Turbulent Autoignition of Hydrogen and Acetylene in a Duct C.N. Markides, E. Mastorakos* Hopkinson Laboratory, Department of Engineering, University of Cambridge Cambridge, UK Abstract This paper presents experimental data on the autoignition of hydrogen and acetylene jets in a turbulent co-flow of hot air. The aim is to understand better how turbulence affects autoignition of non-premixed flows and to obtain data for validating modelling approaches. The conditions under which flashback is observed following autoignition have been quantified as a function of air and jet velocity and air temperature. It is demonstrated that the turbulence causes a randomness in the ignition location, that it can delay the onset of autoignition and that the autoignition time has a higher sensitivity to temperature than that expected from homogeneous mixtures. order to avoid short-circuit problems found with commercial heaters and consist of two pairs of in-line units. Fuel is injected axially and continuously through a d=2.2 mm ID stainless steel nozzle. For most of its length, the nozzle is encased in a ceramic pipe of 2 mm thick walls to insulate the fuel from the hot air. The nozzle is located about 60 mm from the perforated plate to allow the turbulence to develop. Both fuel and air flow rates are controlled by digital mass flow controllers, while nitrogen is measured by calibrated rotameters. The fuel and nitrogen are supplied from bottles, while the air from a compressor, after passing through filters to remove any solid and liquid impurities. Hydrogen and acetylene (99.999% and 99.6% purity respectively) are used as fuels here. The temperature immediately downstream of the heaters is measured by N-Type thermocouples and used to individually control the voltage supplies. 25.0 mm upstream of the injector, another N-Type thermocouple is embedded in the air flow to measure the air temperature, Ta. This can be as high as 1150 K and the air can flow with bulk velocities Ua (calculated from the settings of the mass flow controllers and the measured temperature) up to 75 m/s at the nozzle location. The whole assembly is heavily insulated with a 150 mm thick lagging of non-refractory ceramic blankets in order to keep heat losses to a minimum and the temperature field as uniform as possible within the full length of the tube. A vertical slit about 300 mm long and 30 mm wide is used for optical access. Radial profiles showed that indeed about 80% of the pipe radius is at a uniform temperature, with a thermal boundary layer growing close to the wall. The temperature of the fuel/N2 mixture at injection cannot be measured easily because the bare-wire thermocouple acts like a catalyst and causes ignition. In experiments without fuel and despite the insulation of the nozzle, the exit temperature of the fuel (Tf) has been found to be high (but less than Ta) due to convective heat transfer from the injector wall. By extensive
Introduction The effect of turbulence on autoignition of inhomogeneous mixtures is a topic of fundamental importance in turbulent combustion and of practical relevance in diesel and Homogeneous Charge Compression Ignition engines and in lean-premixed prevapourised gas turbines. The development of these novel engines is hindered by the capability to predict the behaviour of the slow reactions leading to autoignition in the presence of inhomogeneities. In past efforts with DNS, it has been revealed that ignition occurs at a “most reactive” mixture fraction and at spots with low scalar dissipation rates [1-4]. However, these findings have not been confirmed experimentally and the effect of turbulence intensity and lengthscale on ignition time has not been clarified. In addition, the connection between the ensemble-mean behaviour (for example the average ignition timing measured from many cycles in a diesel engine) and the possibility of relatively rare events causing occasional dangerous autoignition (e.g. in a gas turbine premix duct) has not been explored. In an effort to understand these phenomena better and to provide practically-relevant information on ignition times in the presence of turbulence and scalar gradients, an autoignition rig with turbulent flow has been built and the first results from this experiment are presented in this paper. Specifically, the objectives are to describe the main features of autoignition of fuel plumes in fastmoving hot oxidant and to quantify ignition times as a function of flow parameters. The experimental methods are described in the following Section, followed by the results. The paper closes with a summary of the main conclusions. Experimental methods The apparatus is shown in Figure 1. Air at atmospheric pressure is heated electrically and passes through a mixing block and a perforated plate with 3 mm holes in a 1.5 m long quartz pipe of inner diameter D=25 mm. The heaters have been designed in-house in * Corresponding author:
[email protected] Associated Web site: http://www.eng.cam.ac.uk Proceedings of the European Combustion Meeting 2003 1
measurements at various Ta and flow rate conditions just prior to those needed for autoignition, an analysis for heat transfer across the nozzle was validated. This was subsequently used to calculate Tf when autoignition occurs. The range of Tf so calculated for the experiments is between 550 and 750 K. The bulk velocity of the fuel jet Uj was lower, equal or larger than the air, with the ratio Uj/Ua lying between 0.8 and 3. For the hydrogen experiments, images of OH chemiluminescence were taken with an ICCD camera (LaVision Nanostar) fitted with a Nikon UV lens and an interference filter band-centred at 307±10 nm. An exposure time of 150 µs and high gain settings were used to capture the weak chemiluminescence signals. Individual images were processed in stages in order to remove background noise and increase signal-to-noise ratio. The processing included adaptive Weiner filtering, median smoothing and 4th order low-pass filtering of the signal in the frequency domain. The signal was normalized and image background noise was removed by setting the pixel intensity to zero below a certain threshold and to unity above it. Hence, the presence of OH at a pixel location resulted in a normalized unity signal, while the absence of OH in zero. The region of presence of OH is deemed an “autoignition spot”. About 100 images were used to compile the two-dimensional joint probability density function (PDF) of the spot location. Estimations of the mean 〈L〉 and standard deviation Lrms were calculated directly from the PDFs by numerical integration. An alternative definition of autoignition length was based on the minimum axial location of an OH spot observed during a run (Lmin). This was defined as the axial distance at which the PDF has reached a rise height from the background of 5% of the peak value. Due to symmetry the mean cross-stream location of the autoignition spot was zero, but the r.m.s., Yrms, which is also calculated from the PDF, can be used as a measure of the radial spread of the spot. For the acetylene runs, the chemiluminescence apparatus was not available and direct images were taken with a digital camera with and without an externally fitted blue filter. About 30 images were used to quantify the minimum ignition length. Finally, highspeed films were taken with a Kodak digital camera at framing rates up to 40 KHz to visualize the time evolution of the autoignition spots. The method of rig operation is as follows. Once the rig reached thermal equilibrium after about one hour of operation, the required flow rate of air was set and a small flow of N2 passed through the fuel nozzle to keep it cool. Then the fuel was switched on and mixed with nitrogen upstream of the nozzle and final adjustments to the flow rates were made. The dilution of the fuel stream is described by the mass fraction of fuel,Yfu,0, which is 0.13 for all data shown here for hydrogen and 0.71 for acetylene. Limited velocity measurements were taken with a Pitot tube under high temperature conditions to confirm that the air velocity was uniform across the tube. Under cold conditions, velocity measurements were taken with
4.5µm hot wires which were also used to characterize the turbulence. In order to examine the velocity field not just at the nozzle location, but also at downstream locations, the main quartz tube was replaced by shorter lengths in order to provide physical access to the location of the measurement. Results and discussion Initial conditions To estimate the turbulence intensity and lengthscale, as well as to examine the uniformity of the mean velocity across the pipe, hot wire measurements with air at atmospheric temperature were undertaken. For these measurements, a bulk velocity lower than that encountered under hot conditions was used to keep the Reynolds number the same. The jet velocity Uj was similarly varied so that the ratio Uj/Ua was the same as in the autoignition experiments. Figure 2 shows that the axial velocity, normalized by the bulk velocity, is uniform for up to 80% of the pipe radius and decreases towards the wall. The fuel jet is evident, as is also the wake-like behaviour for Uj/Ua
