that the flow rate that had previously been used was nearly half of what FLUENT had predicted. After increasing the flow rate and thus increasing the airfoil temperature to evaporate the higher amount of fuel and eliminate more dripping through the airfoil, flames were consistently ignited.

5.2.5 Flame Structure

The structure of the flame was very similar to Phillips' triple flame (see Figure 1). However, as discussed in Chapter 1, in the Phillips experiment the fuel concentration varied from rich to lean along the height of the duct resulting in half of the flame to be a rich premixed flame, and the other half a lean premixed flame. In the present study, the fuel was concentrated along the center of the duct, with air above and below, presumably equally diffusing above and below. Since the fuel concentration was not quantified, the properties of the three branches of the flame have yet to be determined. Specifically, it is not known which branches are lean or rich. When conditions were optimal, the flame took the shape of a triple flame, with a wing above and below, as well as a trailing branch along middle, as shown in the side view in Figure 39.