Documents

Studies of Relationships Between Free Swelling Index (FSI) and Coal Quality By

Description
swelling index and coal quality
Categories
Published
of 7
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
Share
Transcript
  Studies of relationships between Free Swelling Index (FSI) and coal quality byregression and Adaptive Neuro Fuzzy Inference System M. Tayebi Khorami a , S. Chehreh Chelgani b , James C. Hower c, ⁎ , E. Jorjani a a Department of Mining Engineering, Science and Research Branch, Islamic Azad University, Poonak, Hesarak Tehran, Iran b Surface Science Western, Research Park, University of Western Ontario, London, Ont., Canada N6G0J3 c Center for Applied Energy Research, University of Kentucky, 2540 Research Park Drive, Lexington, KY 40511, USA a b s t r a c ta r t i c l e i n f o  Article history: Received 21 June 2010Received in revised form 30 September 2010Accepted 30 September 2010Available online 8 October 2010 Keywords: Free Swelling IndexCoking coalCoal petrographyUltimate analysisProximate analysisAdaptive Neuro Fuzzy Inference System The results of proximate, ultimate, and petrographic analysis for a wide range of Kentucky coal samples wereused to predict Free Swelling Index (FSI) using multivariable regression and Adaptive Neuro Fuzzy InferenceSystem (ANFIS). Three different input sets: (a) moisture, ash, and volatile matter; (b) carbon, hydrogen,nitrogen, oxygen, sulfur, and mineral matter; and (c) group-maceral analysis, mineral matter, moisture,sulfur, and R  max  were applied for both methods. Non-linear regression achieved the correlation coef  󿬁 cients(R  2 ) of 0.38, 0.49, and 0.70 for input sets (a), (b), and (c), respectively. By using the same input sets, ANFISpredicted FSI with higher R  2 of 0.46, 0.82 and 0.95, respectively. Results show that input set (c) is the bestpredictorofFSIinbothpredictionmethods,andANFISsigni 󿬁 cantlycanbeusedtopredictFSIwhenregressionresults do not have appropriate accuracy.© 2010 Elsevier B.V. All rights reserved. 1. Introduction All types of coal undergo chemical changes, but caking coal exhibitsphysicalchangesinadditiontochemicalchangeswhenheated.(Speight,1994,2005;GhoshandChatterjee,2008).Theseriesofphysicalchangesincakingcoalsduringtheheatingprocessaresoftening,melting,fusing,andresolidifying. These changes are within a speci 󿬁 c temperature range, theplasticrangeofcoal,andthephysicalchangesthatoccurwithinthisrangeare known as the plastic properties (Speight, 1994, 2005).Knowledge of the plastic properties of coal is commonly used topredictcokequality(Pierronetal.,1959;Habermehletal.,1981;Lloydetal., 1990; Riley, 2007). Caking properties are an essential prerequisite forcoking coals (Miller, 2005); subsequently, a number of tests have beendevised to classify the caking properties of coals, including the Roga test,Gray – King test, and Free Swelling Index (Speight, 1994, 2005).The Free Swelling Index (FSI), an important property for bothmetallurgical and steam coals (Hower and Eble, 1996), is a measure of the increase in volume of powdered coal when heated under certainconditions.Inthismethod,1goffreshpowderedcoalsample( − 250 μ  m)isplacedinasilicacrucible.Thecoalisleveledbylightlytapping,covered,and heated to 800°C in a special furnace for 2.5min. After cooling, thecross sectional pro 󿬁 le of the coke button is compared with a series of standard pro 󿬁 les and designated by numbers from zero to nine at anintervalof0.5inincreasingorderofsize;coalwithzeroFSIbeingtheleastswelling. According to this standard, FSI classi 󿬁 ed to 0 – 2, 2 – 4 and 4 – 9,which show weakly, medium, and strongly caking ranges (Table 1)(Speight,1994,2005;ASTMD-720,1999;Riley,2007).Coalsthatarelowin rank, such as lignite, or high in rank, such as anthracite do not usuallyfuseandshownoswellingvalue(Speight,1994,2005;Miller,2005;Riley,2007; Ghosh and Chatterjee, 2008), whereas the Free Swelling Index of bituminous coal enhances as the rank increases (the peaking in mediumvolatilebituminousrankrange)(Rees,1966;Speight,1994,2005;Howeret al., 1994; Riley, 2007).Therelationshipbetweencoal'splasticityanditsphysicalandchemicalpropertieswasstudiedinpreviousworks(SchapiroandGray,1966;Lloydet al., 1984, 1989, 1990; Cole and Williams, 1981; Clark et al., 1984;Reasoner et al., 1985; Hower et al., 1994; Maroto-Valer et al., 1998;Speight, 2005). The caking tendency of coals rises dramatically between25 and 35% w/w volatile matter content (orbetween 81 and 92% w/wof carbon in coals, with a maximum at 89% carbon) and then decreases(Speight, 1994, 2005; Ghosh and Chatterjee, 2008). The amounts of mineral matter content of the coal are signi 󿬁 cant in coke productionbecause of diluting effect of ash (Miller, 2005); subsequently, the cakingtendency of coal decreases with increasing of mineral matter (Speight,1994, 2005).Fluidity is also dependent upon the organic sulfur content in coal(Hower et al., 1994), which is also important for metallurgical processes.Clarketal.(1984)measuredorganicsulfurincoalandsemicokesamples International Journal of Coal Geology 85 (2011) 65 – 71 ⁎  Corresponding author. Tel.: +1 859 257 0261; fax: +1 859 257 0360. E-mail address:  hower@caer.uky.edu (J.C. Hower).0166-5162/$  –  see front matter © 2010 Elsevier B.V. All rights reserved.doi:10.1016/j.coal.2010.09.011 Contents lists available at ScienceDirect International Journal of Coal Geology  journal homepage: www.elsevier.com/locate/ijcoalgeo  fromKentucky,andobservingatrendofincreasingorganicsulfurcontentwith increasing plasticity of coals at the same rank. Cole and Williams(1981) and Senftle and Davis (1982) found that plasticity was high onlyfor the coals with high organic sulfur content. Coals with similar maceralcomposition and rank show a greater tendency to undertake cokeformation upon as the organic sulfur content increased (Ignasiak et al.,1978; Yarzab et al., 1980).Petrographic analysis can provide insight into the thermoplasticproperties of a particular coal, which are of signi 󿬁 cant importance inthe coking industry (Miller, 2005). They display different carboniza-tion behaviors, as expected from their different physicaland chemicalcharacteristics(Krevelen,1993;Stachetal.,1982;FalconandSnyman,1986).SchapiroandGray(1966)classi 󿬁 edthemaceralsinto ‘ reactive ’ including liptinite and vitrinite that enhance the  󿬂 uidity of coal, and ‘ inert ’ , such as inertinite that is infusible during carbonization. FreeSwelling Index is also dependent on the maceral composition, withvitrinitebeingthemostobviouscontributortotheswellingpropertiesin most coals and inertinite group-macerals act as diluent compo-nents, deleterious to swelling (Hower and Eble, 1996).Hower et al. (1994) predicted coking characteristics of a number of easternKentuckycoalsandfoundthatFSIincreasedwithincreasingrank(determinedasvitrinitemaximumre 󿬂 ectance).FSIcanbeaffectedbytherelative amount of fusible components, vitrinite and liptinite, and of diluentinertcomponents.Maroto-Valeretal.(1998)studiedthein 󿬂 uenceofpetrography(densityfractionationofpetrography)on 󿬂 uiditypropertyof coking coals and obtained a signi 󿬁 cant linear relationship (R  2 N 0.96)between 󿬂 uidityandmaceraldensity, 󿬁 ndingthat 󿬂 uiditydecreaseswithincreasing vitrinite and semifusinite density. Vitrinite re 󿬂 ectance, as ameasure of coal rank, is perhaps the most signi 󿬁 cant parameterin 󿬂 uencing the coke making potential of any coking coal (Ghosh andChatterjee, 2008).Some problems associated with the ASTM D-720 FSI method arethe proper heating rate, oxidation or weathering of the coal sample,andanexcessof  󿬁 necoalintheanalysissample(Speight,1994,2005).For more clari 󿬁 cation, a gas burner as a source of heat requiresfrequent calibration because gas pressures are not constant, relativelysmallchangesingaspressureresultinginratherwidevariationsinthetemperatureattainedinacrucible,andthusintheresultantsizeofthecoke button. Furthermore, the quartz crucibles speci 󿬁 ed for the testare not standard equipment, and relatively small variations indimension and wall thickness from those speci 󿬁 ed often result inrather wide variations in the resultant size of the button and the sizespeci 󿬁 ed in the test is not normally used for any other purpose in thelaboratory; So, prediction of FSI from coal analysis data can be useful(Swartzman and Behnke, 1952).Using the experimental data, computing techniques have beenapplied to many aspects of coal processing that were mentioned in thereferences of the paper (Cilek, 2002; Yao et al., 2005; Patel et al., 2007; Jorjanietal.,2007,2008;ChehrehChelganietal.,2008,2010).AdaptiveNeuro Fuzzy Inference System (ANFIS) is one of the most popular andwell documented neural fuzzy systems, which has a good softwaresupport(Lot 󿬁 ,1995). Jangetal.(1997)presentedtheANFISarchitecture andapplicationexamplesinmodellinganon-linearfunction,adynamicsystem identi 󿬁 cation and a chaotic time series prediction. Given itspotential in building neural fuzzy models with good predictioncapabilities (Cai et al., 2003), the ANFIS architecture was chosen formodelling of this work.The aim of the present work is prediction of FSI of coal according tothe proximate and ultimate analysis, group-macerals, mineral matter,and vitrinite maximum re 󿬂 ectance (R  max ) of coal using experimentaldata obtained at a laboratory level. The multivariable regression andAdaptiveNeuroFuzzyInferenceSystem(ANFIS)wasusedtopredict.Toourknowledge,thisisthe 󿬁 rsttimethatANFIShasbeenusedtopredictFree Swelling Index of coal by using mentioned input variables. 2. Experimental data A remarkable mathematical model requires a comprehensivedatabase to cover a wide variety of coal types. That model should be  Table 1 The numbers of samples in database regarding their caking ranges.Coal type Caking range No.Noncaking/weakly caking 0 – 2 107Medium caking 2 – 4 349Strongly caking 4 – 9 352  Table 2 The range of variables for 808 Kentucky coal samples.Variable (%) Minimum Maximum Mean Standard deviationMoisture 0.84 15.56 4.29 2.86Ash 0.64 19.95 8.80 4.46Volatile matter 24.80 44.20 35.38 2.79Fixed carbon 37.65 66.46 51.53 5.92Total sulfur 0.42 8.58 2.11 1.67Pyritic sulfur 0.01 6.50 0.90 1.06Sulfate sulfur 0.00 1.08 0.07 0.13Organic sulfur 0.26 4.66 1.13 0.70Carbon 55.95 84.38 71.71 6.30Hydrogen 4.00 6.60 5.21 0.34Nitrogen 0.80 2.34 1.50 0.20Oxygen 3.60 21.65 10.67 3.30Mineral matter 1.1 24.8 10.7 5.3Vitrinite 21.0 90.6 70.1 11.6Inertinite 0.9 52.6 13.3 8.3Liptinite 0.4 25.8 6.0 3.8R  max  0.39 1.12 0.79 0.17 4.002.000.00-2.00-4.00-6.00 Difference between actual andpredicted FSI 120100806040200    F  r  e  q  u  e  n  c  y Mean~0.00Std. Dev.=1.51N=808 Fig. 1.  Normal distribution of the difference between measured FSI values andestimated FSI values obtained from multivariate regression Eq. (2).  Table 3 The percentage of accurate prediction of FSI by non-linear regression and ANFISprocedure (in testing stage) regarding different caking ranges.Input sets CakingrangeNon-linearregression (%)ANFIS(%)Proximate analysis 0 – 2 0.93 5.562 – 4 65.61 82.764 – 9 78.41 78.79Ultimate analysis, mineral matter 0 – 2 0.93 202 – 4 69.34 74.144 – 9 81.82 100Group-maceral analysis, mineralmatter, moisture, organic sulfur, R  max 0 – 2 29.90 902 – 4 67.62 91.844 – 9 83.81 98.7866  M.T. Khorami et al. / International Journal of Coal Geology 85 (2011) 65 – 71  capable for predicting of FSI with a high validity. Data used to test theproposed approaches are from studies conducted at the University of Kentucky Center for Applied Energy Research. The database includesthe determined proximate and ultimate analysis, petrography,mineral matter, and R  max  analysis as well as Free Swelling Index(FSI) on an as determined basis. More than 800 coal sample analyses(generally falling within the high volatile bituminous rank range)were used. Table 1 shows the number of samples regarding theircoking ranges. 3. Adaptive Neuro Fuzzy Inference System (ANFIS) The main objects of fuzzy systems that decision-making by usingtheknowledgeaboutatarget,includehumanknowledgeandperforminterfacing (Mohanadas and Karimulla, 2001). Fuzzy inference is theprocessofformulatingthemappingfromagiveninputtoanoutputbyusing fuzzylogic ( Jang et al.,1997). In addition,fuzzy logic is all aboutthe relative importance of precision (Lot 󿬁 , 1995) that is closer tohuman thinking and natural language than conventional logicalsystems (Zadeh, 1965).Thefuzzyinferencesystem(FIS)basedontheconceptsoffuzzysettheory, fuzzy if-then rules, and fuzzy reasoning ( Jang et al., 1997) canbeclassi 󿬁 edintothreetypes:Tsukamoto-typeFIS(Tsukamoto,1979),Mamdani-type FIS (Mamdani and Assilan 1975), and Takagi – Sugeno-type FIS (Sugeno and Kang 1988).The methodology of the fuzzy logic controller (FLC) when theprocesses are too complex for analysis or the available sources of informationareinterpretedqualitatively,inexactly,orwithuncertainty,appears very useful by conventional quantitative techniques (Takagiand Sugeno, 1974). The major problems in the fuzzy logic control thathave reduced its application are identifying the appropriate number of rules,thedif  󿬁 cultyofchoice,anddesignofmembershipfunctionstosuita given problem (Mohanadas and Karimulla, 2001; Castellano andFanelli,1996).Neuralnetworksarecapableoflearning,buttheycannotinterpret imprecise data that can be helpful in making decisions. Thislearning capability of the neural network can be combined with thecontrol capabilities of a fuzzy logic system resulting in a neuro-fuzzyinference system ( Jang, 1993; Jantzen, 1998).An Adaptive Neuro Fuzzy Inference System (ANFIS), based on thearchitecture oftheTakagi – Sugeno-typefuzzyinferencesystem,isoneof the most popular neural fuzzy systems (FIS) (Lot 󿬁 , 1995; DemuthandBeale,2002).ANFISdevelopedby Jang(1993)hasagoodsoftware support (Lot 󿬁 , 1995) by using input/output data set, constructs afuzzy inference system (FIS), and tuning it with a back propagationalgorithm alone, or in combination with a least squares type of method(Lot 󿬁 ,1995;Jang,1993;JangandSun,1995;Jangetal.,1997).Fuzzy clustering is the partitioning of a collection of data into fuzzysubsets or clusters based on similarities between the data (Passino andYurkovich,1998)thatdevelopsafuzzyestimationmodel,topredicttheoutput given the input (Lot 󿬁 , 1995). There are some basic methods of fuzzy clustering: fuzzy C-means (FCM) clustering method, the mostpopular method, and subtractive clustering (Lot 󿬁 , 1995; Grabusts,2002). The fuzzy C-means clustering algorithm requires a desirednumberofclusters(C)andinitialguesspositionsforeachclustercenter.Subtractiveclusteringhasanauto-generationcapabilityfordeterminingthe number and initial location of cluster centers in a set of data whenthere is not a clear idea how many clusters should be for a given set of data. This method partitions the data into groups called clusters byspecifyingaclusterradius,andgeneratesaSugeno-typefuzzyinferencesystem(FIS)withtheminimum numberofrulesaccordingtothefuzzyqualitiesassociatedwitheachoftheclusters.ThistypeofFISgenerationcan be accomplished automatically using generate fuzzy inferencesystem structure (GENFIS2) (Lot 󿬁 , 1995). 4. Results and discussion 4.1. Multivariable regression Usingallparametersofproximateorultimateanalysisinoneequationmay increase the bias. For example, proximate analysis contains the  Table 4 FSI estimation deviations from target for non-linear regression equations.FSI deviationfrom targetModel (a) Model (b) Model (c)0 – 2 2 – 4 4 – 9 0 – 2 2 – 4 4 – 9 0 – 2 2 – 4 4 – 9Less than 0.5 0.00% 38.68% 21.88% 4.67% 39.26% 24.15% 18.69% 38.68% 32.39%Between 0.5 and 1 12.15% 30.66% 18.47% 15.89% 30.95% 23.58% 30.84% 32.95% 26.14%More than 1 87.85% 30.66% 59.66% 79.44% 29.80% 52.27% 50.47% 28.37% 41.48% 4.002.000.00-2.00-4.00-6.00 Difference between actual andpredicted FSI 120100806040200    F  r  e  q  u  e  n  c  y Mean~0.00Std. Dev.=1.36N=808 Fig. 2.  Normal distribution of the difference between measured FSI values andestimated FSI values obtained from multivariate regression Eq. (4). 3.002.001.000.00-1.00-2.00-3.00 Difference between actual andpredicted FSI 100806040200    F  r  e  q  u  e  n  c  y Mean~0.00Std. Dev.=1.04N=808 Fig. 3.  Normal distribution of the difference between measured FSI values andestimated FSI values obtained from multivariate regression Eq. (6).67 M.T. Khorami et al. / International Journal of Coal Geology 85 (2011) 65 – 71  determination of moisture, volatile matter, ash, and  󿬁 xed carbon (ASTMD-3172 – D-3175).Fixedcarbondependsonrelativeamountsofmoisture,ash,andvolatilematterthenthedifferenceofthesethreevaluessummedand subtracted from 100, Fixed Carbon=100 − (Moisture+Ash+Volatile Matter). It is not necessary to use all four parameters since, byde 󿬁 nition; the four parameters are a closed system, adding to 100%(Hower, 2006).Topreparelinearequationsthestepwisevariableselectionprocedurewasused,inwhichvariablesaresequentiallyenteredintothemodel.The 󿬁 rst variable considered for entering into the equation is the one withlargest positive or negative correlation with dependent variable. Theprocedurestopswhentherearenovariablesthatmeettheentrycriterion(SPSS Inc, 2004).Unlikelinearregression,whichisrestrictedtoestimatelinearmodels,non-linear regressions create equations with arbitrary relationshipsbetween independent and dependent variables. In this study, someinput variables have non-linear relationships with FSI. Therefore, withrespect to linear regressions, non-linear regression was used to developequations between Free Swelling Index and coal analysis variables.In order to cover the whole structure of coal samples, threedifferent input sets of coal analyses were applied as FSI predictors:(a) Proximate analysis (moisture, ash, and volatile matter);(b) Ultimate analysis (carbon, hydrogen, nitrogen, oxygen,and sulfur), forms of sulfur, and mineral matter;(c) Group-macerals analysis (vitrinite, inertinite, and liptinite),mineral matter, moisture, vitrinite R  max , and forms of sulfur.The statistical parameters of input variables are shown in Table 2.According to the stepwise least square mathematical method (SPSSsoftware) the best variables from each combination were selected topredict FSI. 4.1.1. Proximate analysis By a least square mathematical method, the inter correlations of moisture, ash, and volatile matter with FSI were − 0.52, − 0.23, and − 0.16, respectively. The results show that higher moisture contentsin coal can result lower FSI. The moisture content is, of course, a coalrank parameter. Moisture as a diluent and moisture as an indicator of coal rank are inseparable, providing a good example of the inter-relationship of coal quality parameters. The other variables are notsigni 󿬁 cant.AlinearrelationshipbetweeninputvariablesandFSIcanbeshownas following equation:FSI = 10 : 097 − 0 : 313 M − 0 : 091  A  − 0 : 103  VM  R  2 = 0 : 32 :  ð 1 Þ In addition, the non-linear relationship as following equation: FSI = 9 : 318 − 1 : 781 M − 0 : 054  A  − 0 : 022  VM  + 0 : 21 M 2 − 0 : 008 M 3 R  2 = 0 : 38 : ð 2 Þ Where M, A, and VM denote the percentage of moisture, ash, andvolatile matter, respectively. The other non-linear equations wereexamined, and they did not improve the correlation coef  󿬁 cient. 9.008.007.006.005.004.003.002.001.000.00    P  r  e   d   i  c   t  e   d   F   S   I 9.008.007.006.005.004.003.002.001.000.00 Actual FSI Predicted FSI=1.287+0.701 Actual FSIR2=0.70 Fig. 4. Graphical comparison of experimental FSIs with those estimated by multivariateregression Eq. (6).  Table 5 Parameters of ANFIS based on different input models.Modelno.Input variable No. of trainingepochTrainingdata radiusNo. of rules(a) Proximate analysis 2 0.61 4(b) Ultimate analysis, mineral matter 3 0.9 2(c) Group-maceral analysis, mineral matter,moisture, organic sulfur, R  max 65 0.98 2 Fig. 5.  Normal distribution of the difference between measured FSI values andestimated FSI values obtained from ANFIS (model (a)). 2.001.000.00-1.00-2.00 Difference between actual and ANFISpredicted FSI 50403020100    F  r  e  q  u  e  n  c  y Mean=0.2Std. Dev.=0.53N=200 Fig. 6.  Normal distribution of the difference between measured FSI values andestimated FSI values obtained from ANFIS (model (b)).68  M.T. Khorami et al. / International Journal of Coal Geology 85 (2011) 65 – 71
Search
Similar documents
View more...
Tags
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks