Temperature Profiling of Pulverised Coal Flames Using Multi-Colour Pyrometric and Digital Imaging Techniques

Transcription

1 IMTC 005 Instrumentation and Measurement Technology Conference Ottawa, Canada, May 005 Temperature Profiling of Pulverised Coal Flames Using Multi-Colour Pyrometric and Digital Imaging Techniques Gang Lu 1, Yong Yan 1, Steve Cornwell, and Gerry Riley 1 Department of Electronics, University of Kent, Canterbury, Kent CT 7NT, UK RWE npower plc, Windmill Hill Business Park, Whitehill Way, Swindon SN5 6PB, UK Abstract This paper presents an imaging based multi-colour pyrometric system for the monitoring of temperature and its distribution in a coal-fired flame. A novel optical splitting/filtering device is designed and used to split the light of flame into three beams at three selected wavelengths as required in the multi-colour principle. A high-resolution CCD camera is employed to collect the three beams of the light of flame. The three resulting images provide the basis for the determination of temperature and its distribution in the flame field. The system is evaluated on a 0.5- MW th coal-fired combustion test facility under various combustion conditions. Results obtained demonstrate that the system is capable of measuring the temperature and its distribution concurrently in the flame field. Quantitative relationships between the measured results and the main combustion process data are also discussed. Keywords coal flame, temperature distribution, CCD camera, multi-colour pyrometry, image processing I. INTRODUCTION The temperature and its distribution of a pulverised coal flame in an industrial furnace provide useful information for the in-depth understanding of combustion processes including coal devolatilization, radiative heat transfer, pollutant formation process, and the cause of combustion problems such as slagging and fouling [1]-[3]. However, accurate and reliable temperature profiling of a pulverised coal flame remains challenging. Optical radiation pyrometry has been recognised as the only practical, non-intrusive method of measuring the temperature of flying particles in a flame. However, common total radiation or singlewavelength pyrometers cannot provide the accurate temperature measurement of the flame because the emissivities of particles in the flame are normally unknown. A two-colour pyrometric system measures the temperature by determining the radiative intensities of the object to be measured for two given wavelengths regardless of the absolute emissivity, and has therefore been widely applied to the temperature measurement of flames in various situations. The availability of low-cost CCD sensors and recent advances in digital image processing techniques have overcome the drawback of a conventional two-colour pyrometer where only the averaged temperature within a small area of the flame field is detected. A number of CCD based two-colour systems have been developed in recent years for the measurement of temperature distribution of pulverised coal flames [4][5]. In such systems, the emissivity of the flame had been assumed to be evenly distributed, i.e., exhibits grey-body behaviour. This assumption may be valid for flames fired from coals of the same source, but may not be suitable for one with different radiative and optical properties, particularly in cases where fuel blends, biomass and wastes are fired. This paper presents an instrumentation system for the measurement of temperature and its distribution in a coalfired flame field based on the three-colour technique. A unique beam splitting/filtering device is designed which splits and filtered the light of flame into three narrow-banded beams. A high-resolution CCD camera collects the three beams of the flame light and forms three images which are identical in size but of different wavelengths. The three resulting images give three combinations of either two images. Three two-colour temperature distributions of the flame are then computed simultaneously using the three image pairs. The final temperature of each point in the flame is estimated from the weighted average of the three two-colour temperature values for improved accuracy and reliability. Key design aspects of the system together with experimental results obtained on an industrial-scale combustion test facility are addressed and discussed. It should be stressed that although a number of multi-colour pyrometric devices have been developed and used in various applications [6][7], there are no multi-colour instrumentation systems available for on-line, continuous temperature profiling of a pulverised coal flame. II. SYSTEM DESCRIPTION A. Measurement Principle The principle of the two-colour pyrometry along with a detailed description of the temperature calculation have been reported elsewhere [4][8]. The technique of the three-colour is an extended version of the two-colour method. In the interests of convenience and clarity, a brief summary of the operating principles is given here. In this study, the emissivity of the flame is considered to be unevenly distributed in consideration of fuel blends being fired. By applying Planck s radiation law, the monochromatic emissivity of a non-black body at a given wavelength () can be expressed as: exp( C / T ) -1 = exp( C / T ) -1 a, (1) /05/$ IEEE 1658

2 where T a (K) is the apparent temperature of the non-black body, which is defined as the temperature of a black body which emits the same radiation intensity as the non-black body at temperature T, C is the second Planck s constant ( mk). For sooting particles in a flame, can also be estimated by the widely used empirical equation [9], = 1 exp( KL / ), () where K (m -1 ) is the absorption coefficient, L (m) is the geometrical flame thickness along the optical axis of the imaging system (length of the optical path), and is an empirical parameter depending upon. For the visible spectral range, is considered to be a fixed value of 1.39 for a steady luminous flame [9]. Substituting () into (1) yields exp( C / T ) 1 KL = ln( 1 ). (3) exp( C / T ) 1 The unknown product KL can be eliminated by rearranging (3) for two different wavelengths, 1 and, 1 a exp(c / 1T ) -1 exp(c / T ) (4) exp(c / 1Ta1 ) -1 exp(c / Ta ) - 1 The unknown temperature T can be determined by solving equation (4), provided that the apparent temperatures T a1 and T a are known for 1 and respectively. In practice, T a1 and T a are obtained by calibrating the system using a standard temperature source (Section.D). Since the third optical path at wavelength 3 is available in the three-colour system, three two-colour temperatures can be calculated using (1) for the three image pairs, i.e., 1 /, 1 / 3 and / 3. The true temperature of the flame can then be determined by the weighted average of the three two-colour temperatures, i.e., where T 3 = i = 1 3 i = 1 W i T i W i = 1 where T i is the two-colour temperature for the i th wavelength pair, W i is the weighting factor for T i. A larger weighting factor is assigned to the temperature of smaller variance whilst a smaller weighting factor to the temperature of larger variance. It should be stressed that, in the present study, temperature measurements are made along horizontal optical paths passing through the flame central line. Due to the fact that the temperature may not be constant along the optical path, the measured temperatures represent line-of-sight (5) (6) averages weighted by the particle concentration across the thickness of the flame. It has been proven that the line-ofsight temperatures were reasonably representative of those in the particle-laden regions of the flame, particularly the root region. B. System Setup Fig. 1 shows the block diagram of the system design. The system consists of a CCD camera, a beam splitting/filtering device, a frame gabber and a microcomputer with application software. The beam splitting/filtering device [10] receives the light of the flame through an optical probe and split the light into three identical beams for three different wavelengths. The high-resolution monochromatic camera (/3in CCD, H 1030V, progressive scanning system) collects the three beams of the light of flame and forms three images which are identical in size but responsible for different wavelengths. The arrangement of the light transmitting, splitting and filtering elements is shown in Fig.. The frame grabber, incorporating with the microcomputer, transfers the image signal into two-dimensional digital images with 8-bit digitisation. The entire imaging system provides a frame rate of 4 frames per second. Flame Flame Image acquisition and digitisation Objective Image processing Temperature calculation Fig. 1. Block diagram of the system. Beam splitting unit Filters Fig.. Arrangement of light transmitting, splitting and filtering elements. Data Presentation Camera lens CCD panel C. Choice of wavelengths The choice of the three wavelengths is one of crucial factors in the design of the multi-colour system. Firstly, the wavelengths must be away from the absorption bands of gas molecules and intermediate radicals in the flame due to the fact that Planck s radiation law only fits the continuous spectra of solid particles [11]. Secondly, the wavelengths should be in a region where the outputs of the camera vary sufficiently in view of sensitivity and signal-to-noise ratio. Thirdly, the bandwidths of the filters should be as narrow as possible to acquire single-wavelength radiation and meanwhile allow enough light passing through. Compromising the factors addressed above has given rise to the selection of the three wavelengths at 550 nm, 63 nm and 1659

3 700 nm with a bandwidth of 40 nm. It should be noted that the use of wavelengths in the visible region also minimises the disturbance of the radiation from the refractory wall of the furnace. D. Calibration and accuracy The system was calibrated to determine the apparent temperatures at the chosen wavelengths ( 1, and 3 )so as to establish the relationships between the apparent temperatures and the corresponding grey-level of the images. A gas-filled tungsten lamp was used as a standard light source. The lamp was pre-calibrated at the wavelength of 66.4 nm [1]. The calibration was carried out by reproducing the geometrical relationship between the imaging system and the flame to be measured. The apparent temperature of the lamp varied from 100 C to 1750 C with an interval of about 50 C. Three banded images of the filament were captured and the grey-levels of the images were averaged. The relationships between the apparent temperatures of the tungsten filament and the grey-levels of the images were then identified. It was noted that the calibration results were sensitive to the settings of the imaging system, particularly the camera shutter speed and iris. The calibration was therefore conducted for different camera settings. Fig. 3 shows the calibration curves for the camera shutter speed of 1/400 second and the iris of 5.6. Ta (ºC) 100 1=700nm =63nm 3=550nm Grey-level Fig. 3 Calibration curves (shutter speed: 1/400 s, iris: 5.6). It can be seen that the relationships between the apparent temperatures of the tungsten filament and the grey-levels of the images appear to be non-linear. It is impossible to produce regression equations without losing the accuracy of the measurement. In practice, therefore, the calibration data are stored in the form of lookup tables, through which a unique relationship between every possible grey-level in a banded-image and the corresponding apparent temperature is defined for a given wavelength. The accuracy of the measurement was verified by applying the system to measure the reference temperature generated by the tungsten lamp over the temperature range between 180 C and 1690 C. The results, as shown in Fig. 4, are the averaged temperatures of 0 measurements for each temperature setting. The relative errors is no greater than 1% for all the three pairs. Measured temperature ( C) 1 / / 3 / Reference temperature ( C) Fig. 4. Comparison between the measured and reference temperatures. III. RESULTS AND DISCUSSIONS To evaluate the performance of the system in an industrial environment, a series of tests was conducted on a 0.5MW th coal-fired combustion test facility. The detailed descriptions on the test facility and system installation have previously been given elsewhere [4][5]. In the tests, a typical pulverised coal fired under various operation conditions including variations in excess air and mass flow rate of coal. Two different sized coals (finely and coarsely ground coal from the same source) were also tested to reveal the impact of coal particle size on the thermal characteristics of the flame. Fig. 5 illustrates the typical temperature distribution of the flame field for the excess air of 10.6% and 3.0%, respectively. The excess air indicates the volume of air exceeding stoichiometric air flow required by the coal fed into the furnace, which is considered to be an important factor influencing the temperature in the flame field. As can be seen, there exist a high temperature region (where T>1540 C) in the flame, the size and the location of which varies significantly with the volume of excess air. The flame with a higher volume of excess air has larger high temperature regions, indicating more intensive thermal reaction taking place. Fig. 6 shows the variations of the maximum, mean and minimum temperatures with excess air. The value at each data point in the figure is the average of 5 continuous readings. Standard deviations of the data are also given as error bars. A slight increasing trend can be observed in both maximum and mean temperatures when the volume of excess air increased. The minimum value reflects more or less the refractory wall temperature and remains relatively stable about 140 C. 1660

4 (a) 10.6% (b) 3.0% Fig. 5. Typical temperature distribution of the flame under different volumes of excess air. >1593 >1567 >1541 >1516 >1490 >1464 >1438 >141 >1387 >1361 Temperature ( C ) 1750 Temperature ( C ) Excess air (%) Fig. 6. Variations of flame temperature with excess air. Fig. 7 depicts the temperature distribution of the flame for two different mass flow rates of coal. It is clear that the sizes and locations of the high temperature regions (where T>1558 C) in the flame are affected by the mass flow rate of coal. For instance, at 64 kgh -1 (i.e., the full furnace load) the two high temperature regions are much greater and distinguished in the comparison with that at 39 kgh -1, resulting in a hotter flame. Fig. 8 shows the variations of the maximum, mean and minimum temperatures with the mass flow rate of coal. As expected, the maximum and mean temperatures increase with the mass flow rate of coal. The difference between the maximum and mean temperatures can be as high as 10 C. A significant fluctuation in the mean temperature occurs under the coal flow rate of 39 kgh -1, suggesting an unstable flame under such a condition. The minimum temperature remains stable and much similar to that of the flame under different in excess air. 35 >1607 >158 >1558 >1533 >1508 >1483 >1459 >1434 >1409 > Mass flow rate of coal (kgh -1 ) Fig. 8. Variation of flame temperature with mass flow rate of coal. Fig. 9 shows the temperature distributions of the flames for the fine and coarse coals. It has been found that the coarse coal produces two distinct high temperature regions (where T>1558 C), resulting in a higher temperature overall the flame. This can also be observed in Fig. 10 which illustrates the variations of the maximum, mean and minimum temperatures. The maximum temperature of the coarse coal flame is as high as C whilst that of the fine coal flame about 1660 C. This difference may be attributed by the fact that the measured temperature is the weighted average of the surface temperatures of the particles. It is known that larger particles burn at a higher temperature than smaller ones because of a lower heat loss rate during combustion [13]. An increase in the number of larger particles would therefore result in a higher weighted temperature. In other words, the coarse coal produces a higher temperature and consequently may lead to more serious fouling and slagging problems due to over-heating or melting of particles. (a) Fine coal (b) Coarse coal Fig. 9. Typical temperature distributions of the two particle sized flames. >1607 >158 >1558 >1533 >1508 >1483 >1459 >1434 >1409 > kgh kgh -1 Fig. 7. Temperature distribution of the flame field for two different mass flow rates of coal. 1661

5 1750 Fine coal Coarse coal ACKNOWLEDGEMENTS Temperature ( C) Fig. 10. Variations of flame temperature with coal particle size. IV. CONCLUSIONS An instrumentation system for the monitoring of temperature and its distribution in a coal-fired flame has been developed based on the three-colour pyrometric and digital imaging techniques. A high resolution CCD camera, together with a sophistic beam split/filter device, enables to receive the light of flame and produce three images for three distinct wavelengths. The three resulting images provide the basis of the determination of three two-colour temperatures. The true temperature of the flame is estimated by the three weighted two-colour temperatures so that the accuracy and reliability of the measurement have been improved. Experimental results obtained on a 0.5MW th combustion test facility have demonstrated that the system is capable of measuring the temperature and its distribution in coal-fired flames under a real industrial environment, and therefore providing very useful information on the thermal- and fluid-dynamic characteristics of the flame under various operation conditions. The authors wish to acknowledge the UK Engineering and Physical Sciences Research Council for providing financial support (GR/S76953/01) for the work presented in this paper. REFERENCES [1] T. H. Fletcher, J. Ma, J. R. Rigby, A. L. Brown and B. W. Webb, Soot in coal combustion systems, Progress of Energy and Combustion Science, vol. 3 pp , [] D. W. Shaw and R. H. Essenhigh, Temperature fluctuations in pulverised coal (P.C.) flames Combustion and Flame, vol. 86, pp , [3] T. Merklein, Optimal firing management with varying fuels by means of combustion diagnoses, VGB PowerTech vol. 78, no. 8, pp , [4] Y. Huang and Y. Yan Transient two-dimensional temperature measurement of open flames by dual spectral image analysis, Transactions of the Institute of Measurement and Control, vol., no. 5, pp , 000. [5] G. Lu, Y. Yan, G. Riley and H. C. Bheemul, Concurrent measurements of temperature and soot concentration of pulverised coal flames, IEEE trans - Inst and Meas, vol. 51, no. 5, pp , 00. [6] Y. A. Levendis, K. R. Estrada and H. C. Hottel, Development of multicolour pyrometers to monitor the transient response of burning carbonaceous particles, Review of Scientific Instruments, vol. 63, no. 7. pp , 199. [7] T. Panagiotou, Y. A. Levendis and M. Delichatsios, Measurements of particles flame temperatures using three-colour optical pyrometry Combustion and Flame, vol pp. 7-87, [8] H. Zhao and N. Ladommato, Optical diagnostics for soot and temperature measurement in diesel engines Progress of Energy and Combustion Science, vol.4, pp , [9] H. C. Hottel and F. P. Broughton, Determination of true temperature and total radiation from luminous flame Industrial and Engineering Chemistry (Analytical Edition), vol. 4, no., , 193. [10] G. Lu and Y. Yan, Vision-Based Multi-Functional Flame Monitoring Apparatus, UK Patent, GB , 00. [11] D. P. Dewitt and G. D. Nutter, Theory and Practice of Radiation Thermometry, John Wiley & Sons Inc (New York), [1] National Physical Laboratory, Certificate of Calibration (Tungsten- Ribbon Lamp Vacuum No. P51C), PM06/LN98/06, [13] P. M. Willson and T. E. Chappell, Pulverised fuel flame monitoring in utility boilers, Measurement and Control vol. 18, no. 66-7,