Noninvasive ultrasound measurement of temperature distribution in refractories of coal and biomass gasifiers

A gasifier's temperature is the primary characteristic that must be monitored to ensure its performance and the longevity of its refractory. One of the key technological challenges impacting the reliability and economics of coal and biomass gasification is the lack of temperature sensors that are capable of providing accurate, reliable, and long-life performance in an extreme gasification environment. This research has proposed, demonstrated, and validated a novel approach that uses a noninvasive ultrasound method that provides real-time temperature distribution monitoring across the refractory, especially the hot face temperature of the refractory. The essential idea of the ultrasound measurements of segmental temperature distribution is to use an ultrasound propagation waveguide across a refractory that has been engineered to contain multiple internal partial reflectors at known locations. When an ultrasound excitation pulse is introduced on the cold side of the refractory, it will be partially reflected from each scatterer in the US propagation path in the refractory wall and returned to the receiver as a train of partial echoes. The temperature in the corresponding segment can be determined based on recorded ultrasonic waveform and experimentally defined relationship between the speed of sound and temperature. The ultrasound measurement method offers a powerful solution to provide continuous real-time temperature monitoring for the occasions that conventional thermal, optical, and other sensors are infeasible, such as the impossibility of insertion of temperature sensors, harsh environment, unavailable optical path, and more. Our developed ultrasound system consists of an ultrasound engineered waveguide, ultrasound transducer/receiver, and data acquisition, logging, interpretation, and online display system, which is simple to install on the existing units with minimal modification on the gasifier or to use with new units. This system has been successfully tested with a 100 kW pilot-scale downflow oxyfuel combustor, capturing in real-time temperature changes during all relevant combustion process changes. The ultrasound measurements have excellent agreement with thermocouple measurements, and appear to be more sensitive to temperature changes before the thermocouples response, which is believed to be the first demonstration of ultrasound measurements segmental temperature distribution across refractories.

NONINVASIVE ULTRASOUND MEASUREMENT OF TEMPERATURE DISTRIBUTION IN REFRACTORIES OF COAL AND BIOMASS GASIFIERS by Yunlu Jia A dissertation submitted to the faculty of The University of Utah in partial fulfillment of the requirements for the degree of Doctor of Philosophy Department of Chemical Engineering The University of Utah August 2015 Copyright © Yunlu Jia 2015 All Rights Reserved The Uni v e r s i t y of Utah Graduate School STATEMENT OF DISSERTATION APPROVAL The dissertation of Yunlu Jia has been approved by the following supervisory committee members: Mikhail Skliar Chair 05-18-2015 Date Approved Anthony Butterfield Member 05-04-2015 Date Approved Douglas A. Christensen Member 05-04-2015 Date Approved Eric Eddings Member 05-04-2015 Date Approved Kevin Whitty Member 05-04-2015 Date Approved and by Milind Deo Chair/Dean of the Department/College/School o f ______________Chemical Engineering and by David B. Kieda, Dean of The Graduate School. ABSTRACT A gasifier's temperature is the primary characteristic that must be monitored to ensure its performance and the longevity of its refractory. One of the key tech­nological challenges impacting the reliability and economics of coal and biomass gasification is the lack of temperature sensors that are capable of providing accurate, reliable, and long-life performance in an extreme gasification environment. This research has proposed, demonstrated, and validated a novel approach that uses a noninvasive ultrasound method that provides real-time temperature distribution monitoring across the refractory, especially the hot face temperature of the refractory. The essential idea of the ultrasound measurements of segmental temperature distribution is to use an ultrasound propagation waveguide across a refractory that has been engineered to contain multiple internal partial reflectors at known locations. When an ultrasound excitation pulse is introduced on the cold side of the refractory, it will be partially reflected from each scatterer in the US propagation path in the refractory wall and returned to the receiver as a train of partial echoes. The temperature in the corresponding segment can be determined based on recorded ultrasonic waveform and experimentally defined relationship between the speed of sound and temperature. The ultrasound measurement method offers a powerful solution to provide continuous real-time temperature monitoring for the occasions that conventional thermal, optical, and other sensors are infeasible, such as the impossibility of insertion of temperature sensors, harsh environment, unavailable optical path, and more. Our developed ultrasound system consists of an ultrasound engineered waveguide, ultrasound transducer/receiver, and data acquisition, logging, inter­pretation, and online display system, which is simple to install on the existing units with minimal modification on the gasifier or to use with new units. This system has been successfully tested with a 100 kW pilot-scale downflow oxyfuel combustor, capturing in real-time temperature changes during all rele­vant combustion process changes. The ultrasound measurements have excellent agreement with thermocouple measurements, and appear to be more sensitive to temperature changes before the thermocouples response, which is believed to be the first demonstration of ultrasound measurements segmental temperature distribution across refractories. iv CONTENTS ABSTRACT......................................................................................................................... iii LIST OF F IG U R E S ............................................................................................................viii LIST OF TA B L E S .............................................................................................................. xv ACRONYMS....................................................................................................................... xvi CHAPTERS 1...... INTRODUCTION..................................................................................................... 1 1.1 Background and Motivation......................................................................... 1 1.1.1 Gasficiation and Gasifier....................................................................... 1 1.1.2 Gasifier Lining Wear and Fa ilu re ....................................................... 3 1.2 Current Temperature Measurement Techniques..................................... 6 1.2.1 Direct Measurements.............................................................................. 6 1.2.2 Indirect Measurements......................................................................... 7 1.2.3 Noninvasive Measurements................................................................ 9 1.3 Organization ....................................................................................................... 11 2. NONINVASIVE ULTRASOUND MEASUREMENTS OF SEGMENTAL TEMPERATURE DISTRIBUTION..................................................................... 12 2.1 Ultrasound .......................................................................................................... 12 2.1.1 Basic Instrumentation............................................................................ 13 2.1.2 Ultrasound Propagation Speed............................................................ 14 2.1.3 Impedance and Attenuation................................................................ 15 2.1.4 Reflection and Transmission................................................................ 17 2.2 Physical Basis of the US-MSTD Method ..................................................... 17 2.3 Method ................................................................................................................ 23 2.3.1 Structure US Propagation P a th ............................................................ 24 2.3.2 Acquisition of Echo Waveforms......................................................... 25 2.3.3 Signal Processing..................................................................................... 26 2.3.4 Temperature Dependence of the Ultrasound TOF......................... 28 2.3.5 The Temperature Distribution Estimation....................................... 28 2.3.5.1 Piecewise Constant Distribution.............................................. 28 2.3.5.2 Piecewise Linear Distribution .................................................. 29 2.3.6 Estimation of Heat Flux......................................................................... 29 2.3.7 Parametrization with Thermal Conductivity Model..................... 30 3. LOW TEMPERATURE LABORATORY EXPERIMENTS............................ 32 3.1 Cementitious Waveguide Partial Reflector Structures ............................ 32 3.1.1 High Water-Cement Ratio Sample with Various Curing Times . 35 3.1.2 Low Water-Cement Ratio Sample with Various Curing Times . . 36 3.1.3 Medium Water-Cement Ratio Sample with Long Curing Time . 38 3.1.4 Summary of Partial Internal Reflection Structures ....................... 42 3.1.4.1 Composition ................................................................................... 42 3.1.4.2 Water-Cement R a t io ..................................................................... 42 3.1.4.3 Air Bubbles ..................................................................................... 44 3.1.4.4 The Partial Curing Time .............................................................. 44 3.1.4.5 Number and Spacing of Partial Reflections ............................ 44 3.1.4.6 Sample Length................................................................................ 45 3.1.4.7 Consistency of Sample Properties ............................................ 45 3.2 Experiments of Temperature Measurements ............................................ 46 3.2.1 Structured Cementitious Waveguide ................................................ 46 3.2.2 TOF Acquisition from Ultrasound Waveform ................................ 47 3.2.3 Signal Processing ..................................................................................... 49 3.3 Results ................................................................................................................... 53 3.3.1 The Calibration the SOS and Temperature ....................................... 53 3.3.2 Nonuniform Temperature Distribution ............................................ 53 3.3.2.1 Piecewise Constant Distribution .............................................. 54 3.3.2.2 Piecewise Linear Distribution .................................................. 55 3.3.3 Parametrization with 2D Thermal Conductivity Model .............. 56 3.4 More Cementitious Samples SOS Measurements..................................... 58 4. HIGH-TEMPERATURE LABORATORY EXPERIMENTS......................... 61 4.1 Alumina Refractories ....................................................................................... 61 4.1.1 Castable Alumina ................................................................................... 61 4.1.2 Machinable Alumina.............................................................................. 62 4.2 Experiments....................................................................................................... 65 4.2.1 Partial Reflectors Structures................................................................ 65 4.2.2 Experimental S e tu p ................................................................................ 66 4.3 Results ................................................................................................................... 68 4.3.1 Signal Processing..................................................................................... 68 4.3.2 The SOS versus Temperature Calibration ....................................... 71 4.3.3 Elastic Modulus ....................................................................................... 72 4.3.4 Scaled Calibration ................................................................................... 75 4.3.5 Nonuniform Temperature Experiments............................................ 77 4.3.6 Nonuniform Temperature R e su lts..................................................... 78 4.3.6.1 Piecewise Constant Temperature Distribution..................... 78 4.3.6.2 Piecewise Linear Temperature Distribution ......................... 80 4.3.7 Real-Time Nonuniform Temperature Distribution ....................... 82 4.4 Refractory Degradation Laboratory Tests .................................................. 84 vi 5. ROBUST ULTRASOUND SIGNAL PROCESSING..................................... 86 5.1 The Time of F lig h t............................................................................................ 86 5.1.1 Cross-correlation of W aveform............................................................ 86 5.1.2 Zero Trigger Reference ......................................................................... 89 5.1.3 Cross-correlation of Envelope of Waveform .................................. 91 5.2 Anisotropic Diffusion Filte r............................................................................ 93 5.3 Experiments and Results ................................................................................ 95 6. PILOT-SCALE OXYFUEL COMBUSTOR EXPERIMENTS.......................103 6.1 Pilot-Scale Oxyfuel Combustor.....................................................................103 6.2 US-MSTD System Design and Implementation....................................... 104 6.3 Experimental Conditions................................................................................ 107 6.4 Experiments and Results................................................................................ 108 6.4.1 Natural Gas Preheating......................................................................... 108 6.4.2 Steady State Natural Gas Combustion..............................................108 6.4.3 Transition from Natural Gas to Coal ................................................ 110 6.4.4 Response to Decreased Flow Rate of C o a l....................................... 113 6.4.5 Stable Coal Combustion ....................................................................... 115 6.4.6 Temperatures at Distal E nd ...................................................................118 6.5 Discussion ............................................................................................................ 119 6.5.1 Couplant..................................................................................................... 119 6.5.2 Alumina Refractory Waveguide......................................................... 119 7. REFRACTORY WAVEGUIDE SELECTION .................................................. 122 7.1 Potential Refractory Waveguide ................................................................... 122 7.1.1 Zirconia....................................................................................................... 122 7.1.1.1 Cubic Zirconia ................................................................................ 123 7.1.1.2 Zirconia PSZ ................................................................................... 123 7.1.1.3 Zirconia T Z P ...................................................................................124 7.1.2 Nitrides....................................................................................................... 125 7.1.3 Carbides ..................................................................................................... 125 7.1.3.1 Silicon Carbide .............................................................................. 125 7.1.3.2 Tungsten Carbide ......................................................................... 126 7.2 Ultrasound Properties of Zirconia ................................................................ 126 7.3 Experiments ....................................................................................................... 129 7.4 The SOS Dependence on Temperature ....................................................... 131 8. CONCLUSIONS AND FUTURE W ORK ......................................................... 134 REFERENCES..................................................................................................................... 137 vii LIST OF FIGURES Figure Page 2.1 Block diagram of basic components of an ultrasonic measurement system to generate and detect ultrasonic waves............................................................... 13 2.2 Ultrasound measurements of temperature distribution in the refractory. (A) Refractory material contains embedded planes of scattering material. (B) Layered refractory. (C) Refractory insert with geometric changes in the ultrasound propagation path creates partial back scattering. Left panel shows an ultrasound excitation pulse and the train of partial echoes pro­duced by internal partial ultrasound reflectors. Right panel illustrates an engineered ultrasound waveguide/insert - with internal back scatterers, layers structure, or geometrical changes - embedded into the gasifier refractory. ..................................................................................................................... 24 2.3 The ATOF between echo waveforms at different temperatures is calcu­lated by cross-correlation with a reference waveform acquired at 20 °C. . . 27 3.1 Ultrasound pulse-echo response for the samples with internal interfaces. (A) The pulse-echo ultrasound response of two samples fabricated from the identical cementitious material. One of the samples (Sample B; shown in inserted photograph) contains embedded ultrasound scatterers at the midpoint of ultrasound propagation path, which produces partial reflection (red line). (B) Ultrasound pulse-echo response for the sample with two internal interfaces obtained by sequentially casting three layers of identical formulation and allowing time for a partial cure to occur prior to pouring the next layer................................................................................. 33 3.2 Ultrasound response for samples with the single internal interface ob­tained by allowing the first cementitious layer to cure for 1 hour. Inserts schematically depict the locations creating the acquired echoes: The first echo is the partial internal reflection from the interface of the two layers; the second echo is from the end of the sample. Round-trip echoes from partial internal reflection and the end are also shown...................................... 36 3.3 Ultrasound response for samples with the single internal interface ob­tained by allowing the first cementitious layer to cure for 45 minutes (A) and 15 minutes (B). Inserts schematically depict the locations creating the acquired echoes in (A) both the partial internal reflection and the end of the sample. The reflection from the interface is not well defined and varies with the change in the transducer position in (B).................................. 37 3.4 Ultrasound response for samples with the two internal interfaces ob­tained by allowing the first cementitious layer to cure for 15 minutes, 1 day for the second layer, and at least 1 day for the final third layer. Water-cement ratio equal to 0.36 was used. Inserts schematically depict the location creating the acquired echoes. Three echoes are clearly visible, of which the first two are from the internal interfaces created by multiple layers cast sequentially........................................................................................... 3.5 Ultrasound response for samples with the two internal interfaces ob­tained by allowing the first cementitious layer to cure for 45 minutes (A) and 1 hour (B), both 1 day for the second layer, and at least 1 day for the final third layer. Water-cement ratio equal to 0.36 was used. Echoes from the second interface and the end of the sample are clearly visible. However, the signal from the first interface is rather complex, likely indicating entrapment of air at the interface..................................................... 3.6 Ultrasound response for samples with the two internal interfaces ob­tained by allowing the two initial cementitious layers to cure for 1 day each. The last layer was cured for at least 1 day. Water-cement ratio equal to 0.44 was used. Echoes from both interfaces are well defined, with relatively high SNR......................................................................................... 3.7 Ultrasound response for a sample with a single internal interface ob­tained by allowing the first cementitious layer to cure for 1 day (A) and 2 days (B). Water-cement ratio equal to 0.44 was used. Echoes from the interface and the end of the sample are well defined, with a relatively high SNR...................................................................................................................... 3.8 Ultrasound waveform collected for samples with a single internal in­terface obtained using Portland Fortified (A) and Rapid Set® cast (B). Echoes from the interface and the end of the sample are well defined, with a relatively high SNR...................................................................................... 3.9 Two groups of RS samples tested in the water bath and air show great variability in both the SOS versus temperature relationships and density measurements............................................................................................................. 3.10 Ultrasound waveforms acquired at different temperatures illustrate the TOF shift with temperature.................................................................................... 3.11 Experimental setup of 4 inches, 4 segments cementitious sample for low temperature measurements.................................................................................... 3.12 Comparison of the estimated A TOF (offsets) at different temperatures obtained by cross-correlating the waveforms (panel (A)) or envelopes of the waveforms (panel (B))....................................................................................... 3.13(A) Typical waveforms of ultrasound echoes created by the interface between different layers Li of the cementitious sample (insert) and the sample-air interface at its distal end. The measurements were acquired at the reference temperature of 20 °C. (B) Envelopes of echo waveforms collected at different temperatures....................................................................... 38 39 40 41 43 46 47 48 50 51 ix 3.14 The calibration curves for the SOS as a function of temperature for all four layers of the sample were obtained using envelope cross-correlation data analysis methods. The shown linear fit SOS = f(T ) is based on data for all four layers. The shaded areas show the 95% confidence interval for the obtained linear fit........................................................................................... 54 3.15 Temperature distributions from ultrasound and thermal model parame-terizations compared with thermocouple measurements. (A) Estimated temperature distributions based on piecewise constant and piecewise linear parameterizations are compared with the measurements of the surface temperature obtained with thermocouples attached in the mid­dle of each segment. (B) The parametrization based on 2D thermal conductivity model can be used to estimate the temperature distribution along the centerline and the surface of the waveguide.................................... 55 3.16 Comparison of thermocouple and ultrasound measurements of temper­ature distribution in the sample heated from the bottom................................ 58 3.17 The SOS as function of temperature for all six cementitious samples shows linear correlations........................................................................................... 60 4.1 The ultrasound waveforms acquired from 4-inch and 5-inch alumina samples. Echo signals are seen at approximately 10 and 12 microseconds. 63 4.2 The ultrasound waveform for 1-inch x 12-inch alumina rod collected by transducer with frequency 1 MHz. The distal end echo shows between 140 and 160 ^s, followed with two more echoes................................................ 64 4.3 Possible modes from longitudinal waves at the lateral surface in the long cylindrical rod of alumina......................................................................................... 65 4.4 Three 3/32-inch holes were drilled into the precast alumina rod at 2, 4, and 6 inches away from the distal end of the rod............................................... 65 4.5 The ultrasound echo waveforms for 1 inch x 12 inches alumina rod with drilled holes to produce partial reflections. Plots show the response obtained with transducers of different central frequencies............................. 66 4.6 An overview of the actual lab-scale ultrasound high-temperature mea­surement system for SOS versus temperature calibration tests..................... 67 4.7 The ATOF as a function of temperature was calculated by envelope cross­correlation with the reference waveform acquired at 20 °C............................. 69 4.8 The SOS vs. temperature calibration curve obtained for all tested data points using the envelope cross-correlation method is shown in (A). The results in (B) are shown with and without corrections for the thermal expansion. The error bars indicate the 95% confidence interval for each data point........................................................................................................................ 71 4.9 The Young's modulus as a function of temperature calculated based on our experimental data is shown in (A) and compared with the data from castable alumina refractories from the Gault et al. study [32]. The Young's modulus for sintered alumina from literature [23], [105] is shown in (B). . 73 x 4.10 The Young's modulus of elasticity (normalized to room temperature value £ 0) as a function of the temperature for our experimental results and the sintered alumina material from the literature [23], [105].................. 74 4.11 Both equal percentage and polynomial fitted results of SOS as a function of temperature in high temperature range are compared with experimen­tal data. Equal percentage fitting has a better coefficient of determination. 76 4.12 The experimental setup for nonuniform temperature distribution using US-MSTD system......................................................................................................... 77 4.13 The illustration of calculation of TOF in the i-th segment for determina­tion of nonuniform temperature distribution...................................................... 79 4.14 Nonuniform temperature distribution at steady state. Central tem­peratures from ultrasound measurement compared with thermocouple measurements at the refractory surface when furnace temperature is set to 500 °C, (A) and 1000 °C, (B).................................................................................. 81 4.15 Real-time nonuniform temperature distribution monitored by US-MSTD system captured both heating and cooling process tested in lab-scale furnace setup to 1200 °C............................................................................................. 83 5.1 TOF estimation for two-layer cementitious sample using threshold cross­ing (A), peak of the signal (B), cross-correlation (C) and envelope cross­correlation of waveforms (D).................................................................................... 87 5.2 Ultrasound test of the aluminum shows multiple echoes corresponding to several round-trip travels of the test pulse through the sample............... 88 5.3 The calculated cross-correlation between the first two echoes gives the time shift needed for the best alignment of the two waveforms. This shift is equal to the TOF of the test pulse........................................................................ 89 5.4 The correction of the electrical trigger time is found using the cross­correlation method with an aluminum standard............................................... 90 5.5 Top panel (I) shows echo waveforms produced by four echogenic features drilled along the length of the alumina waveguide and the reflection from its distal end. The corresponding envelopes are shown in the center panel (II). Bottom panel (III) shows echo waveforms and its envelopes produced from interface and distal end of cascaded cementitious sample. 96 5.6 Examples of waveforms with the anisotropic diffusion filter treatment. (I) The original echo waveform, shown in blue, was produced by PR3 echogenic feature; its envelope is shown in A. The application of the filter for 100,500,1000, and 1500 iterations produced filtered envelopes respec­tively shown as B, C, D, and E. Maximum values of filtered envelopes are marked with red triangles. (II) Two echoes, shown as blue traces, were produced by PR1 feature, which were acquired at the same conditions but different times. Their envelopes show variability of the maximum value indicated by green triangles. After applying 3000 iterations, the filtered envelopes of both waveforms are shown in the middle of the panel. Red arrows show that the peak values coincide after filtering......... 98 xi 5.7 Real-time temperature monitoring results comparisons of TOF estima­tion treated without and with anisotropic diffusion filter................................ 6.1 The 100 kW pilot-scale downfired oxyfuel combustor...................................... 6.2 The schematic overview of US-MSTD installation on pilot-scale OFC. A 1x 12-inch alumina waveguide was engineered at predetermined spatial locations along its length located 6, 4, 2, and 1 inches-marked as z1, z2, z3, and Z4, from the hot distal end (DE) of the rod. The data acquisition and interpretation systems provide real-time temperature distribution across the refractory and compare with thermocouple measurements. . . . 6.3 Port-mounted waveguide retention system......................................................... 6.4 Ultrasound measurements with piecewise constant assumption and ther­mocouple measurements for temperature distribution across the refrac­tory as the OFC is being preheated by the natural gas combustion without the electric heaters. Thermocouple measurements were not recorded until 1 hour after the campaign started. ............................................................ 6.5 Ultrasound measurements with piecewise constant parameterization and thermocouple measurements show a same temperature trend during stable natural gas combustion.................................................................................. 6.6 Temperature distribution across the refractory based on piecewise linear assumption during stable natural gas combustion............................................ 6.7 Temperature distributions obtained from ultrasound measurements us­ing two different parameterizations and thermocouple measurements during stable natural gas combustion are compared at all echogenic features' locations and the hot distal end. A great agreement of temper­ature change is shown for all the methods. In addition, both ultrasound measurements appear to be more sensitive to temperature changes than thermocouples............................................................................................................... 6.8 Segmental temperature profiles obtained from both ultrasound with piecewise constant assumption and thermocouple measurements during transition from natural gas to coal combustion. The plots show good agreement with two methods, and ultrasound responses for temperature change are faster than thermocouples................................................................... 6.9 The real-time temperature profiles across the refractory obtained from ultrasound based on piecewise linear assumption and thermocouple measurements during transition from natural gas to coal combustion captured the temperature responses for fuel transition.................................... 6.10 Temperature profiles obtained from the US-MSTD approaches and ther­mocouples during transition from natural gas to coal combustion show comparable temperature results at echogenic features' locations and the hot distal end................................................................................................................. 104 101 105 106 109 109 110 111 112 112 113 xii 6.11 Segmental temperature responses from ultrasound measurements based on piecewise constant assumption and thermocouple measurements are captured during the process of a change feed rate of coal combustion, fol­lowed by the fuel transferred from natural gas to coal. Both temperature measurements present the same trend on temperature changes................... 6.12 Temperature response from ultrasound based on piecewise constant assumption and thermocouple measurements for a change feed rate of coal combustion............................................................................................................ 6.13 Temperature responses from ultrasound and thermocouple measure­ments for a change feed rate of coal combustion at echogenic features' locations and the hot distal end show an excellent agreement of these two approaches, while ultrasound responses are more sensitive than thermocouples............................................................................................................... 6.14 Ultrasound measurements using piecewise constant assumption and thermocouple measurements for temperature distributions during stable coal combustion. Intervals of ultrasound measurements flat lined cor­respond to the application of fresh ultrasound couplant at the transducer-waveguide interface. The vibration of ultrasound measurement is stronger than previous acquired processes due to significant reduction of ultra­sonic waveform strength........................................................................................... 6.15 Ultrasound measurements using piecewise linear assumption and ther­mocouple measurements for temperature distributions during stable coal combustion..................................................................................................................... 6.16 Temperature profile comparisons between ultrasound and thermocouple measurements at echogenic features' locations. The overall temperature distributions from both measurements are comparable, while stronger vibration in ultrasound measurement is observed than in other acquired processes........................................................................................................................ 6.17 The temperature profiles obtained from the US-MSTD system at the hot distal end are verified with the thermocouple measurements at port 2. . . 6.18 SEM images of alumina grains distribution under over 13000x magnifi­cation for alumina refractory Rescor 960 sample without heat treatment (A) short-term heat treatment (B) and long-term heat treatment (C)- pilot-scale tested alumina waveguide................................................................... 7.1 Phase diagram for the zirconia rich portion of the zirconia-yttria system reproduced from the study of Scott [92]................................................................ 7.2 Yttria stabilized zirconia rod as received from Astro Met, Inc....................... 7.3 The ultrasound echo waveforms obtained with this zirconia rod. The response at 9 ^s is produced by the distal end of the sample and is preceded by four partial internal echoes produced by the drilled holes. Panels (a) and (b) show responses obtained with transducers having 5 and 10 MHz central frequencies.............................................................................. 114 115 116 116 117 118 121 124 126 114 127 xiii 7.4 Unexpected echoes in Figure 7.3 are caused by multiple ultrasound reflections between the transducer and the zirconia waveguide...................128 7.5 Waveguide color changes after heat treatment. Top: Color gradient indicates oxidization of the hottest area. Bottom: The whole sample became white after high temperature sintering at a uniform temperature distribution................................................................................................................... 129 7.6 The placement of thermocouples used during real-time temperatures monitoring..................................................................................................................... 130 7.7 Echo waveforms collected at different temperatures. The strength of the signal changes with temperature as crystal structure changes in the waveguide material at different temperatures. Strong temperature dependence of the SOS is evident as temperature increases from room temperature to 1150 °C............................................................................................... 131 7.8 The SOS versus temperature relationship comparison of waveguides of alumina (primary y-axis) and zirconia (secondary y-axis) (A). Both piecewise linear and polynomial fitting models are used to illustrate the SOS and temperature calibration results (B)........................................................132 7.9 The calculated Young's modulus as function of temperature based on experimental SOS results........................................................................................... 133 xiv LIST OF TABLES Table Page 3.1 Summary of cementitious samples for SOS te s ts .............................................. 59 5.1 Signal delay after trigger for different materials .............................................. 91 5.2 Timing errors and corresponding temperature differences............................100 7.1 AmZirOx 86 zirconia properties............................................................................133 ACRONYMS The National Energy Technology Laboratory (NETL) One-dimensional (1D) Two-dimensional (2D) Three-dimensional (3D) Thermocouple (TC) Resistance temperature device (RTD) Coefficient of thermal expansion (CTE) Fiber bragg grating (FBG) Ultrasound (US) Nondestructive testing (NDT) Speed of sound (SOS) Time of flight (TOF) Ultrasound measurements of segmental temperature distribution (US-MSTD) Internal diameter (I.D.) Signal-to-noise ratio (SNR) Oxyfuel combustor (OFC) Partial stabilized zirconia (PSZ) Tetragonal zirconia polycrystal (TZP) Yttria stabilized zirconia polycrystal (Y-TZP) Surface acoustic wave (SAW) CHAPTER 1 INTRODUCTION 1.1 Background and Motivation 1.1.1 Gasficiation and Gasifier Modern day gasification technology was first introduced by the oil industry to process low-value petroleum and its by-products. Since the 1970s, the gasification process has adapted to a variety of carbon-based feedstocks, such as coals, low-cost, widely available petroleum coke, biomass, and agriculture (solid) waste [43]. Gasification technology has been predicted to be a major source of clean-fuel technology for the coming future. Gasification involves the thermal beakdown of carbon rich materials in a hot, reactive environment to produce synthesis gas, or syngas, which is rich in hydrogen (H2) and carbon monoxide (CO). The produced syngas can then be used as an alternative feedstock for many chemical processes, such as those used in methanol, butanol, dimethyl ether, diesel, and gasoline production. When used in power generation, the syngas produced by coal or biomass gasifiers is burned as a fuel in carbon neutral or carbon capture-ready power generation. Gasification plants tend to use less makeup water and produce less solid waste and airborne pollutants than typical coal combustion-based plants. Gasifiers mainly consume oxygen (O2), instead of air, which is more efficient and economical during the conversion of the carbon feedstock, carbon dioxide (CO2) separation, capture, and sequestration [56]. The ability of eliminating most air pollutants and potential greenhouse gases makes gasification a more environmen­tally sustainable technology for the energy plants compared with traditional coal combustion. Gasification-based electric power plants are operating commercially in many countries. There are three major types of commercially-available gasifiers based on their 2 feedstock and end-product requirements: fixed bed gasifiers, fluidized bed gasi-fiers, and entrained bed gasifiers. Gasifiers can also be grouped based on their ash treatment: slagging gasifier and nonslagging gasifier. A nonslagging gasifier normally operates at temperatures below the ash melting point of the feedstock, which is about 500-600 °C. This type of gasifier is much easier to operate, since the wall temperature is significantly low so that it does not require a refractory lining. However, the transformations of the inorganic to ash form in nonslagging gasifier sometimes cause a heavy metal leaching problem [109]. In addition, due to its low operating temperature, air-blown gasification is more favorable than oxygen-blown gasification, which would cause the low efficiency of carbon capture [77]. A slagging gasifier operates at a temperature that is higher than the feedstock ash melting point. The ashes are present in a liquid form (molten slags) that can flow down and be removed regularly. This type of gasifier offers advantages, such as high gas production capacity, lower steam consumption, absence of tars and oils in the product steam, and relatively easy disposal of waste during operation. Examples of the slagging gasifiers include (1) for the fixed-bed system: Lurgi Dry-Ash and British Gas/Lurgi gasifiers; (2) for the fluidized bed system: KBR Transpoint, High Temperature Winkler, and ICC/CAS AFB gasifiers; (3) for the entrained-bed system: GE Energy (formerly Chevron Texaco), Shell, and CB& I E-GasTM gasifiers [25]. However, this high efficient gasification process generally involves high temperatures and pressures, and aggressive chemical composites which would results in an extraordinarily severe environment for the structural components of the slagging gasification system [21]. Typical operating conditions are a temperature of 1300 °C (slurry feed) or 1500­1800 °C (dry feed) and pressure of 0.15 to 2.45 MPa for fixed bed gasifiers, 900­1200 °C and up to 2.94 MPa for a fluidized bed, and 1200-1600 °C and 2-8 MPa for entrained flow gasifiers [70]. The overall reaction of gasification is shown as follows: C + H2O(steam) + O2(shortage) - > CO + H2 + CO2 + minority gases+by-products + heat. The minority gases in a slagging gasifier primarily consist of H2O (steam), H2, 3 CO, and CO2, and small amounts of methane (CH4), N2, NH3, and H2S may also exist. The by-products, in ash or slag form, are basically mixtures of various oxides, such as SiO2, Al2O3, FeO, CaO, MgO, Fe2O3, MnO, Na2O3, K2O, TiO3, TiO2, and P2O5 [9]. The adopted temperatures and pressures in a slagging gasifier generally depend on the specific gasification process and the reactivity of the feedstock. Increasing gasification temperature and pressure can usually increase operational efficiency and reduce the size of gasifier. The high temperatures are not permissible for the vessel shell of the gasifier, and thus the shell is generally protected by a refractory lining system. 1.1.2 Gasifier Lining Wear and Failure The refractory lining system provides resistance to extreme operating conditions and insulates the gasifier from energy loss. The lining system is composed of refractory linings, generally two to six layers of bricks with proper mortar joints and cooling systems. The emphasis on liner materials is on the hot-face refractory materials, which are exposed to the most aggressive environment. At these elevated temperatures and pressures, the processing gases and by-products, especially in slag form, attack the refractory lining in various ways. The steam can oxidize the iron-containing metal shell and cause cracking and spalling problems in the shells during heating cycles. Such problems become very severe at high temperature levels. At high temperatures, steam can affect refractory materials by causing the extraction of soluble oxides or hydroxides, resulting in the reduction of refractory strength and erosion resistance. The feedstock ashes melt into fluid slags in the high temperature and are maintained as a liquid state in slagging gasifiers. Slags run down the wall, flow over the bed of the gasifier, and pass through a slag tap to a quench tank where slags are removed continuously. This concept provides an easy method for waste disposal and creates minimal environmental problems. This slag penetrates and reacts with the refractory, causing degradation and corossion resulting from a combination of mechanical stresses and thermal expansion mismatches. The gasifier operating conditions and slag chemistry have a significant impact on the performance of 4 refractories. Refractory linings under slagging coal gasification experience the combination of a chemical (reaction and phase change) and physical (erosion) effect, followed by a failure in required performance [95]. The molten slag is corrosive to the hot face refractory of the gasifier. The corrosion process involves three mechanisms: dissolution, penetration, and erosion. Refractories exposed to slag might be irregularly dissolved in slag. This causes continuous loss in mass and thickness of the refractory linings. Since refractory material is porous, slag may penetrate into refractories that reacts and chemically dissolves the refractory material, causing degradation and corrosion. The penetration of slags into refractories depends on the porosity and the temperature of the refracoties. Chemical erosion creates local microcracking, weakens material mechanical properties, and causes cracking and spalling problems in the refractories. In addition, the combination of thermal expansion mismatch and boundary confinement between the refractory brick and the slag causes cracking and joint failure in the lining. This leads to gradual development of several microstructural cracks inside the refractory surface, which eventually merge together into bulk removal of material of refractory walls. High-alumina and high-chromia dense refractories are usually used in slagging gasifiers. Erosion is not a governing destructive factor for hot face refractories in slagging gasifiers. However, the chemical dissolution of refractory walls degrades the material mechanical strength and irregular dissolution of the material leads to fatigue damage. Fatigue crack growth leads to further slag penetration that, in turn, causes more chemical dissolution. The slag and harsh gasifier environment are core issues challenging the gasifier lining's refractory service life and key barriers to generalize commercialization of gasification technology. The expected refractory lining should have a reliable life of at least 3 years. Current refractories last 4-18 months, which has yet to meet the desired service life [27]. The replacement cost of a failed refractory lining is over 1 million U.S. dollars, both in terms of the material cost (depending on gasifier size and rebuild requirements) and also in terms of 2-3 weeks lost production time. Research for improving refractory products that can withstand these environments 5 for a continuous, efficient, and reliable gasification process has not stopped [9], [68]. The National Energy Technology Laboratory (NETL) has worked with Harbison- Walker Refractory Company to develop a new refractory, which was designed specifically for longer service life in gasifier. This patented technology has been on market as Aurex® 95P. This new refractory has been tested at several commercial gasifier sites in the United States and showed significantly improved performance relative to other commercially-available materials [8]. Early detection of initial damage in the refractory walls is necessary to prevent unscheduled shutdown of a gasification plant. A real-time diagnosis tool with capabilities of generating early warnings is critical for extending the refractory's service life. The pivotal idea for damage detection and prediction of the refractory wall in particular and the entire gasification system in general is built upon the fact that a local anomaly is likely to influence the temperature gradient in the refractory wall due to changes in the thermal impedance. Current degradation monitoring of slagging gasifiers provides a variety of fault diagnostic methodologies that are primarily built upon microstructural analytical models of damaged refractory bricks. However, lack of real-time sensor-based information is one of the major technical challenges for accurate refractory damage diagnosis. An optimum refractory monitoring system should involve information of ther­momechanical material and system behavior, corrosion behavior, and their in­teraction in the gasification environment. Temperature profile is one of most important sensor-based parameters required for the refractories over an extended period of operation. The mechanical properties of refractories, including com­pressive strength, tensile strength, Young's modulus, and creep rate, are primarily temperature- and load history-dependent. As the critical temperature is reached, refractories lose their strength. This critical temperature depends on the melting point and refractoriness of the materials. The strength loss results in excessive deformation and the loss of load-carrying capability with the consequent loss in integrity of the lining system [21]. Both steady state spatial temperature distributions and dynamic temperature profiles undergo different types of changes that provide important information 6 for detection and identification of an anomalous plant condition. Temperature variations across the cross-section of the gasifer lining at different vertical levels can be used to predict the slag flow and refractory dissolution, something difficult to monitor during the gasification. Using the integrated model of heat transfer, heat loss and heat flux can be calculated based on refractory temperature profiles. [19]. Furthermore, since a gasifier reactor uses a multilayered lining, temperature profiles are important in understanding the thermal interaction between complex layers, choosing proper combinations of materials that will prevent overheating of refractories and improve energy conversion efficiency. Temperature along the refractory wall is also a key input for the 2D and 3D stress analysis for the gasifier support structures, such as the shell and anchors. 1.2 Current Temperature Measurement Techniques 1.2.1 Direct Measurements The current dominant temperature measurement technology for gasifiers is the classic thermocouple (TC) or electrical resistance temperature device (RTD) probe. The most common material used in TCs is precious platinum-rhodium that can handle the extreme temperature measurement range. However, TCs are very susceptible to the harsh operating environment inside the slagging gasifier and often fail within hours of gasifier start-up, leaving the operator with no real-time means of temperature measurement. The conventional approach is to create a more corrosion-resistant thermocouple using a ceramic sheath or an improved filler material. An improved filler material has been developed by NETL, along with a dry-pressing method of manufacture that can be readily adapted to a commercial setting, which has proven to have limited effectiveness in the aggressive gasifier environment [27]. However, the heavy sheathing makes such devices less sensitive to dynamic changes in temperatures. In addition, protective sheathings degrade with time under the attack of the ash and slag, causing erosion damage, which would lead to delayed read-out and faulty readings. Metal-based-improved TCs have a typical life less than 120 days. Failure rates can be up to 50% within 15 days, and 75% within 30 days. This is especially true for entrained flow slagging 7 gasifiers since even the most hardened sensors are unlikely to survive for more than 1 or 2 months as the inner surface of the refractory wall degrades and recesses, exposing sensors directly to the corrosive slagging environment. Alkali vapors and transition metals attachments, and solid coal slag build-ups affect thermocouple measurement and accelerate surface corrosion. Above the material limits, TCs can disintegrate completely [7]. The large thermal ramps and mechanical stresses the TC probe suffers cause multimaterial probe mechanical failure due to coefficient of thermal expansion (CTE) mismatch of probe materials [10]. Although local nitrogen or other purge gas may provide protection for TCs, this would result in understated temperatures. In reality, the gasifier operator would sacrifice the temperature measurement absolute value for a continuous and steady temperature reading. The approach is not to position the TCs' tips flush with the refractory wall, but slightly withdraw the TCs into the wall. In this manner, we can protect the TCs from slag or other erosion damage [43]. The actual temperature measured is closer to that of the refractory than that of the reactor core, and thus is highly dependent on the extent of the depth of withdrawal from the reactor space. 1.2.2 Indirect Measurements Several reports describe how secondary measurements that are relatively easy to obtain-such as temperatures, pressures, and compositions of streams into and out of a gasifier-can be used in conjunction with empirical or theoretical models and correlations to estimate inaccessible operating parameters inside the reaction zone. The advantage of indirect measurement is that sensors are usually located at the downstream of the gasifier, which would be away from the extreme environment of the gasifier. For example, Higman and van der Burgt [43] conclude that the temperature of a dry slurry feed gasifier can be monitored by measuring the concentration of CH4 or CO2 in the product gas. In fact, it was reported that this approach was used to estimate gasification temperature during the Tampa Electric Integrated Gasification Combined-Cycle Demonstration Project [6], [44] and is believed to be in common use by at least some operators of gasification 8 units in the United States. An attempt to correlate a large number of routinely measured process variables to the composition of the produced syngas was also reported [39]. Texaco Inc. [17] has developed an apparatus that monitors the weight of slag as it accumulates and subsequently as it flows form the gasification chamber during a deslagging operation. The resulting data were fed into a preprogrammed computer to evaluate the gasifier operation conditions, including temperature. The evaluation results will cause the fuel composition and its rate of feed to be automatically adjusted by the computer program. Computational study by Sarigul [90] showed close correlation between CH4 concentration and the adiabatic flame temperature of the gasifier. The same approach is used at Eastman; in fact, the gasifier operators routinely report temperature inside the reactor in ppm of methane. The advantage of this method is that it gives an integral measurement of the temperature at the reactor outlet. However, it does not give an indication about local hot spots. Moreover, the measurement has a certain time delay for real-time monitors. A heat flux measurement is comprised of an installation of a small piece of membrane wall in the wall of the refractory and measurement of the increase in water temperature of a known amount of water flowing through the membrane wall. Its response time is relatively slow, usually 10-30 seconds, indicating the local average temperature [65], [85], [97]. Despite limited current use, the inferential sensors remain a promising approach in gasification applications, with further advantages of relatively small invest­ments and retrofit requirements for their deployment. However, two fundamental limitations of inferential measurements must be taken into account. First, the quality of inferences critically depends on modeling errors and uncertainties, and unmodeled changes to the process itself (e.g., due to ware and aging), its feed, and unknown process disturbances. Second, the measurement accuracy, sensitivity, and response time of inferential measurement compare poorly with the corresponding characteristics of the direct measurements. Therefore, the direct measurements in gasification will continue to be desirable despite a long development time, high development cost, and technical challenges that must be 9 overcome. 1.2.3 Noninvasive Measurements An alternative technique to obtain direct measurements of gasification temper­ature is using methods that do not require the direct or partial insertion of a fragile sensing element into the harsh environment. The most widely used techniques in this category are optical measurements, used for combustion-specific measure­ments of temperatures and reaction composition [87]. Typical fiber optic sensors include those based upon optical reflection, scattering, interference, absorption, fluorescence, and thermally generated radiation [26]. Optical pyrometry [72] is a practical method for measuring temperatures of flames if the blackbody radiation emissitivity factor is constant or calibratable. Texaco Inc. [62] used an optical pyrometer in its pilot unit for several years. However, conventional total radiation or single-wavelength pyrometers cannot provide accurate measurement of the flame temperature because of the unknown or nonuniform emissivity of the flame. In addition, the background radiation can interfere the measurement. Two-color pyrometry removes the emissivity limitation by using the ratio of irradiances at two carefully selected wavelengths, which has been used from flame in utility furnaces to various open flames, such as premixed and diffusion flames [66]. The multicolour method has also been developed over recent years. Most existing optical pyrometries can only provide a measurement of average temperature of a single-point or a small area defined by the field of view of the probe. The 3D temperature profile can be reconstructed based on a collected 2D flame temperature image transformed from the color flame images [113]. Though minimally invasive (requiring a transparent access port), optical tech­niques are not suitable for temperature and composition measurements when an optically transparent line-of-sight is difficult or impossible to maintain, as in the case of slagging gasification or when high particle concentration in the reaction zone prevents light transmission. The ash and slag would block optical access ports. Continuously blow nitrogen gas is required to keep the pores open, which make it expensive and results in a loss of reading very frequently. Therefore, any 10 commercialization of optical pyrometry is likely to be in addition to rather than as a replacement of thermocouples. Various optical-based methods have also been studied for extreme temperature measurements [60], including an optical fiber bragg grating (FBG) sensor, optical resonator cavity sensor, optical sapphire fiber-based fluorescence sensor, optical single crystal sapphire-based sensor, etc. Photosensitive FBG sensors and optical resonator cavity sensors, such as Fabry-Perot cavity-based sensor, are both silica fiber based, which cannot survive temperatures over 1000 °C because of material limitations. In long-term thermal tests of FBGs at temperatures close to or above 1000 °C in air for hundreds of hours, unpackaged standard silica single mode fibers lost almost all of their mechanical strength. It is certain that silica fibers experience severe mechanical degradation in the oxidizing atmosphere at high temperature. The silica fibers become extremely brittle and any subsequent handling of the fiber is not possible after the test [38]. Single crystal sapphire fiber or bulk is a more successful sensing element used for extreme temperature sensing applications that have a melting temperature ~ 2050°C [37]. Most sapphire fiber sensors are based on Fabry-Perot structures within the fiber generating broad-band interference fringe pattern that can be monitored as a function of temperature [102]. Sapphire-based sensors have been fabricated and demonstrated for high-temperature measurements in laboratory tests [93], [112]. These sensors will still have to survive the difficult reactor environment, and the fundamental uncertainties of temperature measurement in a gasifier will remain [78]. A well-controlled dopant density is important for accurate measurement performance, which puts strict requirements on its fabri­cation process and increases the cost. It has been difficult to achieve high-quality measurement results, since the interference signal from the fabry-perot cavity are degraded by the multimode electromagnetic fields in multimode sapphire fibers. These systems during industrial service are prone to mechanical failure due to CTE mismatch between optical fibers and sensing elements. 11 1.3 Organization The research of this dissertation illustrates that the development of a novel noninvasive technique using ultrasound measurements of segmental temperature distribution (US-MSTD) method can potentially offer excellent resolution for solv­ing challenging direct temperature distribution monitoring problems on coal and biomass gasifiers. Chapter 2 will discuss the fundamental physical basis of temperature mea­surement based on ultrasound methods, and the essential concept of ultrasound measurements of a segmental temperature distribution system. Chapter 3 will demonstrate the development of an ultrasound system for low temperature mea­surements with cementitious waveguide. In Chapter 4 of this dissertation, the ultrasound system using engineered alumina waveguide for laboratory-scale high temperature tests will be discussed. Chapter 5 will describe a robust signal processing method based on the combination of cross-correlation of envelope of waveform and anisotropic diffusion methods. The validation of the developed US-MSTD system in a pilot-scale downflow oxyfuel combustor will be presented in Chapter 6. Chapter 7 provides improvement of the hybrid US system on waveguide material selections. Finally, the conclusion of this research and possibly future research on this topic will be given in Chapter 8. Chapter 2 and 3 correspond to paper [51], Chapter 5 corresponds to paper [52]. CHAPTER 2 NONINVASIVE ULTRASOUND MEASUREMENTS OF SEGMENTAL TEMPERATURE DISTRIBUTION1 2.1 Ultrasound Ultrasound (US), because of its high sensitivity, high penetrating power, fast time response, great accuracy, and noninvasive operation, has become a potential approach for noninvasive temperature measurement. It became a subject of in­terest to researchers during World War I, but its use in industry did not grow to proportions worthy of note until World War II. Traditional ultrasonic applications have been used almost exclusively for nondestructive testing (NDT), such as macroscopic flaw detection/evaluation and dimensional measurements; material characterization, such as microstructures and associated mechanical properties assessment [89]. Development and perfection of ultrasonic nondestructive evalua­tion techniques are capable of monitoring and controlling the material's production process; the material's stability during transport, storage, and fabrication; and the rate of degradation during the material's in-service life [36]. The application of ultrasound in medicine began in the 1950s, and includes diagnosis, commonly called sonography, and therapy [18], [63]. Sonography is used for evaluating the condition of internal organs and tissues, commonly for neonatal fetuses, heart imaging, and blood flow measurement. Therapy is provided by high intensity waves which heat tissues to provide massage treatment or break stones [30], [73]. 1This chapter is adapted with permission from (Jia, Yunlu, et al. "Ultrasound Measurements of Temperature Profile Across Gasifier Refractories: Method and Initial Validation." Energy & Fuels 27.8 (2013): 4270-4277.). Copyright (2013) American Chemical Society 13 Ultrasound works with vibratory waves at frequencies above those within the hearing range of the average person at frequencies above 20 kHz. Ultrasonic waves are stress waves. Therefore, they can exist only within substance, such as gases, liquids, and solids [29]. The behavior of ultrasound propagation is given as [98] y(x,t) = yocos^cvt- " A") ' (2.1) where y is the particle displacement of the propagating sound wave with respect to distance, x, and time, t, y0 is the amplitude of the wave, w is the angular wave frequency, and A is the wavelength. w and A are constants defined by the medium in which the sound is traveling, which are defined as c = f • A, (2.2) w = 2nf, (2.3) where c is the ultrasound propagating speed (the speed of sound, SOS) within an elastic medium and f is the frequency. 2.1.1 Basic Instrumentation Figure 2.1 shows a block diagram of a basic ultrasonic measurement system used to generate and detect ultrasonic waves in a solid specimen. The synchronization generator gives trigger signals with high repetition rate to the pulser. The pulser provides electrical voltage to the transducer so that the transducer excites ultrasonic Figure 2.1: Block diagram of basic components of an ultrasonic measurement system to generate and detect ultrasonic waves. 14 waves at the same repetition rate. The reflected ultrasonic waves through the specimen are received by the same transducer (called pulser-echo mode) and the resulting voltage of the received signal goes to the display through oscilloscope. The computer is used to collect and analyze the acquired ultrasonic waveforms. 2.1.2 Ultrasound Propagation Speed Considering a sample of a solid material with known thickness L maintained at a uniform temperature, assuming a pulse-echo method (transducer/receiver is the same device), by measuring the time it takes an ultrasonic signal to travel the distance L, the so-called the time of flight (TOF), the speed of sound may be calculated as 2L c = - . (2.4) tof The speed of sound and its propagation mode are directly dependent on the physical composition of the transmitting medium. In solids, the SOS depends upon the type of pulse wave, the elastic properties, the density of the medium, and the frequency in some cases [23]. The ratio of applied stress (force/area) to axial strain (extension/length) is called the elastic modulus, or Young's modulus. A piezoelectric transducer can generate both longitudinal/compression, also known as primary wave/p-wave, and shear waves (s-wave), but one type of mode will be dominant depending on the particular transducer's piezoelectric properties. The relationship between longitudinal SOS and the elastic modulus in an isotropic solid, in which the particle motion is parallel to the axis of the ultrasound propagation, is given by c _ lK+ 3G _ E(1 - v) P5 ) Clongitudinal _ \ l p _ y j p(1 + V)(1 - 2V) . (2.5) When the particle motion in the wave is normal to the direction of propagation, the relationship results in the following equation for shear SOS G cshear _ y ~p' (2.6) where K and G are the bulk modulus and shear modulus of the elastic materials, respectively. E is the Young's modulus, p is the density, and v is the Poisson's 15 ratio, while Young's modulus and Poisson's ratio have a relationship of E = 3K(1 - 2v). The SOS of longitudinal waves depends both on the compression and shear resistance properties of the material, while the speed of shear waves depends on the shear properties only. For a long, thin rod, where its diameter is much smaller than its length, the SOS of longitudinal waves may be simplified and given by: Liquids and gases cannot resist shear stresses. The SOS in fluids is expressed as where /3 = 1/K is the compressibility of liquid. For an "ideal gas," the SOS can be shown to be where y is the adiabatic index, which is the ratio of specific heats of a gas at a constant-pressure to a gas at a constant-volume(Cp/Cv), P is the ambient pressure, and R is the molar gas constant. Young's modulus of elasticity is sensitive to most of the common microstructural evolutions and damage (microcracking, densification, phase transition, etc.). The quantitative assessment of microstructural changes can be carried out through the measurement of ultrasound properties. 2.1.3 Impedance and Attenuation Acoustic impedance, another major characteristic to describe ultrasonic prop­erties of material, is the quantification of the resistance of ultrasound propagation in a medium [22], [55] An ultrasound propagating through a material has its acoustic energy loss. The reduction in the amplitude of a ultrasonic waveform is attenuation [55], [63], (2.7) (2.8) (2.9) Z = p • c. (2.10) 16 expressed as the logarithm of the ratio of the magnitudes of the original to the attenuated pressure amplitudes, a and a0, measured in decibels (dB): A(dB) = 20log10( - ) . (2.11) a0 Two types of process affect the attenuation of ultrasound propagating wave­form, which involve material responses and wave interaction. Material repose processes include geometric attenuation, such as beam spreading or focusing, energy absorption, dispersion, and nonlinearity. Transmission across interfaces, scattering by material variation, inhomogeneities of grains, grain boundaries, pores, and more, and the Doppler effect are three main aspects of wave interaction processes that cause acoustic attenuation and defects [14], [74]. Scattering is the primary mechanism by which ultrasonic energy is lost during propagation, which also affects the feasibility and effectiveness of engineered waveguides used for proposed temperature measurement. Scattering is the redi­rection of an ultrasonic wave, as a result of the interaction between a primary ultrasonic propagating wave and the anisotropic grains (inhomogeneities) inside of the medium [58]. If their physical properties such as density or elasticity are different from those of the surrounding medium, it causes a discontinuity in ultrasound propagating speed at each grain boundary, which leads to the reflection at the grain boundaries denoted by scattered wave and energy loss. The magnitude of scattering depends on the particles per volume, size, acoustic impedance, and frequency [41]. A scattering is created at a single grain if the dimensions of the heterogeneities are smaller than the wavelength, as D <= A, where A = c/f , f is ultrasound propagating central frequency. This scattering problem of sound was first solved by Lord Rayleigh and is therefore called Rayleigh scattering [84]. There are a number of causes of ultrasonic energy loss in solids, and Rayleigh scattering is one of the main reasons [13]. The process of energy loss usually refers to the changes of ultrasonic energy into heat. The presence of microscopic structural defects, such as point defects and dislocations and macroscopic defects, affects the degree of hardness and the elastic properties of the material and gives rise to absorption that occurs in both metals and nonmetals. 17 2.1.4 Reflection and Transmission When an ultrasonic wave encounters an interface, several phenomena may occur, including reflection, transmission, refraction, and mode conversion. These interactions are the phenomena upon which our proposed ultrasound method relies. Ultrasonic waves are reflected at the interface of two media if there is a difference in acoustic impedances (Z) of the materials on each side of the interface. This difference in Z is commonly referred to as the impedance mismatch [29]. Assuming the incident is normal to the interface, the fraction of the incident wave intensity that is reflected can be derived because particle velocity and local particle pressures must be continuous across the boundary, which is calculated from the acoustic impedances of the materials on both sides of the interface: Re = ( Z2-Z1 f , (2.12) \Z2 + ZlJ where z1 and z2 are acoustic impedances of media 1 and 2. The value produced by Equation 2.12 is known as the reflection coefficient, while the transmission coefficient is calculated by Tr = 4+1Z2 , (2.13) (Z2 + Z1)2 or from Re + Tr = 1. The reflection and transmission coefficients represent the percentage of acoustic energy which is either reflected or transmitted at a boundary. The greater the impedance mismatch, the greater the percentage of energy that will be reflected at the interface. In addition, the preceding equations for reflection and transmission apply to ideal interfaces that have no thickness. 2.2 Physical Basis of the US-MSTD Method The physical basis of the proposed noninvasive ultrasound approach for tem­perature measurement is temperature dependence of the speed of sound in gas, liquid, and solid: c = f(T). (2.14) 18 If relationship 2.14 is known, either from theoretical considerations or empirical correlations, the measured tof can be used to obtain the uniform temperature of material of interest as: T = f-1 ( , (215) assuming that the inversion of 2.14 is unique. In addition to the TOF measurements emphasized in this research, other ultrasonic characteristics, such as a phase change of echoes produced by a tone burst excitation, may be used to define temperature-dependent variations in the SOS [67]. The application of this idea to measure the temperature in gases is known as acoustic pyrometry and is well established [34], [53] and commercially used in many high-temperature applications, such as the cement industry, combustion, and incineration industries [15], [57], [106]. The advantage of the approach is the ability to obtain real-time temperature measurements over an extremely large range of temperatures (from 0 to 3500 °F), which makes it applicable to process monitoring from a cold start up to normal high-temperature operation. Disadvantages include significant measurement uncertainties when temperature along the propagation path between the transducer and the receiver varies significantly and unknown changes in the adiabatic constant due to variability in the gas composition. The utilized acoustic frequency range is low (typically, <3kHz) because higher fre­quency ultrasound does not propagate well through gases. The consequence of low excitation frequencies is interference from combustion instabilities, sounds produced by a turbulent flow, and other disturbances, collectively known as a passive acoustic signature. Such low frequencies also limit the achievable spatial resolution of measurements when multiple transducers-receivers are used in order to measure the temperature distribution inside of a containment [48]. In addition, acoustic thermometry has been used to detect temperature changes in the ocean by receiving low-frequency ultrasound (< 100 kHz) transmitted across an ocean basin [31]. Ultrasonic method offers a powerful noninvasive or minimally invasive alter­native for temperature measurement in solids. However, very little work has been 19 done in this area. Techniques like this are particularly useful for applications such as when: a) insertion of temperature probes is undesirable, difficult, or impossible; b) extreme environments affect longevity of conventional sensors, as is the case for many energy conversion processes; and c) optical line-of-sight measurements are not practical because the medium is opaque or optically dissipative. The dependence of the speed of sound on the temperature, c = f (T), obtained experimentally or theoretically, would then allow us to estimate the temperature of the sample. Several notable examples using acoustic thermometry to measure temperature have been shown in microelectronic and medical applications. Most conventional ultrasonic remote temperature sensing methods rely on an assump­tion of constant temperature along the propagation path in the solid. Lee et al. [61] reported the development of an acoustical temperature measurement system which uses the TOF measurements of an acoustic wave introduced into the wafer through an excitation quartz rod. The wave, partially reflected from the quartz-silicon interface, travels through the wafer until reaching a second quartz rod through which the wave reaches the receiver. The difference between an arrival time of the reflected wave and the wave reaching the receiver through the second rod gives the time of flight through the wafer, which is used to estimate the wafer temperature. Lee et al. reported that ± 5 °C accuracy was achieved in the range from room temperature to 1000 °C (with a proposed use up to approximately 1800°C [2]). Arthur et al. [1] investigated the use of backscattered ultrasound energy in temperature measurements to monitor and control noninvasive thermal therapies of tumors. Using a 7 MHz linear ultrasound phased array transducer, they demonstrated temperature measurements in ex vivo phantom tissue from 37 to 50 °C in 0.5 °C steps. The project did not progress towards in vivo testing because the quality of temperature measurements was severely affected by subject motion, unavoidable in subjects due to breathing and other disturbance. Simon et al. [94] developed a 2D temperature estimation method based on the detection of shifts in echo location of the backscattered ultrasound from a tissue undergoing thermal 20 therapy. They suggested that 0.5 °C accuracy is possible. Another example of high-temperature application of the US method is in metallurgy. Balasubramaniam et al. [3] used an ultrasonic sensor to measure viscosity and temperature of molten material up to 1000 °C with the temperature resolution of 5 °C. However, when temperature distribution of the sample T(z) along the propaga­tion path is nonuniform, the overall TOF depends on the temperature in a complex and unknown way and there are many arbitrary temperature distributions T(z) that will result in the estimated TOF matching the measured value. bution may be resolved by adding constraints on the feasible solution so that an estimation of fixed temperature distribution based on TOF measurements in the regularized by imposing additional constraints on the temperature distribution. This has an effect of parameterizing the "admissible" temperature distribution by prescribing a functional form that depends on one or more unknown parameters, which are then found from ultrasound and, perhaps, other unrelated measure­ments. Parameterizations may include an assumption that the temperature along the US propagation is constant and given by Equation 2.15; the temperature distribution is linear with the slope and intercept found from tof and at least one additional independent measurement; and the requirement that T(z) satisfies a heat transfer model with appropriately selected parameters (such as thermal conductivity) and boundary conditions. In the case of the linear parametrization, a slope and intercept are needed to determine the parameterization. However, both parameters cannot be determined from a single measurement of tof . At least one additional measurement, such as the thermocouple measurement of the temperature at the location of the transducer, is required to determine both unknowns. If assumption of a constant temperature Ta is used across the ultrasound propagation path, 2.15 calculated from using the measurement of tof may only (2.16) The lack of unique dependence of the measured tof on the temperature distri integral form (2.16) becomes possible. The first approach is to have the problem 21 the suitable for small temperature gradient cases. The relationship between the calculated average Ta and the unknown temperature distribution T(z) is given by the following equation: When strong thermal gradients are present, using Equation 2.17 to approximate the temperature distribution across the containment of extreme environments, such as refractories of the gasifier and other energy conversion processes, would result in less accurate temperature estimation. The TOF measurements alone are not sufficient to reconstruct the parameterized temperature distribution with more than a single unknown. Reducing the number of unknowns for temperature distribution determination is to devise an approach that provides more data than a single measurement of the TOF. This can be achieved by using multiple transducers and receivers to measure transmit and echo delay times along different ultrasound propagation paths, followed by the simultaneous interpretation of the measurements to reconstruct the temperature distribution (e.g., [35], [48], [80]). This approach shares common features with X-ray computer tomography, which reconstructs the density (attenuation) distribution inside the sample and other noninvasive measurement modalities in which the acquired data depend on the spatial integral of the property of interest [111]. Model-based temperature estimation using TOF measurements is another ap­proach and is found the most in literature. Takahashi and Ihara [47], [96] tested a 30 mm steel plate, with a single side heated at 300 °C and 700 °C at steady state. The temperature at the transducer location was independently measured. A linear relationship between SOS and the temperature 2.14 was assumed. In addition, they assumed a 1D heat transfer model for temperature estimation along the length of ultrasound propagation direction. Thus, the hot face temperature was assumed to predict the temperature distribution T(z) along the propagation path and the corresponding model prediction of the TOF from the model. The unknown distal temperature was then estimated as the value that minimizes the difference between the TOF predication given by Equation 2.16 and its measured value. An approach developed by Schmidt et al. [91] also uses a 1D heat transport model to obtain (2.17) 22 T(z), but it adjusts a different single parameter, the boundary heat flux at the distal end of the ultrasound propagation path, to match the measured and the predicted values of tof . Heyman et al. [42] invented their dynamic acoustic thermometer to measure the temperature at a remote location by relating the measured change in inte­grated acoustic delay to the remote location temperature with a combined 1D thermal-acoustic model. The integrated acoustic delay is determined from the measurements of phase change between points of interest and reference location, at which a constant frequency is applied. The combined model relates temperature to acoustic propagation speed of sound along the path. The experiment of measuring steel rod temperature using this system in a nonstirring water tank showed a temperature resolution of better than 6 at room temperature and a 110 resolution for a 60 rpm magnetic stirring test. The experiments also showed a faster response to the thermal energy change, long before the thermal wave had propagated from the heat source to the sensor location. Yuhas et al. [108] designed an apparatus for determining local temperatures of inaccessible surface heat fluxes based on measured propagation time using ultrasound pulse-echo mode. They also used 1D thermal model approach and assumed that the dependence of SOS on temperature is linear. The temperature at the point of interest may be not calculated to estimate its heat flux as a step-wise constant function of time. The relationship between SOS and temperature was calibrated with known constant heat flux. The verification experiment was carried with a naval ship gun barrel Mark 45 Navel Gun with thickness of 0.0635 m during a firing regimen using three assumed heat flux profiles. Their tests showed that the estimated maximum heat flux is underestimated by no more than 6%, but the maximum measured temperature has an over 150 °C error. The third distinct approach is to devise ways to extract more information from the response of each ultrasound transducer-receiver other than a single time of flight measurement. This is possible if the ultrasound pulse produces multiple ultrasound reflections, caused by echogenic features encountered as the excitation propagates through a waveguide, which has been presented in optical fiber Bragg 23 grating to measure temperature [83]. Hanscombe and Richardson in Schlum-berger [40] proposed a method in which an ultrasound waveguide is engineered to have a number of randomly spaced notches formed by arbitrary grating length. The dominant frequency reflected at each notch is different, which is determined by grating dimensions, which would change due to temperature changes through thermal expansion. Multiple frequency-separated echoes propagate along the waveguide; the cross-talk between overlapping echoes is reduced and causes the SOS of the waves within the notches to change because the temperature variations are encoded by the changes in echoes' frequency content. The described innovative approach apparently has not been tested in experiments, leaving many unanswered questions. For example, it is not clear how long each grated zone should be to ensure narrow frequency content of each echo. The achievable accuracy as a function of grating design remains unknown. Furthermore, the accuracy of temperature measurements when thermal gradients are present within each grated zone, leading to a wider frequency band of each echo, has not been quantified. 2.3 Method Our proposed ultrasound measurements of the segmental temperature distribu­tion method are shown in Figure 2.2. The overall system for measuring temperature distribution across refractory and other aggressive process containments consists of: (a) the engineered ultrasound propagation path with echogenic features cre­ating partial reflections from known locations, either embedded as an insert or incorporated into the refractory to provide partial ultrasound reflections; (b) an ultrasound transducer and receiver, which can be implemented as single or distinct components ; (c) the analog and digital ultrasound instrumentation used to generate the excitation pulse and then acquire and amplify the return echoes; (d) the signal processing system that accurately determines the TOF for each echo and then uses this information to calculate the SOS or its change in the corresponding segment of the refractory; (e) the relationship between the SOS and the temperature; and 24 Ultrasound Pulse & Echoes H- - --------H -*- TOF1 «- » TOF2 «---------- • TOF3 «--------------- » Figure 2.2: Ultrasound measurements of temperature distribution in the refractory. (A) Refractory material contains embedded planes of scattering material. (B) Layered refractory. (C) Refractory insert with geometric changes in the ultrasound propagation path creates partial back scattering. Left panel shows an ultrasound ex­citation pulse and the train of partial echoes produced by internal partial ultrasound reflectors. Right panel illustrates an engineered ultrasound waveguide/insert - with internal back scatterers, layers structure, or geometrical changes - embedded into the gasifier refractory. (f) the method to translate the segmental SOS into the temperature distribution such that the predicted TOF, according to Equation 2.16, matches the measurement values. Each of the components will be discussed in more detail in the following chapters. In the described approach, the sensitive electronic components are kept away from harsh gasification environments and it is only required that the US transducer be acoustically coupled to the cold side of the refractory, representing minimal modifications to the gasifier. 2.3.1 Structure US Propagation Path The central idea of the US-MSTD method is to create an ultrasound propaga­tion path inside the refractory (or material of interest) which incorporates partial ultrasound reflectors (back scatters) at known locations that redirect a portion of US energy of the excitation pulse back to the transducer as multiple echoes. Figure 2.2 illustrates three different alternatives to creating such ultrasound backscattering. In this illustration, it is assumed that the same element serves as a transducer and receiver; modification for the case of a separate transducer and ress Reaction i Zone - / US Transducer/ Receiver Gasifief 25 receiver and an angled US beam are straightforward. In Figure 2.2(A), partial US reflections are created by planes of scattering material embedded into the refractory. The second option is depicted in Figure 2.2(B), where the refractory material is layered, with slightly different acoustical impedance in each layer. Figure 2.2(C) shows an embodiment in which partial reflections are created by geometric changes in the US propagation path through an embedded refractory insert. Such an insert can have a geometry (e.g., as shown in Figure 2.2(C)), designed to produce distinct US reflections at predetermined spatial positions, or layered properties, as in the case of Figure 2.2(A) and 2.2(B). Separately produced inserts can be introduced into the refractory during its replacement, service, or relining, as illustrated in the right panel of Figure 2.2. A measurement of the temperature distribution begins with a US pulse, gener­ated by an ultrasound transducer. This pulse will be partially reflected from each scatterer in the insert and return to the receiver as a train of partial echoes at time TOF1, TOF2, TOF3, ..., as shown conceptually in Figure 2.2 (left panel). The temperature distribution in the i-th segment of the propagation path is inferred from the difference in the time of flight, tofi, between consecutive echoes produced by echogenic features which bound the segment at locations Z/_1 and Zf. where (z/ _ Z/_1) is the segment's length. The TOF of the first echo gives an indication on the temperatures in the 1st zone of the refractory, between the cold surface and the first scatterer based on information specific to the temperature distribution with that segment. The next return echo will originate from the second scatterer. By subtracting the TOF of the second and the first echoes, the temperatures between Scatterers 1 and 2 can be estimated, and so on until the estimate of the temperature 2.3.2 Acquisition of Echo Waveforms The time of flight f f of the echo produced by a feature located at z/ is equal to (2.18) (2.19) 26 distribution throughout the refractory is obtained. With that distribution known, the last echo, reflected from the refractory-reactor zone surface, can be used to determine the temperature of the refractory's interior hot surface. The first segment between the transducer and the first scatterer is often used as a delay line and references the time of flight of all subsequent echoes to the arrival time of the first echo. Then the difference in the TOF between the second and the first echoes gives the information on the temperature distribution in the second segment, and so on. 2.3.3 Signal Processing Since the speed of sound is calculated as the distance traveled by an ultrasound pulse divided by the time of propagation (or time of flight, TOF), a method for precise measurements of the time of flight is essential to accurate measurements of temperature distribution. The simplest approach to the measurements of the TOF and its changes is to use temporal location of a single-point waveform feature, such as the first zero crossing or the peak value of the waveform. Though standard, these timing techniques are sensitive to measurement noises. Furthermore, when broad-band excitations are used, the timing accuracy of single-feature methods de­teriorates further due to waveform distortions and broadening caused by stronger attenuation of higher-frequency content of ultrasound pulses. More robust and accurate measurements of tof and tof may be achieved when the entire shape of the waveform is utilized in timing. In this case, both amplitude and phase information are taken into account [45], which makes timing results less sensitive to measurement noises and shape distortions. Mathematically, the cross-correlation between two signal f (t) and g(t) is represented as: f *(t) g(t + T)dt/ (2.20) TO where f * is the complex conjugate of f and t is the lag time between two signals. The temporal shift t needed to obtain the best match between the waveforms may be found by maximizing their cross-correlation <pfg [24], minimizing I1 and I2 norms of their difference [50], [75], [101], or by maximum likelihood [20]. Figure 2.3 shows that A TOF between two echoes is obtained by finding the best match of the entire normalized shape of the two waveforms as based on both phase and 27 Figure 2.3: The ATOF between echo waveforms at different temperatures is calculated by cross-correlation with a reference waveform acquired at 20 °C. amplitude information, which makes timing results less sensitive to noises and shape distortions. Though cross-correlation and other shape-matching methods perform better than single-point timing, the results may still be unacceptable when significant distortion of ultrasound waveforms occurs, as is often the case when the pulse propagates through attenuating and dissipative materials. It was suggested by Le [59] that for waveform distorting samples, a higher precision can be achieved if the envelopes of the waveforms are used in timing. The analytic signal, sa(t), of the waveform, s(t), is the following complex function: Sa (t) = s(t) + js(t), (2.21) where j 2 = -1 and S(t) is the Hilbert transform of s(t): lim f 0 J £ 1 ,. f " s(t + t ) - s (t- t) , s(t) = - --------------------- dT. (2.22) n £- The envelope of the waveform s(t) is then calculated at the amplitude of its analytic signal: A(t) = |s»(f)| = ^ s 2(t) + s2(t). (2.23) 28 We have found [52] that further improvements in timing accuracy can be achieved by iteratively applying a nonlinear anisotropic diffusion filter to the envelopes of the echo waveforms, which will be discussed in a later chapter. 2.3.4 Temperature Dependence of the Ultrasound TOF The unknown temperature distribution is estimated from the measurements of the time of fight of ultrasound echoes. The SOS and length of the propagation path both change with the temperature of the waveguide and thus affect the echoes' TOF. It is possible to separate their contributions to the changes in the TOF. However, as long as the calibration curve 2.14 is obtained without differentiating between the two phenomena, there is no practical need to distinguish the contribution of each one. The subsequent discussion assumes that the correlation between the SOS vs. temperature, Equation 2.14, was not corrected for the thermal expansion. This simplifies the method, as it becomes unnecessary to adjust the length of the propagation path in Equations 2.16 and 2.18 for the thermal expansion. 2.3.5 The Temperature Distribution Estimation The measurements of the segmental time of flight tofi encode the information on the temperature distribution within i-th segment. As before, additional as­sumptions are needed to estimate the segmental temperature distribution from the measurement model 2.18. All parametrization options discussed in the context of deconvoluting model Equation 2.16 maybe used for this purpose, and are discussed below. 2.3.5.1 Piecewise Constant Distribution This distribution is obtained by assuming constant speed of sound within each segment. Using this assumption in Equation 2.18, the constant SOS in the i-th segment of the waveguide is obtained as C = 2(Zi_ Zi-1), (2.24) tofi and the corresponding constant temperature is obtained by inverting the correlation 2.14. After repeating the process for all segments, the entire temperature distribu­tion along the waveguide is approximated as a piecewise constant function. The 29 infeasible temperature discontinuity at the locations of the echogenic features is an undesirable feature of such approximation. By using a larger number of echogenic features and the correspondently finer segmentation, this approximation can be further improved. 2.3.5.2 Piecewise Linear Distribution Assuming that the temperature changes linearly within each segment signif­icantly enforce temperature continuity. For a segmental sample with several echogenic features, the temperature at the transducer location, z = 0, is measured independently and equal to T(0) = n1. By using the measured TOF of the first echo and the linear temperature distribution in Equation 2.19, the following equation is obtained r z2 1 t0f = 2 ------------ - dz, (2.25) 0f1 Jz\ f (m z + «1) from which the unknown slope of the distribution, m1, can be found. Similarly for the i-th segment, the unknown slope m/ and intercept n are obtained from the solution of the following two equations: r zi 1 'of'=2JL m + n * dz (2.26) and n/ = (m'-1 - mi)zi-1 + n -1 , (2.27) where t0f is the difference in the TOF of the i and i - 1 echoes, and Equation 2.27 enforces the continuity of the temperature at z = z ^ . The process continues for all remaining segments until the piecewise linear approximation of the temperature distribution over the entire sample is obtained. 2.3.6 Estimation of Heat Flux The measurement of conductive heat fluxes through a solid is commonly obtained by attaching a flux sensor to the surface to the sample. It is therefore capable of estimating only a localized heat flux in the immediate proximity of the flux sensor. The approach proposed in this dissertation can be used to profile the temperature distribution over a significant distance away form the surface. 30 Thus, it can be used to noninvasively profile conductive heat fluxes through the sample, at a considerable distance from its surface. Specifically, by differentiating the estimated temperature distribution T(z), the conductive heat flux, q, across the sample is estimated as The piecewise linear temperature distribution will result in a piecewise constant ultrasound transducer. An even more detailed estimation is possible when the heat flux is calculated based on the temperature profile that satisfies the heat conduction model. The temperature parameterization by a one-dimensional heat conduction mod­els was usually adapted. When temperature of the distal end Th is to be determined, the temperature at the location of the transducer Tc is independently measured, and the temperature distribution T(z) can be estimated by adjusting a single boundary produced by a reflection of the excitation pulse from the distal end of the ultrasound propagation path: where p, C, and k are refractory density, heat capacity, and thermal conductivity, respectively. When a 2D or 3D model is needed to provide an adequately accurate description of the temperature distribution in the sample, additional measurements will be required to reconstruct the temperature distribution. For example, consider the case of a cylindrical waveguide with the transducer, used to launch an excitation pulse in the axial direction, coupled to one of its ends. Assuming the radial symmetry of the temperature distribution, constant density p, heat conductivity k, (2.28) For the case of piecewise constant temperature profile, dT _ Ti _ Ti_1 dz zi _ z/_1 (2.29) estimation of the heat flux distribution in the direction normal to the plane of the 2.3.7 Parametrization with Thermal Conductivity Model condition in order to match the predicted and the measured TOF of an echo (2.30) 31 and capacity Cp, the temperature distribution inside the sample must satisfy the following 2D heat transport model in the cylindrical coordinates: define the problem, three boundary conditions-at the proximal, distal, and cylin-only a single US echo will be produced by a reflection from a distal end of the sample, and measurement of its TOF will allow us to estimate only one of the three needed boundary conditions. The other two boundary conditions must be obtained from independent measurements. For example, if the temperatures of the distal and proximal ends of the waveguide are independently measured, then the measured Atof can be used to estimate the overall heat transfer coefficient h and define the remaining boundary condition given as the heat flux through the cylindrical boundary of the waveguide: where Te is the ambient temperature of the environment. The time of flight of multiple echoes received when the excitation pulse propagates through a structured waveguide provide sufficient data to estimate all required boundary conditions without the need for the additional independent measurements. When such independent measurements are available, they can still be incorporated into the US-MSTD method and may help improve the accuracy and the robustness of the estimated temperature distribution. where r is the radial position relative to the centerline of the sample. To completely (2.31) drical surfaces of the waveguide-are required. If the waveguide is unstructured, q = h(Te - T), (2.32) CHAPTER 3 LOW TEMPERATURE LABORATORY EXPERIMENTS1 The feasibility of the proposed approach hinges on two questions: 1. Is it possible to create partial internal reflections along the path of the ultra­sound propagation and what are the methods that can be used to create such reflections? 2. Is the speed of ultrasound propagation in the refractory temperature-dependent? 3.1 Cementitious Waveguide Partial Reflector Structures Creating partial internal ultrasound reflections from known spatial locations inside the sample is the key prerequisite for the proposed approach to work. Two solutions, illustrated in Figures 2.2(A) and 2.2(B), were investigated. Figure 3.1(A) compares the ultrasound echo waveforms from two similar 4 cm long cementitious samples, one of which (waveforms in red) contains a few 0.5 mm steel shots placed in the middle of the sample during its casting. The result clearly shows a partial echo from inside of the sample created by embedded scatterers, confirming the viability of the concept depicted in Figure 2.2(A). The range of other material has been investigated in order to find the most appropriate selection for internal scatterers. An ideal choice for partial reflectors would be a material with identical thermal expansion, and chemical and mechanical resistances similar to that of the surrounding refractory material; steel clearly does not satisfy these specifications. 1This chapter is adapted with permission from (Jia, Yunlu, et al. "Ultrasound Measurements of Temperature Profile Across Gasifier Refractories: Method and Initial Validation." Energy & Fuels 27.8 (2013): 4270-4277.). Copyright (2013) American Chemical Society 33 -5 0 5 10 15 20 25 30 35 40 Time (^sec) 0 10 20 30 40 Time (^sec) Figure 3.1: Ultrasound pulse-echo response for the samples with internal inter­faces. (A) The pulse-echo ultrasound response of two samples fabricated from the identical cementitious material. One of the samples (Sample B; shown in inserted photograph) contains embedded ultrasound scatterers at the midpoint of ultrasound propagation path, which produces partial reflection (red line). (B) Ultrasound pulse-echo response for the sample with two internal interfaces obtained by sequentially casting three layers of identical formulation and allowing time for a partial cure to occur prior to pouring the next layer. 34 We, therefore, investigated if the concept depicted in Figure 2.2(B) can be imple­mented by using small variations in the composition of the layered cementitious materials, creating partial internal reflections at the interface between the layers. This indeed was found to be the case. In fact, it was found that by casting multiple layers of the same composition and allowing for a partial curing before casting the next layer, enough variation in acoustic impedance is introduced to create partial US reflections at the interface. Such implementation of the refractory with an embedded partial internal ultrasound reflector is particularly appealing since each layer will have essentially identical thermal, chemical and mechanical properties. Figure 3.1(B) illustrates this approach. It depicts the results obtained with the cementitious sample (shown in the insert) obtained by casting three 1-inch thick layers of identical cement mixture and allowing for a partial cure before the next layer is cast. Note three distinct echoes, produced at the two internal interfaces and the distal end of the sample. To determine the conditions needed to create detectable partial internal reflec­tions from the interfaces corresponding to consecutively cast layers, three groups of cementitious samples were made by layered casting using Portland type I/II cement. Two inches of I.D. PVC tubing was cut in lengths of 2, 3 and 4 inches and used as a mold. Water-cement mixture was poured into the vertically oriented PVC mold in several layers approximately 1 inch thick, altering the duration (cure time) between the previous and the subsequent pours. To help with uniform setting of each layer and removing of air bubbles, the mold was vibrated by high speed vibrator on the outside surface; alternative vibration methods are currently being tried. The curing time for different layers varied from 15 minutes to days. Fresh cementitious mixtures of identical composition were prepared right before the casting of each new layer. The water-cement ratio by weight also changed which was found to have a significant effect on the outcome. All samples were cured in air and at least one week was allowed after the casting of the final layer before ultrasound testing; this long cure eliminated short-term aging effects. The ultrasound tests of cementitious samples were carried out using a Panametrics pulser/receiver (model 5072PR) and a Panametrics immersion transducer with a 35 central frequency of 1 MHz (model V302), coupled to a sample using ultrasound gel. The data were acquired using a Tektronix oscilloscope (model MSO 2024) interfaced to a computer. 3.1.1 High Water-Cement Ratio Sample with Various Curing Times The water-cement ratio used for this group of cementitious samples was 0.5, which is the highest water portion from manufacture's recommendation. With this ratio, the mix can be easily and uniformly poured into PVC molds. With two pours, each 1 inch thick, two layer samples are created that have a single internal interface. To investigate the effect of the cementitious curing time, several samples were created in which we varied the time the first layer was allowed to cure prior to completing the sample with the second pour. The curing time for the first cementitious layer was set to 15, 30,45 minutes or 1 hour for different samples. The results of ultrasound tests show that the clearest partial internal reflections are observed with samples in which the first cementitious layer was allowed to cure for 1 hour (Figure 3.2). The reflections from the internal interface are less clearly defined when the curing time was 45 minutes, indicating a smaller change in acoustic impedance between the layers that were cast with little delay (Figure 3.3(A)). Further reduction in time allowed for the first layer to cure makes the two layers even less distinguish­able to ultrasound testing, indicating a very small change in ultrasound impedance between the layers under these circumstances. For the samples cured for only 15 minutes, we did notice the change in the ultrasound signature depending on the position of the ultrasound transducer relative to the center axis of the mold (Figure 3.3(B)). The signature is smaller closer to the center of the mold. We speculate that this may be due to the difference in temperature with radial position, caused by the cementitious curing (which is an exothermic hydration reaction), resulting in different curing rates in different spatial position. Apparently, the rate of curing is higher close to the mold. Pouring the second layer after 15 minutes of curing results in a situation where the center of the first layer is less cured, allowing partial mixing with the newly poured second layer, which results in a minimal variation 36 Figure 3.2: Ultrasound response for samples with the single internal interface obtained by allowing the first cementitious layer to cure for 1 hour. Inserts schematically depict the locations creating the acquired echoes: The first echo is the partial internal reflection from the interface of the two layers; the second echo is from the end of the sample. Round-trip echoes from partial internal reflection and the end are also shown. in the ultrasound impedance of the two layers in the center of the sample. 3.1.2 Low Water-Cement Ratio Sample with Various Curing Times This group of samples were cast with two internal interfaces obtained by sequentially casting three cementitious layers, 1 inch thick, using the water-cement ratio of 0.36, which was selected to be as small as possible while still allowing for layered casting of the sample with minimal air entrapment. The cure time for the first cementitious layer was also set to 15,30,45 or 60 minutes for different samples before pouring the second layer. The final, third layer was poured after the sample was allowed to dry for one day. The ultrasound tests showed that all cementitious samples made according to this recipe exhibit partial internal reflections from each of two internal interfaces, even with short curing time of 15 minutes (Figure 3.4). However, the test also showed the signal-to-noise ratio (SNR) in this group (Figures 3.4,3.5) is high, likely 37 Figure 3.3: Ultrasound response for samples with the single internal interface obtained by allowing the first cementitious layer to cure for 45 minutes (A) and 15 minutes (B). Inserts schematically depict the locations creating the acquired echoes in (A) both the partial internal reflection and the end of the sample. The reflection from the interface is not well defined and varies with the change in the transducer position in (B). 38 Figure 3.4: Ultrasound response for samples with the two internal interfaces obtained by allowing the first cementitious layer to cure for 15 minutes, 1 day for the second layer, and at least 1 day for the final third layer. Water-cement ratio equal to 0.36 was used. Inserts schematically depict the location creating the acquired echoes. Three echoes are clearly visible, of which the first two are from the internal interfaces created by multiple layers cast sequentially. due to air entrapment during multiple castings of this mixture with a relatively low water content. 3.1.3 Medium Water-Cement Ratio Sample with Long Curing Time The water-cement ratio equal to 0.44 was used to make the next series of samples. This ratio was chosen to be roughly an average of 0.36 and 0.5 values used with the already described samples. Our goal is to find the composition that gives an optimal trade-off between the ease of casting air free layers, which is easier to achieve with large ratios, without sacrificing the hardened strength of cementitious after curing, observed with high water content in the mixture. The maximum of three layers were cast to make samples. Each layer was equal to or less than 1 inch in thickness. The purpose of these samples was to obtain initial experimental evidence on the shortest spacing between the interfaces, introduced to create internal ultrasound reflections, without overlap in consecutive echoes. By 39 Figure 3.5: Ultrasound response for samples with the two internal interfaces obtained by allowing the first cementitious layer to cure for 45 minutes (A) and 1 hour (B), both 1 day for the second layer, and at least 1 day for the final third layer. Water-cement ratio equal to 0.36 was used. Echoes from the second interface and the end of the sample are clearly visible. However, the signal from the first interface is rather complex, likely indicating entrapment of air at the interface. 40 avoiding such overlap the signal analysis and the measurements of the ultrasound TOF are simplified. Each layer of prepared samples in this batch was allowed to cure for a long time that varied from one day to several days before the next layer was poured. The ultrasound testing revealed that all samples in this group produce clear partial internal reflections from each interface. The reflections are clear without waveform overlap for 3-layered samples (Figure 3.6). This suggests that 1 inch or better spatial resolution of the ultrasound measurements of temperature distribu­tion may be achieved with the proposed approach. The signal distortions in this and other samples prepared following this recipe (such as the two-layer samples for which the results are shown in Figures 3.7) are small, which is important for the precise measurements of the ultrasound time of flight. Figure 3.6: Ultrasound response for samples with the two internal interfaces obtained by allowing the two initial cementitious layers to cure for 1 day each. The last layer was cured for at least 1 day. Water-cement ratio equal to 0.44 was used. Echoes from both interfaces are well defined, with relatively high SNR. 41 Time (s) -5 X 10 Figure 3.7: Ultrasound response for a sample with a single internal interface obtained by allowing the first cementitious layer to cure for 1 day (A) and 2 days (B). Water-cement ratio equal to 0.44 was used. Echoes from the interface and the end of the sample are well defined, with a relatively high SNR. 42 3.1.4 Summary of Partial Internal Reflection Structures 3.1.4.1 Composition It was found that the Portland cement that was used produces samples that often cracked at the interface after soaking in water. Two new cement formulations were then examined to obtain a more stable model of the refractory. 1. In the first case, a fortifier was added to the Portland cement mixture to increase strength of the samples, adhesion between multiple layers and reduce permeability to water. In the following, samples prepared with this modifier are designated as PF (Portland Fortified samples). 2. Rapid Set® concrete mixture is the second formulation that we tested. It produced fast setting, high strength samples, with excellent bond between layers and crack resistance. Samples prepared using Rapid Set mix are designated as RS. Samples prepared according to these two formulations have higher strength and less water permeability than samples prepared with the traditional Portland cement mix. The Rapid Set concrete samples were found to be much denser than other concrete samples. Their ultrasound waveforms are shown in Figure 3.8. 3.1.4.2 Water-Cement Ratio Higher water-cement ratios, such as samples produced using the ratio equal to 0.44 and 0.5, lead to better defined partial internal reflections of the ultrasound test pulse and the measurements are characterized by higher SNRs. This is most likely due to reduced trapping of air bubbles in samples with higher moisture content and more planar interfaces obtained by casting of more "fluid" cementitious mixtures. Typically, higher water content is associated with the reduced mechanical strength of cementitious samples. For a particular cementitious mix used by us, it is recommended that the water-cement ratio should not exceed 0.5. A lower water-cement ratio is known to lead to higher strength and durability, but may make casting of uniform layers more difficult. For a particular cement mixture used by us, a minimum of 0.25 of water-cement ratio is required for cementitious mix to harden. A mix with too much water would result in internal cracks and fractures which will reduce the final strength. A maximum of 0.5 of water-cement 43 Time (s) Time (s) Figure 3.8: Ultrasound waveform collected for samples with a single internal interface obtained using Portland Fortified (A) and Rapid Set® cast (B). Echoes from the interface and the end of the sample are well defined, with a relatively high SNR. x 10 x 10 44 ratio was used for making cementitious samples-the ratio that is still within the recommended range. 3.1.4.3 Air Bubbles Vibration on freshly poured cementitious layers helps to reduce the amount of trapped air bubbles significantly. The presence of trapped air has a very significant negative influence on the quality of the required echoes. At the same time, vibrating samples prior to setting of the previously cast layers (short cure time of 15 minutes or less) likely contributes to the lack or poor quality of the measured echo signals. To vibrate the samples immediately following the pour of the newly prepared cementitious mixtures, we used a Fisher Sonic Dismembrator (Model 300) which provides high-frequency vibrations (on the order of 20 kHz). 3.1.4.4 The Partial Curing Time For samples with water-cement ratio equal to 0.5 (the maximum value in the recommended range of the Portland cement mixture used by us), the curing time of less than 1 hour produced complex reflection patterns which suggest a diffused interface between sequentially cast layers. It was found that better defined reflections can only be observed on the edge of a cylindrical sample but not in the samples' center. To ensure the reflection can be observed everywhere on the sample surface, each layer curing time of at least 1 hour was found to be necessary. However, curing interlayers over days significantly reduced the bond effectiveness and sample strength at the interfaces. 3.1.4.5 Number and Spacing of Partial Reflections We have demonstrated the acquisition of partial internal reflections from two interfaces inside the samples obtained by a sequential casting of three cement layers, allowing some time to cure each layer. At this time, we have shown that with the central frequency of 1 MHz, the echoes do not overlap when interfaces are spaced 1 inch apart. The distance between the features (interfaces between layers, in this case) that produce partial reflections determines the achievable spatial resolution of the temperature distribution measurements. Though a 1 inch spacing was 45 demonstrated so far, a closer spacing of partial reflectors may be possible with higher frequency of excitation, the sharper defined changes in acoustic impedance and the more advanced signal analysis techniques that can handle overlapping echoes. 3.1.4.6 Sample Length We have demonstrated that with the current ultrasound pulser used in the pulse-echo mode, it is possible to characterize samples 2-4 inches long and obtain clear measurements of partial internal reflections from, at least, two internal interfaces. Longer samples (over 4 inches) can be characterized in transmission mode in which a separate transducer is used to generate the test pulse and the second transducer acquires the ultrasound signature after the excitation has propagated through the sample. 3.1.4.7 Consistency of Sample Properties Consistency of the sample properties prepared following the same recipe was tested with RS samples. Three 2-inch samples were obtained in a single casting and had no internal ultrasound reflectors. The other three samples in the group were 3 inches long and had a single partial internal reflector created by casting two cement layers of identical mixture and allowing for a partial cure between consecutive layers. All samples were prepared and made at the same condition, from the same mixture ratio. The speed of sound measurements for the two groups of samples were carried out, both in the water bath and the air. All samples were tested in the following order: (1) in air, (2) fully immerged in water and saturated, (3) in air again, after one week drying at room temperature. The time of flight was measured as a time delay between two echoes from the distal end of the sample produced by an ultrasound pulse acquired after a single or two round trips. SOS was determined as: c = 2L/tof , where the length of the sample, L, was measured with a micrometer. The results for all six samples are shown in Figure 3.9. The solid and hollow shapes represent the SOS results from three 2-inch and three 3-inch samples, respectively, presented by the primary y-axis. The secondary y-axis shows the densities of all 46 <x>O <x> 4320 4300 4280 4260 4240 4220 4200 4180 4160 4140 4120 □ X o *□ • 2" 1st Air for SOS ▲ 2" Water for SOS ■ 2" 2nd A ir for SOS 2 3" 1st Air for SOS 3" Water for SOS □ 3" 2nd A ir for SOS X 2" Density + 3" Density 2.26 2.24 2.22 2.2 2.18 2.16 2.14 2.12 2.1 2.08 2.06 Sample order Figure 3.9: Two groups of RS samples tested in the water bath and air show great variability in both the SOS versus temperature relationships and density measurements. samples. The results indicate some variability between samples in the same group, like significant SOS increments when sample is saturated, and higher variabilities between different groups, both in the SOS and the density. 3.2 Experiments of Temperature Measurements 3.2.1 Structured Cementitious Waveguide In this study, a cementitious sample obtained by casting of Portland Type I/II cement was selected as a model of castable refractory. A 2-inch I.D. PVC tubing was used as a mold and a 4-inch sample was created by sequentially casting four layers (each 1 inch thick) of cement mixture and allowing 30 minutes curing time between consecutive layers. The mold was vibrated by an external vibrator after each pouring to ensure uniform setting of each layer and to remove air bubbles. The samples were cured and aged at ambient temperature until their ultrasound properties stabilized. The ultrasound tests of cementitious samples were carried out using a Panamet-rics pulser/receiver (model 5072PR) and a Panametrics immersion transducer with a central frequency of 1 MHz (model V302), coupled to a sample using ultrasound 47 gel. The data were acquired using a Tektronix oscilloscope (model MSO 2024) interfaced to a computer. 3.2.2 TOF Acquisition from Ultrasound Waveform The other essential prior condition to apply the proposed ultrasound method is the SOS dependence on the temperature. Figure 3.10 illustrates typical ultrasound waveforms acquired in this configuration from the same sample maintained at different uniform temperatures. The echoes are produced at the distal end of the sample. The first echo seen in the figure corresponds to the ultrasound pulse that traveled the length of the sample and back (a single round trip), while the second measured echo corresponds to the same pulse after it made the second round trip through the sample. The direct inspection of waveforms in Figure 3.10 indicates that the speed of sound in the model refractory indeed depends on temperature, decreasing as the temperature goes up, leading to longer time of flight of ultrasound pulse at higher temperatures. To establish the correlation between the speed of sound and temperature, the 4- ■ a c o initial Pulse Echoes Time (psec) Figure 3.10: Ultrasound waveforms acquired at different temperatures illustrate the TOF shift with temperature. 48 inch sample was placed inside the fabricated heating fixture depicted in Figure 3.11, which consisted of a thermally insulated steel container and an internal heating blanket (silicon rubber blanket by BriskHeat®) that tightly surrounded the sample. The temperature of the blanket was measured by a thermocouple and controlled by a PID controller. The surface temperature of the sample was measured by four Omega Precision Fine Wire Thermocouples attached with high-temperature adhesive tape in the middle of each layer of the model refractory. Two additional thermocouples of the same type were used to measure the temperature of the top and bottom surfaces of the sample. The ultrasound transducer was coupled to the surface of the top layer of the sample (Layer 1, L1). To prevent damage to the transducer, the top surface of To Panametrics Pulser/Receiver Figure 3.11: Experimental setup of 4 inches, 4 segments cementitious sample for low temperature measurements. 49 Layer 1 extended above the fixture to allow for partial cooling of the sample; in this arrangement, Layer 1 is effectively used as a delay line. The test temperatures were changed in 10 °C increments and spanned from 20 to 100 °C in this study. After each temperature change, sufficient time was allowed for thermal equilibration to occur before attempting the time of flight measurements. The sequence of temperatures for which the SOS measurements were conducted was randomized. The randomization included all repeat experiments for each temperature. Such randomization avoids measurement potential bias introduced when temperatures always changed in one direction (either an increasing or decreasing temperature) and by a fixed increment of temperatures from one experiment to the next. We noticed that the interface between the consecutive layers is not entirely flat or smooth (a consequence of coning, partial penetration and/or mixing between layers). To account for unevenness of the interfaces, during the repeat experiments the positioning of the transducer was slightly shifted relative to the centerline of the sample and kept at the same location for all the tests. 3.2.3 Signal Processing Several methods for determining the TOF from the waveforms typified in Figures 3.10 were investigated. Our initial approach was to use a delay line and the echo signal from sample-delay interface as a reference "zero time" from which the time of flight is calculated. The time of flight is then calculated by matching single-point features (e.g., peak value or zero crossing) in the reference waveform from the sample-delay interface and the waveform of the reflection produced by internal interfaces and the end of the sample. Though this approach is standard, we encountered difficulties in its applications. Cementitious refractory materials are dissipative and have higher absorption of higher frequency components of the ultrasound wave, which leads to distortion and broadening of the echo waveform and thus errors in determining the time of flight based on a single-point feature matching. We therefore opted to use the cross-correlation between the echo waveforms obtained at different temperatures to determine the difference in the time of flight at two different temperatures, ATOF. 50 For dissipative samples higher accuracy may be obtained if cross-correlation is performed between the analytical envelopes of the waveforms, rather than the waveforms themselves [59]. To test the potential improvements, we implemented the envelope cross-correlation method and compared its performance with the TOF measurements based on the waveform cross-correlation. A numerical procedure based on the Hilbert Transform is applied to the waveform in the time domain to create an envelope of the waveform. This is a representation of the amplitude modulation on the carrier wave frequency. The procedure also creates a time-domain phase function which has application to interpreting dispersion. Figure 3.12 shows the comparison of results obtained using the cross-correlation between the waveforms (left 2 columns) and the envelopes of the waveforms. The offsets (ATOF) between the reference waveform and its envelope (green lines) and the waveforms and their envelopes of the echo signals acquired when the sample was maintained at different temperatures (blue lines) are listed for each subfigure. The two meth