microwave radiometry for cement kiln temperature ... - CiteSeerX

MICROWAVE RADIOMETRY FOR CEMENT KILN TEMPERATURE MEASUREMENTS Karl D. Stephan1, Lingyun Wang2, and Eric Ryza3 Department of Engineering and Technology, Texas State University, San Marcos, TX, USA Department of Electrical and Computer Engineering, The University of Texas, Austin, TX, USA 3 National Instruments, Inc., Austin, TX, USA 1

2

The maximum temperature inside a cement kiln is a critical operating parameter, but is often difficult or impossible to measure. We present here the first data that show a correlation between cement kiln temperature measured using a microwave radiometer and product chemistry over an eight-hour period. The microwave radiometer senses radiation in the 12-13 GHz range and has been described previously [Stephan and Pearce (2002), JMPEE 37: 112-124]. Submission Date: 10 March 2006 Acceptance Date: 30 January 2007 Publication Date: 5 March 2007 INTRODUCTION The manufacture of hydraulic cement is an important industry, with worldwide annual production exceeding 1.5 billion tons [U. S. Geological Survey, 1999]. The most critical process step involves heating a mixture of unslaked lime (CaO), silica (SiO2), and certain other chemicals in a large rotary kiln to a temperature of about 1350 C. The resulting exothermic reaction produces nodules containing calcium silicate and allied compounds, which are cooled and ground to make finished cement. While the temperature of the hottest part of the kiln must be maintained within fairly close limits to produce cement of uniformly high quality without wasting energy, this temperature is very hard to measure. Corrosion and abrasion inside the kiln destroy contact temperature probes such as thermocouples, and infrared remote temperature sensors work only if the optical path Keywords: Cement industry, ceramics industry, microwave radiometry, temperature measurement 140

between the burn zone and one end of the kiln is sufficiently clear of dust. Many shorter kilns have a strong air draft and are therefore too dusty to allow IR temperature measurements, because the high concentrations of airborne particulates scatter the desired radiation out of the path between the product and the sensor. However, a reliable measure of maximum kiln temperature is still desirable for reasons of product quality, process control, and energy conservation. While airborne particulates can scatter infrared rays severely, the much longer wavelengths at microwave frequencies reduce such scattering to negligible levels. Nyfors reported the measurement of temperature in a cement kiln using a microwave radiometer in 1989, but no quantitative data were presented [Nyfors, 1989]. Following initial experiments with a low-cost radiometric temperature sensor [Stephan and Pearce, 2002], we reported the development of a 12-GHz radiometer for industrial temperature sensing applications [Stephan, 2004] and used it with an offset-parabolic antenna to estimate the temperature of regions well inside a cement

Journal of Microwave Power & Electromagnetic Energy

Vol. 40, No. 3, 2007

FIGURE 1. Kiln and measurement system setup showing radiometer and antenna viewing burn zone of kiln through viewport in firehood. kiln [Stephan, 2005]. The purpose of this paper is to report the first microwave radiometric temperature measurements which correlate well with product chemistry. A brief explanation of the basics of microwave radiometry is in order. All objects at a temperature above 0 K radiate electromagnetic energy in accordance with the Einstein-Planck radiation law which relates the intensity of radiation to its wavelength and the absolute temperature of the object. Strictly speaking, only an ideal uniform-temperature “black-body” radiator produces the full amount of radiation predicted by the Einstein-Planck relation, but in practice, real objects radiate a constant fraction of the black-body radiation. This fraction is called the emissivity of the object, and ranges between 0 and 1. At microwave frequencies, the emissivity of a substance can be calculated from knowledge of its dielectric and magnetic properties. At temperatures above 100 K and wavelengths longer than 1 cm, the EinsteinPlanck relation can be simplified to the following expression: P = kTB

(1)

in which P = maximum power available in a single-mode waveguiding structure coupled to the object, k = Boltzmann’s constant (1.38 x 10-23 J/K), T = (uniform) object temperature (K), International Microwave Power Institute

and B = equivalent noise bandwidth over which the power P is measured. This simple relation says that the absolute temperature of an object is directly proportional to the noise power received from it in a given bandwidth. This equation is the basis of microwave radiometry, which uses precise measurements of microwave noise power to remotely measure the temperature of objects of interest. While most industrial remote temperature sensors use infrared radiometers based on similar principles, these instruments have the drawback that atmospheres with a high concentration of airborne particulates scatter infrared radiation severely. This means that unless the infrared sensor can be physically close to the object whose temperature is desired, the sensor measures the average temperature of the intervening scatterers and not the temperature of the object. However, microwaves having a wavelength of 1 cm or greater are essentially not scattered by airborne particulates, and so can be used over relatively long distances to measure temperatures remotely. This is the approach we have taken in the method we will now describe. METHODS The kiln used in this study is shown in Fig. 1. It was 4.6 m in outside diameter by 58 m long, 141

lined with a 46-cm layer of firebrick. Preheated raw cement is introduced into the cool end of the kiln, which rotates at about 4 RPM to move the product toward the coal-fired burner at the hot end. The maximum temperature is reached about 12-24 m from the hot end in a region called the “burn zone.” The product falls out of the hot end into an enclosed chamber called the firehood, where it passes to a cooler. Access to the kiln interior during operation is limited to a small viewport about 25 cm in diameter next to the burner inlet at the hot end. This viewport was used as the aperture of a single-pixel microwave staring sensor. An offset parabolic antenna about 50 cm in diameter was stationed 3 m behind the viewport, and the antenna’s scalar feed horn was adjusted to produce a near-field focus at the viewport rather than at infinity. We estimate that this system produces a beam inside the kiln with a 3-dB beamwidth of about 4.6 degrees (full width at half maximum). Previous experiments [Stephan, 2004] showed that this beam is sufficiently narrow to distinguish between hotter and cooler portions of the kiln interior. The antenna’s vertical and horizontal position, as well as its aiming direction in elevation and azimuth, were set to maximize the apparent temperature at the antenna output, which ensures that the system is viewing the hottest portion of the kiln accessible to the viewport. The antenna was connected to a noise-injection radiometer operating at a frequency of 12.5 GHz with a bandwidth of 500 MHz and a noise temperature of about 1120 K. Further details about the radiometer are given in [Stephan, 2004]. Since the effective microwave emissivity of the heated product and the antenna-aperture system efficiency were not known, and no independent direct kiln temperature measurements were possible, we were not able to calibrate the system in an absolute sense. As pointed out above, direct contact temperature measurements of the burn zone in a rotary cement kiln are difficult, and installation of temporary contact probes would have required an extremely costly 142

kiln shutdown, which was not feasible for this experiment. Instead, we took data as the kiln operated continuously in its normal production mode. Fluctuations in kiln operating parameters were expected to create some variability in the kiln temperature, and we planned to correlate these variations with variations in product chemistry. As the data presented below will show, these plans succeeded. During an eight-hour period of normal kiln operation, chemical samples of the product were analyzed hourly for a number of characteristics, including the percentage of free lime (CaO). The percentage of free lime is a sensitive measure of overall cement quality and reflects the integrated time-temperature history of the product. All else being equal, if the kiln temperature is too low, the reaction is incomplete, free lime is too high (above about 3%), and the cement’s quality suffers. If the kiln is too hot, the free lime is excessively low (below 1%) and energy is wasted [Bye, 1983]. As we explained, direct contact measurement of kiln temperatures was not possible, and indirect measurements using infrared radiometry were also not feasible because of the large quantity of dust in the kiln. Assuming that the radiometer output was linearly proportional to kiln temperature in the region of interest, we require two calibration points to establish a linear relation between the apparent temperature TA present at the radiometer antenna output and the estimated kiln (or other target) temperature TK. One calibration point was obtained by placing an ambient-temperature absorber (30 C) with known emissivity in front of the antenna. In this situation, both TA and TK can be measured directly, giving us one of the two calibration points required. The second calibration point was chosen so that an apparent temperature TA of 235 C corresponded to a kiln temperature of 1320 C. This point was chosen so that the resulting estimated kiln temperature ranged from about 1350 C to 1470 C. Given that the kiln produced cement with a free lime percentage


Vol. 40, No. 3, 2007

FIGURE 2. Estimated kiln temperature (C) and free-lime percentage during eight-hour field trial of temperature-measurement radiometer. Note that highest temperatures correspond to lowest free-lime percentages. To avoid cluttering the graph, error bars are shown on every tenth temperature data point only. in the acceptable range of 1% to 2.1%, and that according to Bye the lowest kiln temperature for acceptable free lime percentage is about 1320 C, we feel that the choice of this second calibration point produces the best estimate of kiln temperature in the absence of direct calibration measurements. Converting all temperatures to degrees K, the two calibration points we chose produce the equation TK = 6.288TA (K) – 1602.5K.

(2)

This equation was used for all data presented below. The radiometer was independently calibrated with known-temperature microwave loads so that TA was known to an accuracy of approximately ±4 K, which translates into an equivalent kiln-temperature error of approximately ±25 K. During the experimental run, the radiometer output was averaged over one minute for each data point, and once an hour the antenna feed horn was exposed to an ambient-temperature load for two or three International Microwave Power Institute

minutes. These hourly ambient-temperature reference checks were used to compensate for offset drift during the experiment, and are omitted from the data shown below. RESULTS Results of the eight-hour field trial are shown in Fig. 2, with temperature data converted from K to C. The short-term temperature changes on the order of ±15 C are probably due to the inhomogeneous nature of the kiln charge, which consists of irregular semi-liquid lumps of material that randomly cling to and fall from the lining as the kiln rotates. Superimposed on these small variations is a clearly visible secular change which ranged from a low of about 1350 C at t = 1 hr to a high of about 1470 C at t = 3.8 hr. During this run, the free-lime percentage as measured by chemical analysis varied over nearly its entire allowable range, from 1.0% up to 2.1%. The error in the determination of 143

free-lime percentage is approximately ±0.1%, and is shown on the figure together with the temperature error bars of ±25 C. [Jarl, 2006]

ACKNOWLEDGEMENTS

DISCUSSION As Fig. 2 shows, there is good correlation between the measured kiln temperature and the percentage of free lime, with the lowest freelime figures tending to occur at the highest measured temperatures and vice-versa. Other factors that can affect the free-lime percentage besides kiln temperature include composition of the incoming product, kiln speed variations under load, and product feed rate. Since we did not have sufficient data and insight into the process to control for these extraneous factors with a theoretical model of the system, the correlation between measured kiln temperature and product chemistry that we present here is persuasive rather than conclusive. The comparison of microwave radiometric temperature data with directly measured “ground truth” temperature data still remains to be performed. In the case of industrial controls, absolute accuracy is not always as important as repeatability. The ultimate goal is not to measure temperature, but to achieve higher consistency of product quality. Any measurement which enables operators to approach this ideal more closely will be useful, whether or not it represents an absolute physical temperature. We have demonstrated that the radiometer we used was stable and repeatable enough to provide data that appear to be useful for monitoring and adjustment of kiln operating parameters in a production environment. We believe the demonstrated correlation between measured estimates of kiln temperature and the important product parameter of free lime percentage shows that microwavebased radiometric measurements of cement kiln temperatures will be an important addition to instrumentation in this and other energy-intensive industries where critical temperature mea144

surements are currently difficult or impossible to perform.

We gratefully acknowledge the cooperation and assistance of Randy McKee and Jim Jarl of Texas-Lehigh Cement Co., Buda, Texas, for making their kiln and analytical results available for this study. We are also grateful for the support from the U. S. Department of Energy’s I&I Program under contract no. DE-PS3602GO90014. REFERENCES Bye, G. C. (1983). Portland Cement: Composition, Production, and Properties, Pergamon Press, Oxford, U. K. Jarl, J. (2006). Private communication, Jan. 20, 2006. Nyfors, E. (1989). Industrial Microwave Sensors. Artech House, Norwood, MA. Stephan, K. D., and Pearce, J. A. (2002). “Low-Cost Remote Temperature Sensor for Microwave and RF Heating Using Microwave Radiometry.” Journal of Microwave Power & Electromagnetic Energy. 37(2), pp.112-124. Stephan, K. D., Pearce, J. A., Wang, L., and Ryza, E. (2004). “Prospects for Industrial Remote Temperature Sensing Using Microwave Radiometry.” Proc. of the 2004 IEEE MTT-S International Microwave Symposium, session WE4B-1, 651-654, Fort Worth, Texas. Stephan, K. D., Pearce, J. A., Wang, L., and Ryza, E. (2005). “Cement Kiln Temperature Measurements Using Microwave Radiometry.” Proc. of the 2005 IEEE MTT-S International Microwave Symposium, session TU3A-6, Long Beach, Calif. U.S.Geological Survey (1999). Website: http:// minerals.usgs.gov/minerals/ pubs/commodity/ cement/170399.pdf

An online version published at www.jmpee.org.


Vol. 40, No. 3, 2007