Combustion Sensors: Gas Turbine Applications A Final Report submitted to NASA Ames Research Center Grant# NAG 2-1430 Mel Human Department of Mechanical Engineering North Carolina A&T State University Table of Contents IBackground II TraditionalSensor Devices m Optical Methods W Other Techniques V Implementation VI Conclusions and Recommendations Appendix ALaboratory Configuration Summary Appendix BComputer Aided Engineering(CA[/) Support Figures 1-10 Experiment Schematics References EXECUTIVE SUMMARY This report documents efforts to survey the current research directions in sensor technology for gas turbine systems. The work is driven by the current and future requirements on system performance and optimization. Accurate real time measurements of velocities, pressure, temperatures and species concentrations will be required for objectives such as: combustion instability attenuation, pollutant reduction, engine health management, exhaust profile control via active control, etc. Changing combustor conditions - engine aging, flow path slagging, or rapid maneuvering - will require adaptive responses; the effectiveness of such will be only as good as the dynamic information available for processing. MI of these issues point toward the importance of continued sensor development. For adequate control of the combustion process, sensor data must include information about the above mentioned quantities along with equivalence ratios and radical concentrations, and also include both temporal and spatial velocity resolution. Ultimately these devices must transfer from the laboratory to field installations, and thus must become low weight and cost, reliable and maintainable. A primary conclusion from this study is that the optics based sensor science will be the primary diagnostic in future gas turbine technologies. Accordingly, these techniques dominate the following discussions. The various procedures involve some type of illumination of the flow filed, and subsequent signal interrogation. While "classical measurement devices will still have usage, it is believed that laser driven system provides the overall performance and responses which will be required by the mentioned system objectives. Particle Image Velocimetry (-PIV) allows velocity time measurements over a region of space. Planar Doppler Veiocimetry (PDV) and Phase Doppler Particle Analysis (PDPA) are similar to PIV, but demonstrates superior three dimensional resolution due to a use oftbe scattered particle's frequency shift. Further advances in laser and camera technology will enhanced its effectiveness. Laser Doppler Anemometry (LDA) also allows instantaneous flow rate determinations, and is particularly effective for assessing injection droplet velocities, sizes and number densities. Phase Doppler methods involve information processing of laser generated fringe patterns, and has the advantage of not requiring calibration, apotentially significant advantage in field applications. The scattering analysis isfurther exploited by, Filtered Raleigh Scattering fiRS), where vapor filters are used to suppress background noise, and allow more accurate estimates of quantities such as temperature fields. Investigators have used this technique to measure temperature in a combustion environment, and the simultaneous measurement of density, temperature, pressure and velocity. Laser Induced Incandescence (LII) involves invoking a radiant response by the heating of target particle surfaces, and subsequent data processing via charge coupled device (CCD) techniques. It has proven to be effective in determining soot concentrations in exhaust streams an important result, as soot isa good indicator of combustion efficiency. The observation of chemiluminescent emissions from particular species in exhaust gases can lead to conclusions about combustion energy release rates and zone equivalence ratios. Very promising results have shown how the relatively rugged diode laser can be incorporated in systems for harsh environments. The research efforts in these areas show great promise. However, there appears to be a relative dearth of information concerning actual implementation ingas turbine systems. Issues to be resolved include the size, complexity and ruggedness of these devices. Thermal, shock and vibratory environments in real world systems will be more severe than laboratory conditions. Limited optical access, restricted geometry, background luminosity, spectral interference, and fully developed turbulence, are all issues, which will require attention. Line of sight geometry must also be established in very constrained flow passages. Field system implementation and cost reduction are significant issues for consideration. A major area for investigation is defining exactly how these advance systems are to be used. System performance, health monitoring, active control are all candidate objectives. The actual usage will influence what variables are to measured, where such interrogations must be performed, and to what accuracy and bandwidth. Accordingly, a major recommendation is reviewing the gas turbine system role and initiating and continuing: in-depth investigations for establishing objectives and associated measures of merit, defining system architecture, and performing simulation studies. This will assist in answering a number of questions about sensor requirements and expected performance. In addition to the optical systems, a short discussion is done on acoustic and electromagnetic methods; these techniques could also prove useful inspecific applications. IBackground Performanlecevelsfornewgeneratioonfsprimemoverso,fspecial interest here - gas turbine systems will have to meet higher levels of requirements. The challenge isa formidable one as it involves the simultaneous and somewhat conflicting tasks of improved thermodynamic performance and reduction inpolluting emissions. Obtaining these enhancements inperformances will rely significantly in the ability to monitor, measure and control various processes throughout the flow path. Accordingly there will be great need for high fidelity sensor devices and the corresponding data fusion methodologies for optimizing engine parameters. The ability to continuously monitor the physical and chemical environments associated with combustion processes isand will be an important function in the energy and propulsion technology areas. These diagnostics provide information for supplementing the understanding of the chemical kinetics; specie consumption and depletion rates and data input for real time applications such as active control. The development of digital control modules along with microprocessor capabilities have positively impacted the aviation and gas turbine engine community with the capacity to improve performance and reliability of such systems (Alden-99, Eckbreath-96, Gord-01, Kiel-01, McManus-93, Schadow-96, Zinn- 97). Sensors are essential for active control using state feedback methodologies. Combustor control requires devices that can respond to fluctuations inheat release rate, variations in local fuel-air ratio, and temperature estimates at the exhaust plane. For aero-gas turbine engine applications, control of combustor exit temperature, combustion instabilities, and pollutant emissions are desirable control goals, all, which require accurate real time data streams. New techniques inengine health monitoring will require real time and historical data recording. Chemical reactions such as combustion processes are dependent on the relative composition on the reactants, the initial thermal state -temperature and pressure, the presence of any catalytic substance, and the physical dynamics of the initial state such as a fluid velocity. In real processes while these parameters may be initially known, as the process proceeds measurements must be taken to constantly monitor the reacting environment. It is a complex coupled phenomena where effects of the reactions such as exothermic energy release impacts the environment which inturn influences the reactions, etc., etc. Real time data acquisition is imperative in the understanding of the process evolution and sensitivity of the reactant-product dynamics with respect to the controlling factors. Advances in the improvement of gas turbine systems are strongly linked to greater understanding about the physics and chemistry of multiphase turbulent reacting flows. Current research areas such as supersonic combustion, pollutant and soot formation, turbulent combustion and interactive control will continue to rely heavily on real time temperature and chemical sensing along the flow path environment. Higher performance requirements (up to 4000 F temperature and 30-40 ATM. pressure) will demand accurate scaling and design laws that may be confirmed only from actual measured data. Higher fidelity validation and benchmarking of computational fluid dynamics codes used inengine analysis and design will require more accurate measurements. Turbulent closure of these models may also be assisted by such data. Automonous and ongoing health management for engine and vehicle require accurate evaluation systems. This will lead to reduced maintenance and life cycle costs. Laser based methods such as CW high-resolution spectroscopy and ultra-fast spectroscopy will are prime candidates for the task of real time non-intrusive temperature, velocity vector, and specie concentration assessments in gas turbine systems. Improvements must be made to current Doppler and anemometry velocity measurements as they involve bulky systems unsuitable for field usage. Specie measurement isless practical as procedures such as absorption, laser induced florescence and Raman scattering are even more complex. Traditionally, measurements are considered as either intrusive or non-intrusive. Intrusive techniques, such as hot wires and Pitot tubes while fairly accurate, interferes with the flow and does not provide spatial resolution (Fraden-97). High sensitivities are needed in many applications. For example, scramjet engines generally exhibit relatively small end to end increase of momentum flux. Thus proper thrust measurements require high accuracy estimates of density, temperature and velocity. Injection systems involve phenomena such as jet instability, fuel droplet coalescence and disintegration, impingement and gas spray interaction. For accurate estimates, one must couple both downstream and upstream oscillating flows. Clearly, both experimental research and actual field implementation will benefit from continued advancements in sensor technology. IITraditionaSlensoDrevices Inthischaptewr,ebrieflydiscustrsaditionaolr"classicaml"ethodfosrmeasurinveglocityp,ressurfelo,w andtemperature. Velocity, Flow and Pressure Pitot tubes are well known instruments for velocity measurements, using Bernoulli's principle for pressure- velocity head conservation (Moore-65). However they present obstruction issues and can only be used when this is not a problem. Also they can only withstand moderate thermal conditions. Venturi meters are standard flow measurement devices (Baker-81). Pressure measurement devices generally rely on one of three basic ideas: comparison of equivalent heights of a known liquid, measuring force acting on a given area, or determining some pressure correlated physical of property change ina substance. Common instruments employing the first two methods include manometers, Bourdon and diaphragm gages (Benedict-77). An example of the third type iseither the Pirani or thermocouple gage (Moore). Utilizing the change in conductivity with respect to pressure, the current changes in a heated wire are measured and correlated. Pressure sensors have the capability for monitoring fluctuations for combustion instability control although their temperature sensitive calibration characteristics complicate the application. In addition, they cannot withstand combustor liner temperatures and therefore must be encapsulated which interferes with reading integrity. Thermal anemometry (TA) 011- measures fluid velocity by sensing changes in heat transfer from an electrically heated wire or thin film exposed to the fluid. Anemometers are generally classified as either constant-current or constant-temperature, although recent developments of constant voltage devices may make that mode prevalent (Kegerise-00). A control circuit maintains the wire at a constant temperature. Convective transport, which is proportional to the velocity, cools the wire; the required current adjustment for constant temperature ismeasured and calibrated. Advantages include low cost, flexible applications and ready automation. TA is also has utility when fast response time is needed. Clearly there are limitations inthe in-situ positioning for engine applications. The technique can be applied to most "clean" flows where the intrusion of the sensor does not adversely impact the flow field. These systems have deficiencies with respect to bandwidth, sensitivity and signal to noise ratio, which limits their applicability to high speed, high Reynolds numbers flows. Also as just mentioned, contaminates can degrade or damage the relatively fragile sensors. A low level of turbulence is also a desire operational requirement, although some studies have focused on extending the applicability against this shortcoming (Ljus-00). For enhancing the single point statistics of three dimensional flow fields, methods of using two or more probes have been developed (Maciel-00). Cross wire techniques are the most commonly used, although they suffer from relatively long measuring times, inability to resolve simultaneous three component velocities, and variable unsteady flows. Studies have been performed which unite anemometry methods with laser based devices (Dawson-91) Dynamic calibration is difficult for the frequency dependent hot film probes; instead corrections involving turbulent energy spectrum has been investigated, particularly in the area of gas-particle flows. Temperature Temperature measurements are often performed with thermocouples. When two wires of dissimilar materials are joined at two ends which are exposed to two different temperatures, a continuous circuit is established - the Seeback effect. The current magnitude is a function of the temperature difference between the wires, inaddition to the specific materials. In actuality the Seeback voltage isthe measured quantity, and it must be calibrated accordingly. Several factors must be areas of concern when using thermocouples. Positioning of the thermocouple is critical as to correctly measure the intended location. They are unsuitable for insertion into high temperature gas flows, recent experiments at elevated temperature has exposed significant durability issues (Delaat-00). Thermoweils and protective tubes could extend usage for harsh environments; embedding in structural elements may be done for obtaining wall temperatures. Care must be taken that the device's presence does not disturb the flow environment as to give inaccurate estimates to what is truly going on. Thermocouple arrays are often located between high and low pressure stages as combustor exit temperatures prohibits their use at those positions. While this serves to assist in providing engine health monitorinagndignitiondetection, this application isunable to resolve time lags between combustion chamber and sensor location events. As flow has just left the high pressure turbine, flow data representing combustor conditions has been distorted. Inaccuracies due to stagnation pressure and gas dissociation losses misrepresent upstream temperature profiles. Poor mixing isalso not captured. However, the importance ofthermocouples should not be diminishedl Whenever their inherent intrusive nature isnot a factor, they are excellent measurement devices. They have been used as diagnostic tools in turbulent gaseous and spray flames. They are often used as calibration instruments for laser sensors. Response time may be improved by grounding the junction to a protection device. Some temperature measurement results have been favorably compared with those given by laser Raman spectroscopy, and indeed may be superior to lasers on sooty flames and when high sampling rates are required. Recent work has been performed in temperature ranges of 2100-2700 K, which istraditionally beyond thermocouple application (Gokoglu-00). The technique involves monitoring the sharp increase in emittance of certain metal oxide fibers as the material approaches known melting points. In this regime, greater care must be taken for temperature correction, due to enhanced radiative losses. Another temperature measuring device isthe optical pyrometer. A lens focuses unto a calibrated tungsten filament. The light intensity on the filament iskept constant by maintaining a constant current flow. An optic wedge is positioned inthe target light path until the two intensities appear to be equal, and the calibrated temperature is read. In a similar configuration, the radiation pyrometer uses a sensing element instead of a filament. This device is useful for temperatures in excess ofg00 K, As it isa non-contact apparatus; the pyrometer isuseful for measuring moving, remote, or inaccessible objects or surfaces. Radiation thermometers estimate temperature by measuring the radiative emissive power of the source over awavelength bandwidth. It can be shown that higher emissivity surfaces give more accurate readings. Accordingly, coating the targets with thin layers of low reflectance materials improve this application. Specie Concentrations Techniques have been developed for the simultaneous measurement of flow velocity and concentration fluctuations, using a dual hot wire sensor (Sakai-01). A digital data processing algorithm uses the voltage from both sensors and a calibration map. Construction of the probe depends on the conductivity of the specimen gas. The procedure utilizes the so called overheat ratio (OHR), and has the advantage of not requiring constant adjustments of the OHR. Surface mounted hot-film sensors can also be used to obtain pressure recovery values (Jones-01). Results show that high mean and low level rms voltages correlate well with improved pressure recovery. The procedure is also useful for determining flow distortion levels. While not a typical measurement ingas turbine applications, active control efforts may be assisted with real time air stream humidity determinations, giving the not insignificant effect water vapor has on system performance. Psychrometers and hygrometers are standard instruments for humidity measurements. For engine applications, an electronic sensing device is needed. For example, a coil impregnated with a hygroscopic salt demonstrates a humidity dependent resistance where the resulting current can be calibrated to water content values, Similar devices for other gases use Wheatstone bridge circuitry for determining concentrations, which affects the medium's thermal conductivity. Other While not adirect input in combustion control schemes, knowledge of turbulence statistics improves fundamental understanding of flow field behavior and thus could influence gas turbine design thought. Hot wire anemometry, particularly multi wire probes have been used for assessing three dimensional velocity fields along with Reynolds stress estimates (Chert-00). HIOptical Methods In this chapter, we discuss recently developed sensor efforts, namely laser based systems. Non-intrusive sensing has obvious advantages over many traditional procedures, as the flow field is not perturbed by the presence of a sensor. Optical sensors have the capability of gathering data in hostile environments and over larger local areas than traditional devices. Passive optics often are relatively simple, thus enhancing maintainability. Photodiodes offer fast response, low power requirements, and spectrally tunable operation. These devices offer high spatial, temporal and spectral resolutions for temperature, velocity and specie concentration measurements. Early non-intrusive techniques include schilieren and shadowgraphs provided high spatial resolution of density gradients, but were limited by line of sight restrictions of the equipment. Today's laser based systems alleviates some of these restrictions although each method had its own set of advantages and disadvantages (Mayinger-01, Kiel-01, Gord-01). This chapter's discussion follows a somewhat different pattern vs. Chapter 2. Because of their inherent capability to measure a number of quantities, the technologies discussed here are not segregated by measurement variable. Laser Doppler Anemeometry (LDA) The general setup involves alaser beam passing through a splitter, which provides a two-beam illumination of a test volume. The beam pair share a shifted frequency, depending on the spatial displacement. The intersected scattered light is directed unto a photomultiplier whose signal is processed via an interface board. In general a number of interrogations, perhaps in the thousands, are needed for an accurate processing of velocity profiles (Ismailov-01). Planar patterns are often observed using a laser sheet an optical array of lenses focuses the laser output into a plane pattern. Such a measurement procedure would provide information for: a) Flow dynamics in combustion chamber fuel mixing region - combustion process optimization requires fuel droplet size temporal distribution b) Integrated fuel mass flow rate - requires droplet number density c) Jet characteristics - high levels of stratification may occur inthe jet stream: high concentrations near nozzle, aregion near the conical edge, and also along a vortex edge. In addition, instantaneous flow patterns may be visualized by using high speed Charge Coupled Device (CCD) cameras which may give a resolution as close as 3-5 microns. Laser Doppler Velocimetry (LDV) This technique gives accurate velocity information. Particles ina flow field arelaser illuminated; the scattered light isthen collected and processed. Usually, a single ray is projected through a beam splitter, and the two equal intensity beams are focused at a common point. An interference pattern is formed which isthe measuring volume. A photodetector collects the scattered light, and the resulting frequency is related directly to the particle velocity. If one of the two beams isfrequency shifted, even flow reversal patterns can be interrogated. LDV has the capability to measure three velocity components by using different frequencies, giving an instantaneous snap shot of the flow field (Thurow-01, Elliot-99, Lempert-96, Samimy-00). Temporal resolution remains the primary handicap of the method (McKenzie-96, Smith-98). In practice, the frequency shift cannot be directly measured. Typically, a molecular filter consisting of a glass cell containing a gas such as iodine, isused a frequency discriminator. Due to rotational and vibrational molecular transitions of the cell gas, light transmission is a function of frequency, and gas pressure and temperature. Therefore, the light's frequency shift can be determined by scattered light's transmission through the filter. LDV would appear to be agood candidate for exhaust nozzle flow profile assessment. Particle Image Velocimetry (PIV) These systems measure velocity by determining particle displacement with a pulsed laser technique (Raffel- 98). Particle positions are illuminated ina plane by a laser sheet, and recorded using a digital camera. At a prescribed time segment, another pulse highlights the same plane, creating a second particle image. Using these two images, data processing algorithms based on Fast Fourier Techniques (FFT) correlation methods determine displacements for the entire flow regime, a major result, including velocity information. An important feature is that turbulent statistics can be assessed. Separation points and re-attachment lengths can also be determined. Because of this, PIV could be a useful monitor for diffuser passages, and if line of sight issues addressed, turbine blade flow surfaces. Phase Doppler Particle Analysis (PDPA) This method uses light scattering interferometry. Two laser rays intersecting inthe control volume create a fringe pattern. Crossing particles scatters light and projects the fringe, which is, detected at several off-axis detectors. Each detector produces aDoppler burst signal with a particle velocity dependent frequency. The phase shift between two different detectors is proportional to particle size. The method isattractive, as it requires no calibration and the two outputs: velocity and size are dependent only on laser wavelength and optical configuration. Thus the procedure is quite robust for dense particle and combustion environments. Laser Induced Incandescence (LII) LII is based on optical emission and absorption in the visible to mid infrared. LII measurements of particulates such as soot (Wainner-99) has been shown to be as accurate as 10(10)/m3 particle density (Quay-94, Melton-84). This is a key technique in the measurement of major specie concentration, high concentration CO and gas temperature (Xin-01, Morrell-01). Also general exhaust radiative signatures may be registered. Access to ultraviolet ranges has also been recently investigated (Brown-98, Yang-98). Laser induced incandescence is typically observed as a pulse on the order of a microsecond of grey body radiation from a particular matter that is heated to near vaporization temperatures by a pulsed laser. Energy absorption heats the surface up to 4000 K, followed by rapid cooling which depends heavily on partical size. The magnitude and decay rate of the luminescent pulse can respectively be used to determine respectively, particle concentration and size. The "two color" or dual wavelength method can perform the data processing where the temperature can be determined from the spectral distribution of the response. Laser Induced Luminescence (LIL) When a laser pulse heats certain reactants, they emit an excited spectra response. Using alaser which is tuned to an atomic or molecular absorption transition and subsequent detection of the emission, it is possible to infer both specie concentration and temperature. The main drawback is the so-called quenching phenomena where coilisional deexcitation attenuates LIL signals, but research efforts have determined methods to minimize this effect. Extensive research has been performed on chemiluminescent emissions from OH*, CH*, and C2' molecules (Mizutani-89, Samaniego-95). OH* and CH* have been found to be good indicators of reaction heat release rate as they exist in high concentrations inthe flame zone, the former also providing good information on post combustion zone conditions. The planar induced fluorescence inOH* is an excellent flame marker (Jensen-86, Keller-87). Correlation between OH* and heat release Q, and burning rate have been studied. OH* and C2" measurements have been used to estimate flame front movement (Roby-95). The characteristic of having relatively strong emissions, which are spectrally narrow, makes these excellent characteristics for monitoring. Useful linear relations have been found between the equivalence ratio and the C2*/OH* value. The simultaneous determination of local combustion conditions such as Q and reaction zone equivalence would play a major role in active control schemes. Combustion radicals demonstrate flame emission wavelengths inthe range 305 - 515 nm In addition, we must deal with the broadband emitter CO2" which may obscure other signals. As chemiluminescense form these radicals result from chemical reactions and have very short lifetimes compared to convective time scales, their signals provide good information on the reaction zone ina combustor. Species tracking in a linear flow path has shown to be quite accurate for velocity determinations (Sanders-00, 01, Littleton-00). Accordingly, in addition to flame structure information, LIL and LII are useful exhaust measuring techniques. Filtered Raleigh Scattering fiRS) In this technique, the scattered spectrum from an illuminated flow field ispassed through avapor filter whose absorption line is tuned with the laser frequency (EUiot-92, 99, 01, McKenzie-96, Hoffman-96, Meyers-91). This improves flow visualization by filtering out strong background scattering. Key parameters controlling this application include laser intensity, frequency, and polarization properties. If the laser istuned to the center of the absorption profile, most &the background interference will be absorbed (Miles-91). Since the Doppler shift is quite small, is possible that the particle scattering could shift completely out of the absorption line. This can beavoided by the proper selection of incident and observation angles. FRS is ahighly thought of procedure for performing multi-property measurements (Elliot-96, Forkay-96). All of the characteristics of aflow field - temperature, density, pressure and velocity have some effect on the spectral intensity of the Raleigh scattering profile. For example, density is proportional to the integrated intensity, width of scattering isrelated to temperature, and profile center position is correlated to the Doppler shift or the velocity (Lempert-97). The issue becomes the separation of these effects for real time flow calculations. A proposed improvement to the methodology is using a frequency scanning technique (Elliot-01). The laser frequency is scanned across the absorption line and the subsequent scattering transmission through an iodine cell is measured. A typical setup has measurements performed at five points. Coherent Anti-Stokes Raman Scattering (CARS) This technique employs two or three laser beams, which excite a Raman transition in such a way that a third beam is produced. This resultant beam is related to the local specie concentration. The majority of experiments have been performed by interrogating rotational/vibrational Raman transitions with a 1000- 4100/cm shift, or rotational transitions below 300/cm The latter has the advantage of being able to detect simultaneously several species. However, the method has serious signal resolution issues. A major advantage over the LII for gas turbine applications is better possible in-situ configurations: spectral interference from background fluorescence isminimized, and a smaller detection angle is needed. However, the proper alignment of multiple laser paths is an issue. Fourier Transform Infra Red Spectrometry (FTIR) FTIR spectrometry is very capable nonintrusive method for measuring spectral emission signals, particularly those present inhigh temperature exhaust streams (Heland-97). For typical combustion temperatures, emissions from flame specie arise from thermal excitation of vibrational and rotational states in molecular ground states, the resulting spectra consisting of many closely spaced lines. Primary quantities of interest include the emission intensity of each spectrum, the transition frequency, and the particle density, which is given by the Boltzmann distribution equation. Using these relationships, temperature dependent line by line emission spectra can be determined. Thus we can simulate flame molecular emission for a given species concentration and temperature. Simple diatomic specie such as CO and NO can have their spectral line data calculated with known spectroscopic constants. FTIR can be used for specie determination in the 2000- 2200/cm range for CO2, CO, NO and H20 measurements. The integrated area of four absorption lineshapes, which are only temperature dependent, can measure H20 density. One proposal for the combustor exit temperature metric isthe infrared absorption lines, primarily inthe 1310-1600 nm range (Allen-98, 01). Measured absorption ratios are functions of temperature. Tunable diode lasers have been used this way by path averaging techniques, although soot presence degrades the signal to noise ratio (Wang-00). Transient Grating Spectroscopy (TGS) This is a technique where amedia's electronic state is excited, and atemporal grating gives the wave speed of the gas. Of course this isrelated to the mean molecular weight and specific heat, from which temperature is derived. This isa promising procedure as point measurements could be made in the internal combustor cavity. TGS can also compensate for soot presence. Far and Mid-infra red Many of the discussed techniques involve measurements in the ultraviolet, visible, and near infrared section of the spectrum. Mid and far infrared could yield even greater insights inthe thermal flow field; numerous species do not possess electronic transitions; however all molecules exhibit vibrational spectra which can be