THE DEVELOPMENT OF A TUNABLE DIODE LASER ABSORPTION TOMOGRAPHY SYSTEM ON A DIRECT- CONNECT SUPERSONIC COMBUSTION WIND TUNNEL

Transcription

1 THE DEVELOPMENT OF A TUNABLE DIODE LASER ABSORPTION TOMOGRAPHY SYSTEM ON A DIRECT- CONNECT SUPERSONIC COMBUSTION WIND TUNNEL Erik N. Ellison, The University of Virginia Advisor: James McDaniel A Tunable Diode Laser Absorption Tomography (TDLAT) system been developed and is being tested on a supersonic wind tunnel at NASA Langley. TDLAT is a diagnostic technique that is designed to measure water vapor concentration and temperature at the exit plane of the wind tunnel. Prior to this project TDLAT had already been used on the smaller supersonic wind tunnel at the University of Virginia. The differences between the two wind tunnels resulted in several challenges. This paper explains how these challenges were addressed and presents the progress to-date, as of mid March as well as plans for the near future. The author s role in this project is mainly focused on the software system including: motion control, laser control, and data acquisition. The changes to these three components resulted in a system designed for a larger test section with a reduced runtime for implementation on the NASA tunnel, compared to the UVa tunnel. Introduction The American military and scientific community are both interested in the research and development of hypersonic technology. Hypersonic technology has several potential applications ranging from high speed, high energy missiles, to high speed passenger travel, to space access. Hypersonic missiles have several tactical advantages, mostly due to their elevated speed. Travelling at Mach 5 makes them very difficult to intercept, disable, or destroy once at cruise velocity. A defensive missile with the task of interception would have to be even faster in order to catch the hypersonic vehicle. Another notable advantage is the energy that is associated with such a weapon, not only could there be a warhead contained within the missile, but its immense velocity would also result in significant kinetic energy. On the other side, if this technology was applied on a passenger carrier, one could imagine a plane similar to the Concorde, except with the ability to fly twice as fast. In 2011 at the Paris Air Show EADS proposed a concept hypersonic plane that could carry 100 passengers and fly from London to Sydney, in 3 hours and 30 minutes 1. EADS has admitted that the technology for an aircraft with this speed and range is about four decades away, but this is why hypersonic research is important today 1. Hypersonic technologies could also be applied in Two-Stage-to-Orbit (TSTO) vehicles. The first stage would be a turbinebased combined cycle (TBCC) 2. This means that at low speeds a turbine would propel the vehicle, and would be followed by a transition to ramjet then scramjet power all in that stage 2. The second stage would be the rocket stage that would take the vehicle into orbit. Background Historically, the most common method of hypersonic propulsion has been the rocket Ellison 1

2 engine; however, it is not the most efficient. Hypersonic aircraft with air breathing propulsion systems have a few notable advantages over rocket based propulsion systems. For one, they are reusable; there is no need to build another rocket for each successive mission. Since they are air breathing, the oxidizer used to propel the vehicle, oxygen, is present in the atmosphere, eliminating the need for the vehicle to carry its own oxygen. Instead all that is required is the hydrogen or hydrocarbon fuel that will mix and combust with the oxygen. As a result, all this extra mass can translate into a larger payload or larger operational ranges. For flight Mach numbers above about 5 the engine operates in a supersonic mode, hence the name Scramjet. Methodology Tomography If the combustion efficiency in such aircraft can be measured accurately, steps can be taken in the design process to maximize the hydrogen combustion in the scramjet. The approach used was to measure the hydrogen converted into water post-combustion and divide this value by the total hydrogen fuel injected into the scramjet. The relationship is expressed by Equation 1: (1) where is the water density distribution and is the axial velocity distribution of the exhaust and N is the total hydrogen injected 3. The technique that was implemented to find the water density distribution is called Tunable Diode Laser Absorption Tomography (TDLAT). There are two parts to this technique: absorption spectroscopy and tomography. Absorption spectroscopy is governed by the Beer-Lambert law shown below, which relates the absorption of light through a certain material to the properties of that material: (2) where I is the transmitted intensity, I 0 is the incident intensity, L is the path length and is the spectral absorption coefficient 4. The spectral absorption coefficient must be determined as well, and the following equation is used: (3) where P is pressure, is the mole fraction of the water in the exhaust, is the transition line strength and is the lineshape function 4. Absorption spectroscopy results in an integrated line-of-sight measurement, so a number of different beams must be emitted in different positions and angles to be able to reconstruct the density and temperature distributions This reconstruction method is what is known as tomography. There are several beam geometries that could have been used. Two examples of such geometries are parallel beam and fan beam (see Figure 1 on the next page). The fan beam geometry was used in this research because it was easier to implement in this system as the laser could remain in one position, rotate its angle and collect data along several rays, before shifting a few degrees around the outside of the cross section of the exhaust. At this new vertex, the process is repeated until the laser has gone completely around the test section. Once the procedure is complete, various data interpretation techniques, some found in the MATLAB computing software, are used to reconstruct the water vapor density and temperature distribution 3. In the ambient atmosphere there Ellison 2

3 Figure 1: Absorption laser beam path geometries 4 is humidity, so in order to avoid the humidity influencing the results, the UVa Aerospace Research Lab injected nitrogen or dry air around the optics while testing to keep the ambient environment clean of humidity 5. The other part to a combustion efficiency measurement involves Stereoscopic Particle Imaging Velocimetry (SPIV). This part involves placing two cameras at the wind tunnel exhaust and injecting very small alumina particles to track the fuel and air flow 3. These small particles are illuminated by a laser sheet, which the cameras can detect and track to find the velocity of the exhaust flow 3. By measuring the water density distribution and the axial velocity of the flow, the combustion efficiency can be directly calculated, as seen in Equation 1. This portion of the research, although performed at UVa, is not part of the NASA Langley project. A significant advantage of TDLAT/SPIV is that they are nonintrusive. Other methods have been used to try to measure combustion efficiency but were not as effective. One technique involves using the first law of thermodynamics to compare actual heat release with the predicted maximum heat release 3. The benefit of TDLAT/SPIV is that the hydrogen flux is measured when they are injected and the water flux at the combustor exit is measured 3. The measurement of combustion efficiency is therefore made nonintrusively. This experiment has been tested at UVa with a flat flame using a Hencken burner as a proof of concept 5. Afterwards, the experiment was tested on UVa s dual mode combustion wind tunnel. UVa s wind tunnel has the special ability to run continuously, making it ideal for flow field diagnostics 3. As a result, tunnel run-time was not a limiting factor for the amount of time available for each TDLAT/SPIV experiment. Project Objectives and Challenges TDLAT has been performed on UVa s supersonic wind tunnel and as a result of a grant from NASA, the same technique is being used on a wind tunnel at NASA Langley. There are several notable differences between these two wind tunnels. Ellison 3

4 The Direct Connect Research wind tunnel at NASA Langley can only run about 30 seconds at a time, and it has a test section crosssectional area of 28.9 in 2 compared to the 1 in 2 cross-sectional area of the UVa wind tunnel. As a result, new hardware and software systems had to be created in order to accommodate the larger scale and shorter run time. The overall objectives guiding this project are listed as follows: 1. Create hardware set-up to use for NASA Langley testing. 2. Create software to use for NASA Langley testing. 3. Set-up and carry out experiment. 4. Analyze data. This paper focuses on the second and third objectives. Figure 2 shows the previous hardware installation used to find the density profile on a Hencken Burner or the UVa wind tunnel. The laser comes in through an optical cable, beams through the flame and then is reflected back to the tomography emitter detector (TED) box where it runs to the signal detector 3. The installation setup also has a purge that minimizes the beam s contact with water in the ambient air 4. Figure 2: Setup for tomographic scanning at tunnel exit 3 As seen in the diagram, the rotational stage is not much larger than the tunnel exit. With a larger exit area, a new, larger rotational stage was needed. In order to increase the data collection rate by a factor of 5, the new configuration uses 5 TED boxes (one for each of the beams). Figure 3 shows the new stage with improved TED boxes. Now instead of having to rotate 360 degrees to get a complete view of the cross-section the stage only needs to rotate 72 degrees (360/5). Figure 3: New set-up for tomographic scanning (McGovern) The software used to control the motion of the experiment, as well as the lasers and the data collection, had a number of changes that had to be made. On the one hand a new computer and new data acquisition cards were purchased to accommodate faster collection rates as well as an increase in data lines (5 instead of 1). The data cards (NI USB 9263) can now collect data at a rate of 100,000 samples per second. This data acquisition speed allows for sufficient data to be collected in 0.1 seconds rather than 1.0 seconds, decreasing the overall runtime. The rotation of the TED boxes was also made more efficient further reducing runtime. Overall the total time it takes to collect a full set of data has been reduced from about 58 minutes to 13 Ellison 4

5 minutes, if one TED box is used, or 2 minutes and 36 seconds when the new 5 TED box hardware configuration is used. Another change that was made to the software was the addition of a third laser scanning at a different wavelength in order to get a better understanding of the water concentration and water vapor temperatures. These three signals are split, and run either to a signal summer, that is used to distinguish the three signals during data analysis, or to a laser controller which converts the signals into a laser wavelength that is emitted through the test section. Each of these signals, as they are created by the computer, consist the same voltage ramping sweep from 2 to 6 volts. Since each of the lasers share the same fiber as they enter each TED box, they can each only be on for one third of the time and are actually only on for 27% of the time (this will be explained later). The next task was to configure the data collection cards. In the UVa experiment only three data collection lines were needed. The three were the laser signal (or current) generated by the LabVIEW program, the absorption data from the detector in the TED box, and an etalon (a reflective glass prism). In this experiment, a small portion of the laser signal is beamed through the etalon and is used to measure the relative frequency of the laser as it sweeps 4. One card using 3 simultaneous inputs was therefore sufficient. On the new setup, one summed current value is needed on each card for synchronization purposes, absorption data is need from the 5 TED box detectors, and again one etalon is needed. The configuration is as follows: Card 1 Card 2 Summed Signal Summed Signal TED box 1 TED box 4 TED box 2 TED box 5 TED box 3 Etalon Table 1: Data input configuration In order for the three signals to get to the TED boxes correctly, laser switches were purchased. These switches would ensure that lasers 1, 2, and 3 were passed along at the correct time. Two 2-to-1 switches were purchased from Thorlabs and were connected in the manner shown in Figure 4. Switch 1 Laser 1 Laser 2 Out Switch 2 Figure 4: Laser switch setup Laser 3 These switches could be controlled in several ways, either using a USB connection or through falling edge detection of a digital 5v signal. The method used in this project was the digital signal because it was easier to implement in LabVIEW and synchronized with the output signal better. In each waveform, the switches needed to switch positions twice so they began and ended each waveform in the same position. As a result each digital signal had to have two drops per waveform. The way this was accomplished was by creating two separate waveforms per switch and then combining them using logic statements to keep the signal high except when the switch position needed changing. This is shown in Figure 5. Ellison 5

6 System Testing Once the code modifications were made, the next step was to put all the different components together at UVa and test the motion control, laser control, and data collection to ensure that they all worked harmoniously before making a trip down to NASA Langley. NASA Langley Major Trip 1 Figure 5: Switch control waveform The switches first received did not operate as fast as they should have. The specification in the manual said that the switch time was 1 millisecond, or less, when in reality these switches were taking 7 milliseconds. This was a problem because the 1 millisecond downtime was needed to keep the laser-on time maximized. In the one tenth of a second collection time, there are 30 laser sweeps and a switch is needed in between each sweep, so the switches were sent back and new switches which took about 0.6 milliseconds were used. It was determined that with a 27% uptime per laser, the 0.6 ms switches could be used without signal loss. Figure 6 shows the relative uptime vs downtime for the lasers to give the switches sufficient time to respond. Lasers on Lasers off (downtime) Figure 6: Laser uptime vs. downtime With the setup verified and tested, the next step was to test the system on a flat flame burner in the laser diagnostic lab at NASA Langley. The purpose of this test was to check that both hardware and software worked together to collect data correctly. A series of tests were run at different fuel conditions and it was determined that several minor modifications had to be made to both the code and hardware. The numbering of the rays within each fan was changed to indicate the angle with more precision (in increments in tenths of a degree). The original plan was to have each fan consist of 25 rays at an increment of 1 degree. Since the flat flame burner is much smaller than the test section area, each fan still consisted of 25 rays, but at increments of 0.3 degrees. This would increase the precision of the tomographic reconstruction of the flat flame. With the idea of shrinking the increments in mind, the next data set attempted, was modified to cover the 25 degree fan angle in increments of 0.25 degrees. This resulted in 101 rays per fan. As this test was being run, the program gradually got slower and slower and the TED boxes started over-rotating and the test had to be aborted. The cause of this malfunction has not been determined with certainty, but it would seem that the commands running from the computer to the TED box stages started to get misread over time when the increments were that small. Ellison 6

7 Referring back to Figure 1, the sensors in the fan beam geometry appear on the far side of the test section from the origin of the fan. Logistically speaking it is not easy to cover one arc, much less five arcs with an array of sensors. Instead what was done was to cover the arc in a reflective tape that would send the laser back through the same path it had just traveled and a mirror inside the TED box would deflect the signal to the detector, or sensor. The problem with the first 5 TED box design was that the 24 degree fans did not quite cover the entire 28.9 square inch test section on the NASA tunnel exit. This was a problem because by not covering the entire test section, inaccuracies in the reconstruction could have formed. This issue could not be immediately fixed on that trip, but new wider reflector arcs were machined and each fan was programmed to be 26 degrees wide instead of 24 degrees. Another issue which came up was that over time the laser signals and the laser switch controls slowly became desynchronized within each run. This meant that gradually the switches were being activated a bit later every fan. This led to a gradual loss of the beginning of each sweep on the lower (2 volt) side. The absorbance peaks would not be lost in a 2.5 minute test, but in more extended tests, the absorbance peaks would eventually be cut out because of the switch lag. Fortunately, the test times in the NASA Langley wind tunnel will be short so this problem will not be apparent. At this point in time (mid March) this problem has not been solved, but it has been determined that the two data output cards sample at very slightly different rates and it could be due to the internal hardware within the cards. The lag is such that after a minute of test, one card ends up 20 samples behind the other out of the 6 million samples created. Although this may not seem like much of a drift, since the data collection windows are only one tenth of a second, it is enough for concern during longer tests. NASA Langley Trip: New Retroreflector After the new retroreflector arcs were machined and installed another quick trip was made to NASA Langley to test the tomographic reconstruction method in the laser diagnostic lab. Rather than measuring temperatures and absorbance, an obstruction consisting of two D batteries taped side by side was placed off center within the imaginary test section plane. This obstruction would block the laser signals and using tomographic reconstruction the shape of the obstruction was successfully reconstructed. This reconstruction, as well as the absorption data from the first major trip, validated the system s data collection. At this time the project was deemed ready to move to the Direct-Connect wind tunnel to begin testing on the week of March 19 th. This paper will not be able to cover notes from these tests, but this next section will outline considerations and concerns that will have to be addressed while running on the tunnel. NASA Langley Tunnel Tests The same equipment setup will be used on the NASA combustion tunnel, but due to noise and access concerns components will be divided up into three locations: the test cell itself, a room next to the test cell, and a control room above the test cell. Most of the laser equipment will be in the room next to the test cell to protect the lasers from the noisy environment. Because some of the motion controller cables could not be extended they will have to remain in the tunnel run. The computer itself will be in the room adjacent, and its monitor will be in the control room to control the experiment when the wind tunnel is running. Ellison 7

8 Another consideration that had to be taken into account for the NASA wind tunnel experiments was the synchronization between the TDLAT system and the wind tunnel. The facility will send a 5 volt signal to indicate the beginning of a tunnel run. This signal will be connected to a digital input line and, when the signal goes from low to high, the program will start. Earlier in the project, a second input card was added. In order to check-in on the data collection program as it is running, a second tab was added to display card 2. This change was not absolutely necessary, but it was beneficial to be able view data from both cards as it was being collected. A full data set is composed over a total of 18 fans at an even spacing of 4 degrees. This ensures a 360 degree coverage since there are 5 TED boxes. Given the very short runtimes of about 30 seconds when fueled, each program run will only include 2 fans at a spacing of 36 degrees. In total it will take 9 tunnel runs to obtain a complete data set. The unfueled runs can last a bit longer, so each run will contain 3 fans spaced 24 degrees apart and a complete data set will be collected after 6 runs. To-Date Program Configuration The current LabVIEW user interface configuration is shown below in Figure 7. It includes control of the fans and rays, as well as laser and motion starting and stopping, streaming input data, error messages, and other controls more specific to the inner workings of the program. Between when this paper is written and the end of the final tests as NASA Langley, this code is subject to change. Further Importance As mentioned earlier, this project will not only help advance the knowledge of Fan and Ray Input Control Position Start Program Start Waveform Error Messages Figure 7: Screenshot of most important program feature Ellison 8

9 hypersonic technology and scramjets, but it will also advance the technique of TDLAT. This technique is still fairly new and it has potential as a diagnostic tool. A similar tomographic system could be attached to the exhaust of actual engines in a larger wind tunnel to obtain similar information on water vapor concentration, temperatures, and combustion efficiency. Conclusions Much work has been done on this project over the past year. A number of challenges were confronted and have been addressed by making modifications to the lasers, motion control, and data acquisition on the software side. As the NASA Langley wind tunnel tests are approaching, there is confidence that the system will work as envisioned and as demonstrated. The results from this project will aid in the understanding of scramjets, and will hopefully help make the idea of a hypersonic airibreathing aircraft become a reality down the road. 2 Background, National Center for Combined Cycle Propulsion, [ center.us/about/background/ Accessed 3/22/12]. 3 McDaniel, J., Goyne, C., and Diskin, G., Combustion Efficiency Measurement for Ground Test and Basic Hypersonic Research, NRA # NNX07AC35A Final Report, Dec Busa, K., Bryner, E., McDaniel, J., Goyne, C., Smith, C., and Diskin, G., Demonstration of Capability of Water Flux Measurement in a Scramjet Combustor using Tunable Diode Laser Absorption Tomography and Stereoscopic PIV, AIAA , Jan Bryner, E., Sharma, M., Goyne, C., McDaniel, J., Snyder, M., Krauss, R., Martin, E., and Diskin, G., Tunable Diode Laser Absorption Technique Development for Determination of Spatially Resolved Water Concentration and Temperature. AIAA , Jan Acknowledgements This project was supported by the Virginia Space Grant Consortium, the University of Virginia Aerospace Research Lab, and NASA Langley. The author would like to thank Dr. Glenn Diskin and Kristin Busa for their support and assistance, as well as fellow undergraduate student Brian McGovern for his contribution to the hardware design. References 1 McDermott, N., London to Tokyo in two hours: Blueprints for 3,000mph hypersonic plane are unveiled but it will take 40 years to build, Daily Mail, [ /article /london-tokyo-2-hours-blueprints-3-000mph-hypersonic-plane-unveiled.html Accessed 3/17/12.] Ellison 9