O. Rama MohanaRao, K.VenkataSubbaiah, K.NarayanaRao, T. SrinivasaRao / InternationalJournal of Engineering Research and Applications (IJERA) ISSN: 2248-9622 www.ijera.comVol. 3, Issue 3, May-Jun 2013, pp.635-641635 | P a g eApplication of Univariate Statistical Process Control Charts forImprovement of Hot Metal Quality- A Case StudyO. Rama MohanaRao*, K.VenkataSubbaiah**, K.NarayanaRao***, T.SrinivasaRao*****Human Resource Department, Visakhapatnam Steel plant, Visakhapatnam**Department of Mechanical Engg, Andhra University, Visakhapatnam***Department of Mechanical Engg, Govt. Polytechnic, Visakhapatnam****Design &Engg, Department, Visakhapatnam Steel plant, VisakhapatnamAbstractStatistical Process Control (SPC)techniques are employed to monitor productionprocesses over time to detect changes. The basicfundamentals of statistical process control andcontrol charting were proposed by WalterShewhart. Shewhart chart can be used formonitoring both the mean and the variance of aprocess, however sensitivity of chart to shiftsin the variance is often considered inadequate.So, it is common to use the chart coupled witheither R chart or S chart, to monitor changes inmean and variance of process. This paperpresents the application of univariate controlchart for monitoring hot metal making processin a blast furnace of a steel industry forcontinuous quality improvement.Key Words - Control Chart, Regression Analysis,Statistical Process Control, Univariate, chartIntroductionStatistical process control is defined as theapplication of statistical techniques to control aprocess. SPC is concerned with quality ofconformance. There are a number of tools availableto the quality engineer that is effective for problemsolving process. The seven quality tools arerelatively simple but very powerful tools whichevery quality engineer should aware. The tools are:flow chart, run chart, process control chart, checksheet, pareto diagram, cause and effect diagram,and scatter diagram (Juran&Gryna, 1998) .The primary function of a control chart isto determine which type of variation is present andwhether adjustments need to be made to theprocess. Variables data are those data which can bemeasured on a continuous scale. Variable data areplotted on a combination of two charts- usually achart and a range (R) chart. The chart plotssample means. It is a measure of between-samplevariation and is used to assess the centering andlong term variation of the process. The range chartmeasure the within sample variation and asses theshort term variation of the process.A control chart is a statistical tool used todistinguish between variation in a process resultingfrom common causes and variation resulting fromspecial causes. It presents a graphic display ofprocess stability or instability over time. Everyprocess has variation. Some variation may be theresult of causes which are not normally present inthe process. This could be special cause variation.Some variation is simply the result of numerous,ever-present differences in the process. This iscommon cause variation. Control Chartsdifferentiate between these two types of variation.One goal of using a Control Chart is to achieve andmaintain process stability. Process stability isdefined as a state in which a process has displayeda certain degree of consistency in the past and isexpected to continue to do so in the future. Thisconsistency is characterized by a stream of datafalling within control limits based on plus or minus3 Sigma (standard deviation) of the centerline.Control charts are useful, i) To monitor processvariation over time ii) To differentiate betweenspecial cause and common cause variation iii) Toassess the effectiveness of changes to improve aprocess iv) To communicate how a processperformed during a specific period. There aredifferent types of control charts, and the chart to beused is determined largely by the type of data to beplotted. Two important types of data are:Continuous (measurement) data and discrete (orcount or attribute) data. Continuous data involvemeasurement. Discrete data involve counts(integers). For continuous data that are chart, Rchart are often appropriate.SPC is founded on the principle that aprocess will demonstrate consistent results unless itis performed inconsistently. Thus, we can definecontrol limits for a consistent process and checknew process outputs in order to determine whetherthere is a discrepancy or not. In the manufacturingarena, it is not difficult to figure out therelationship between product quality and thecorresponding production process. Therefore wecan measure process attributes, work on them,improve according to the results and produce high
O. Rama MohanaRao, K.VenkataSubbaiah, K.NarayanaRao, T. SrinivasaRao / InternationalJournal of Engineering Research and Applications (IJERA) ISSN: 2248-9622 www.ijera.comVol. 3, Issue 3, May-Jun 2013, pp.635-641636 | P a g equality products. There is a repetitive production ofthe same products in high numbers and this bringsan opportunity to obtain large sample size for themeasured attributes. Moreover, the product isconcrete, and the attributes and variables to bemeasured are easily defined. Consequently, theonly difficulty left is to define correct attributes andcollect data for utilizing the tools of StatisticalProcess Control.Literature reviewMany businesses use Univariate StatisticalProcess Control (USPC) in both theirmanufacturing and service operations. Automateddata collection, low-cost computation, products andprocesses designed to facilitate measurement,demands for higher quality, lower cost, andincreased reliability have accelerated the use ofUSPC.A more modern approach for monitoringprocess variability is to calculate the standarddeviation of each subgroup and use these values tomonitor the process standard deviation(Montgomery & Runger, 2003). Samanta andBhattacherjee (2004) analyzed qualitycharacteristic through construction of the Shewartcontrol chart for mining applications. Woodall andFaltin, F. W. (1993)  presented an overview andperspective on control charting. The role of SPC inunderstanding, modeling, and reducing variabilityover time remains very important. Weller (2000) discussed some practical applications ofStatistical Process Control. Mohammed (2004) adopted Statistical Process Control to improve thequality of health care. Mohammed et al.,(2008) illustrated the selection and construction of fourcommonly used control charts(xmr-chart, p-chart,u-chart, c-chart) using examples from healthcare.Grigg et al., (1998) presented a case studyStatistical Process Control in fish productpackaging. Srikaeo, K., & Hourigan, J.A. (2002) discussed the use Statistical Process Control toenhance the validation of critical control points(CCPs) in shell egg washing. Rashed (2005) made a performance Analysis of Univariate andMultivariate Quality Control Charts for OptimalProcess Control. Statistical Process Controlinvolves measurements of process performance thataim to identify common and assignable causes ofquality variation and maintain process performancewithin specified limits. (Mukbelbaarz, 2012).Sharaf El-Din et al (2006) , made a comparisonof the univariate out-of-control signals with themultivariate out-of-control signals using a casestudy of Steel making.Methodology3.1 Identification of critical process variablesGenerally, not all quality attributes andprocess variables are equally important. Some ofthem may be very important (critical) for quality ofthe product performance and some of them may beless important. The practitioners should know whatinput variables need to be stable in order to achievestable output, and then these variables areappropriately to be monitored. The critical processvariable of the process may be identified byRegression Analysis. Regression analysis is astatistical technique for estimating the relationshipsamong variables in process and to predict adependent variable(s) from a number of inputvariables.T-Statistics is an aid in determiningwhether an independent variable should beincluded in a model or not. A variable is typicallyincluded in a model if it exceeds a pre-determinedthreshold level or ‘critical value’. Thresholds aredetermined for different levels of confidence. Fore.g. to be 95% confident that a variable should beincluded in a model, or in other words to tolerateonly a 5% chance that a variable doesn’t belong ina model, a T-statistic of greater than 1.98 (if thecoefficient is positive) or less than -1.98 (if thecoefficient is negative) is considered statisticallysignificant.3.2 Construction of Control ChartsTo produce with consistent quality,manufacturing processes need to be closelymonitored for any deviations in the process. Properanalysis of control charts that are used to determinethe state of the process not only requires a thoroughknowledge and understanding of the underlyingdistribution theories associated with control charts,but also the experience of an expert in decisionmaking. There are many different types of controlchart and the chart to be used is determined largelyby the type of data to be plotted. This paperformulates Shewhart mean ( ) and R- Chart fordiagnosis and interpretation.Case studyThe hot metal production process in BlastFurnace of an integrated Steel Plant is shown inFig. 1. The purpose of a blast furnace is tochemically reduce and physically convert ironoxides into liquid iron called "hot metal". The blastfurnace is a huge, steel stack lined with refractorybrick, where the inputs are iron ore, sinter, cokeand limestone are dumped into the top, andpreheated air (sometimes with Oxygen Enrichment)is blown into the bottom.
O. Rama MohanaRao, K.VenkataSubbaiah, K.NarayanaRao, T. SrinivasaRao / InternationalJournal of Engineering Research and Applications (IJERA) ISSN: 2248-9622 www.ijera.comVol. 3, Issue 3, May-Jun 2013, pp.635-641637 | P a g eFig.1: Process flow diagram of Hot Metal processin Blast FurnaceIn which, inputs like sinter, coke are pre-processed before using in Blast Furnace. The hotair that was blown into the bottom of the furnaceascends to the top after going through numerouschemical reactions. These raw materials require 6to 8 hours to descend to the bottom of the furnacewhere they become the final product of liquid ironand slag. The output liquid iron known as hot metaldrained from the furnace at regular intervals fromthe bottom through tap hole.The hot metal with lower silicon andSulphur contents is required for the production ofSteel at Steel Melt Shop. Blast Furnace is supposedto supply the hot metal with the followingcomposition to reduce defectives in Steel making atSteel Melt Shop (SMS).Silicon (Si) = 0.3 - 0.60%Manganese (Mn) = 0.0 - 0.25%Phosphorous (P) = 0.0 - 0.15%Sulphur (S) = 0.0 - 0.04%To produce desired quality hot metal, it isessential to identify the critical process variablefrom the given inputs and optimize them.The various inputs for this process areBlast Volume (M3/Min), Blast Pressure (Kg/cm2),Blast Temperature (0C), Steam (t/hr.), OxygenEnrichment(%), Oxygen (M3/hr.), Ash, Moisture,Volatile material, Fe(%),FeO (%), SiO2(%),Al2O3(%), CaO (%), MgO (%), Mn (% ), SiO,Sulpher (S), Phosphorus(P), Manganese (Mn),Silica(Si), MnO (%) etc. The production data with370 observations grouped by 46 days was collectedfor the study.Results & Discussion5.1 Identification of critical process variablesIn order to understand the relationshipbetween the input and output variables of the hotmetal, the data is analyzed and Regression analysishas been carried out with the help of MINITABsoftware. In the analysis, each output variable istested individually to find out relationship betweeninput process variables. A set of data containingobservations on 370 samples were analyzed. Theregression equation for each output variable is asfollows:(1) The regression equation for Silicon toinput variables isSi = - 36.8 - 0.000016 Blast Volume (M3/Min) +0.864 Blast Pressure (Kg/cm2)- 0.830 Top Pressure(Kg/cm2) - 0.00130 Blast Temp (oC) + 0.00413Steam (t/hr) + 0.0402 % Oxygen Enrichment -0.000014 Oxygen (M3/hr.) + 0.420 Ash - 0.154Moist + 0.524 VM + 0.389 FC - 0.0165 %Fe +0.0397 %FeO- 0.372 %SiO2 + 0.793 %Al2O3 +0.0609 %CaO - 0.0220 %MgO + 0.973 %Mn- 4.16SiO2Table 1. T & P values for Silicon.Predictor T pConstant -0.47 0.636Blast Volume -0.24 0.811Blast Pressure 1.89 0.06Top Pressure -1.75 0.08Blast Temp -3.34 0.001Steam (t/hr) 1.18 0.239Oxygen Enrichment 0.69 0.493Oxygen -0.85 0.397Ash 0.55 0.586Moist -1.06 0.291VM 0.67 0.501FC 0.5 0.616%Fe -0.24 0.814%FeO 2.55 0.011%SiO2 -3.66 0.000%Al2O3 3.36 0.001%CaO 1.1 0.271%MgO -0.33 0.739%Mn 3.19 0.002SiO2 -3.09 0.002(2) The regression equation for Manganese toinput variables isMn = - 7.00 + 0.000012 Blast Volume (M3/Min) -0.0169 Blast Pressure (Kg/cm2) - 0.0141 TopPressure (Kg/cm2) + 0.000314 Blast Temp (oC) +0.000390 Steam (t/hr) - 0.0115 % OxygenEnrichment + 0.000003 Oxygen (M3/hr.) + 0.0622Ash - 0.0049 Moist + 0.106 VM + 0.0746 FC -0.00946 %Fe + 0.00173 %FeO - 0.0406 %SiO2 +0.0784 %Al2O3 + 0.0166 %CaO - 0.0233 %MgO+ 0.141 %Mn - 0.187 SiO2
O. Rama MohanaRao, K.VenkataSubbaiah, K.NarayanaRao, T. SrinivasaRao / InternationalJournal of Engineering Research and Applications (IJERA) ISSN: 2248-9622 www.ijera.comVol. 3, Issue 3, May-Jun 2013, pp.635-641638 | P a g eTable 2. T & P values for Manganese.Predictor T pConstant -0.73 0.465Blast volume 1.43 0.154Blast Pressure -0.3 0.765Top Pressure -0.24 0.81Blast Temp 6.55 0.000Steam (t/hr) 0.9 0.367Oxygen Enrichment -1.59 0.113Oxygen 1.54 0.125Ash 0.65 0.513Moist -0.27 0.787VM 1.11 0.269FC 0.78 0.435%Fe -1.1 0.274%FeO 0.9 0.369%SiO2 -3.23 0.001%Al2O3 2.69 0.007%CaO 2.44 0.015%MgO -2.87 0.004%Mn 3.74 0.000SiO2 -1.13 0.261(3) The regression equation for Sulpher toinput variables isS = - 0.88 + 0.000009 Blast Volume (M3/Min) -0.0490 Blast Pressure (Kg/cm2) + 0.0459 TopPressure (Kg/cm2) + 0.000013 Blast Temp (oC) -0.00116 Steam (t/hr) - 0.00924 % OxygenEnrichment + 0.000002 Oxygen (M3/hr.) + 0.0088Ash + 0.0102 Moist - 0.0075 VM + 0.0076 FC +0.00305 %Fe - 0.00090 %FeO + 0.00167 %SiO2 +0.0120 %Al2O3 - 0.00326 %CaO + 0.00840 %MgO+ 0.0575 %Mn - 0.0846 SiO2Table 3. T & P values for Sulpher.Predictor T pConstant -0.16 0.874Blast Volume 1.86 0.064Blast Pressure -1.51 0.132Top Pressure 1.37 0.173Blast Temp 0.49 0.626Steam -4.66 0.000Oxygen Enrichment -2.22 0.027Oxygen 2.15 0.032Ash 0.16 0.872Moist 0.98 0.328VM -0.14 0.892Predictor T pFC 0.14 0.89%Fe 0.61 0.539%FeO -0.81 0.418%SiO2 0.23 0.818%Al2O3 0.72 0.474%CaO -0.83 0.406%MgO 1.8 0.073%Mn 2.66 0.008SiO2 -0.89 0.377(4) The regression equation for Phosphorousto input variables isP = - 3.16 + 0.000023 Blast Volume (M3/Min) -0.0421 Blast Pressure (Kg/cm2) + 0.0115 TopPressure (Kg/cm2) - 0.000071 Blast Temp (oC) +0.000868 Steam (t/hr) - 0.00572 % OxygenEnrichment + 0.000000 Oxygen (M3/hr.) + 0.0323Ash - 0.0026 Moist + 0.0276 VM + 0.0358 FC -0.00458 %Fe - 0.00238 %FeO + 0.00923 %SiO2 -0.0139 %Al2O3 + 0.00114 %CaO - 0.0114 %MgO- 0.0404 %Mn + 0.182 SiO2Table 4. T & P values for Phosphorus.Predictor T pConstant -0.43 0.668Blast Volume 3.62 0.000Blast Pressure -0.97 0.332Top Pressure 0.26 0.798Blast Temp -1.93 0.055Steam 2.61 0.009Oxygen Enrichment -1.03 0.304Oxygen 0.14 0.887Ash 0.44 0.659Moist -0.19 0.853VM 0.37 0.708FC 0.49 0.626%Fe -0.69 0.491%FeO -1.61 0.109%SiO2 0.95 0.341%Al2O3 -0.62 0.536%CaO 0.22 0.828%MgO -1.83 0.067%Mn -1.4 0.163SiO2 1.42 0.155It is also necessary to examine thedependency between these variables and also findthe critical process variables (p value < 0.05) which
O. Rama MohanaRao, K.VenkataSubbaiah, K.NarayanaRao, T. SrinivasaRao / InternationalJournal of Engineering Research and Applications (IJERA) ISSN: 2248-9622 www.ijera.comVol. 3, Issue 3, May-Jun 2013, pp.635-641639 | P a g emay influence the quality of hot metal. The ‘p’ and‘T’ values from the Table 1 to Table 4 of theregression analysis are tabulated in Table 5 forwhich ‘p’ value is less than 0.05. It is predictedfrom the above, the critical process variables whichmay influence the quality of Hot metal in thisprocess are Blast Volume, Blast Pressure, Steam,Oxygen Enrichment, Oxygen, %FeO, %MgO,%Mn, SiO2 and %Al2O3.The T-statistic values for the above criticalprocess variables are also higher than the thresholdvalues (i.e. plus or minus 1.98 for 95 % confidencelevel) indicating their significance presence ofdependence which may influence the quality of Hotmetal.Table 5. Diagnosis of critical process variablesPredictor Variable T pBlast Volume 3.62 0.000Blast Pressure -3.34 0.001Steam (t/hr) 2.61 0.009Oxygen Enrichment -2.22 0.027Oxygen 2.15 0.032%FeO 2.55 0.011%MgO 2.44 0.015%Mn 2.66 0.008SiO2 -3.09 0.002%Al2O3 3.36 0.0015.2 Control limits and construction of ControlchartsThe data has been analyzed using chartwith customary plus/minus three sigma controllimits to identify the problematic observations. Theindividual Control charts for the critical processvariables are drawn and shown (Fig.2 to Fig. 11).Chart of Blast Volume is shown inFig.2 and the observations on the days of 9, 25 and45falls outside the control limits, indicating anunstable process. Test Results for Chart of BlastPressure is shown in Fig.3 and the observations onthe days of 9, 25, 45 and 46 falls outside the controllimits, indicating an unstable process. Test Resultsfor Chart of Steam is shown in Fig.4 and theobservations on the days of 6, 15, 25, 32, 35and 45falls outside the control limits, indicating anunstable process. Test Results for Chart of %Oxygen Enrichment is shown in Fig. 5 and theobservations on the days of 7, 35, 36, 37, 38, and45 falls outside the control limits, indicating anunstable process. Test Results for Chart ofOxygen is shown in Fig. 6 and the observations onthe days of 7, 35, 36, 37, 38, and 45 falls outsidethe control limits, indicating an unstable process.Test Results for Chart of % FeO is shown in Fig.7 and the observations on the days of 5, 13, 14, 17,18, and 22 falls outside the control limits,indicating an unstable process. Test Results forChart of %MgO is shown in Fig. and theobservations on the days of 10, 20, 21, 22, 29, 30,34, 35, 40, 42, and 45 falls outside the controllimits, indicating an unstable process. Test Resultsfor Chart of %Mn is shown in Fig. 9 and theobservations on the days of 31, 32, 33, 34, 35, 40,45 and 46 falls outside the control limits, indicatingan unstable process. Test Results for Chart ofSiO2is shown in Fig. 10 and the observations on thedays of 1, 12, 13, 14, 30, 33 and 46 falls outside thecontrol limits, indicating an unstable process. TestResults for Chart of %Al2O3is shown in Fig. 11and the observations on the days of 15, 19, 20, 40,42, 45 and 46 falls outside the control limits,indicating an unstable process.The input values for Blast Volume, BlastPressure, Steam, Oxygen Enrichment and Oxygenremain unchanged for the entire day and there is nodifference in sample range with in a day. Hence thesignificance of R- Chart does not exist for thesevariables. It is evident from the R-Chart drawn forthe above variables in the Fig. 2 to Fig. 6 that manyof the observations fall outside control limits,hence ignored.Whereas the input values of othervariables like %FeO, %MgO, %Mn, SiO2, and%Al2O3 may vary for each observation andcorresponding R- Charts were drawn and shown inthe Fig.7 to Fig.11 respectively. The out of controllimit points for MgO is on 46thday, for Mn is on10thday, for SiO2 is on 3rdand 30thday and for%Al2O3 is on 4th, 8th, 10th, 37thand 46thdays.4641363126211611616000500040003000DaysSampleMean__X=4691UCL=5122LCL=42604641363126211611612000150010005000DaysSampleRange_R=1156UCL=2154LCL=1571111111111111111111111111111111111111111111Figure 2. Xbar-R Chart of Blast Volume4641363126211611614.03.53.02.52.0DaysSampleMean__X=3.22UCL=3.502LCL=2.93846413631262116116220.127.116.11.40.0DaysSampleRange_R=0.757UCL=1.412LCL=0.10311111111111111111111111111111111111111111111Figure 3. Xbar-R Chart of Blast Pressure
O. Rama MohanaRao, K.VenkataSubbaiah, K.NarayanaRao, T. SrinivasaRao / InternationalJournal of Engineering Research and Applications (IJERA) ISSN: 2248-9622 www.ijera.comVol. 3, Issue 3, May-Jun 2013, pp.635-641640 | P a g e464136312621161161151050DaysSampleMean__X=6.44UCL=9.84LCL=3.0546413631262116116120151050DaysSampleRange_R=9.11UCL=16.98LCL=1.241111111111111111111111111111111111111111111111Figure 4. Xbar-R Chart of Steam4641363126211611613210-1DaysSampleMean__X=0.948UCL=1.850LCL=0.04746413631262116116143210DaysSampleRange_R=2.420UCL=4.510LCL=0.3291111111111111111111111111111111111111111111111111Figure 5. Xbar-R Chart of % Enrichment4641363126211611611000050000-5000DaysSampleMean__X=3550UCL=6944LCL=15646413631262116116120000150001000050000DaysSampleRange_R=9110UCL=16980LCL=124011111111111111111111111111111111111111111111111111Figure 6. Xbar-R Chart of Oxygen46413631262116116112.011.511.010.5DaysSampleMean__X=11.332UCL=11.720LCL=10.9444641363126211611612.01.51.00.50.0DaysSampleRange_R=1.042UCL=1.942LCL=0.142111111Figure 7. Xbar-R Chart of % FeO4641363126211611618.104.22.168DaysSampleMean__X=2.21UCL=2.3450LCL=2.07504641363126211611610.80.60.40.20.0DaysSampleRange_R=0.3624UCL=0.6755LCL=0.0493111111111111Figure 8. Xbar-R Chart of % MgO4641363126211611610.150.120.090.06DaysSampleMean__X=0.108UCL=0.1293LCL=0.08674641363126211611610.1000.0750.0500.0250.000DaysSampleRange_R=0.0572UCL=0.1067LCL=0.0078111111111Figure 9. Xbar-R Chart of % Mn4641363126211611616.86.46.05.65.2DaysSampleMean__X=5.99UCL=6.327LCL=5.6534641363126211611612.01.51.00.50.0DaysSampleRange_R=0.905UCL=1.687LCL=0.123111111111Figure 10. Xbar-R Chart of % SiO24641363126211611622.214.171.124.22.0DaysSampleMean__X=2.447UCL=2.6581LCL=2.23594641363126211611611.000.750.500.250.00DaysSampleRange_R=0.567UCL=1.056LCL=0.077111111111111111111Figure 11. Xbar-R Chart of % Al2O35.3 Results and discussionEven if the variation in input variableswere known but the exact reason was difficult toidentify due to complexities in Blast FurnaceProcess. Blast furnace slag composition has veryimportant behavior on its physicochemicalcharacteristics which affects the degree ofdesulphurization, smoothness of operation, cokeconsumption, hot metal productivity and its quality.Al2O3, MgO and CaO that entered withthe iron ore, pellets, sinter or coke Si with the cokeash and Sulphur enters through coke. In the normalpractice of blast furnace, slag is generallyaccounted for by adjusting the overall compositionof CaO, SiO2, Al2O3 and MgO components. Sincethe limestone (flux) is melted to become the slagwhich removes Sulphur and other impurities, theblast furnace operator may blend the differentgrades of flux to produce the desired slag chemistryand produce optimum hot metal quality. High toppressure in Blast Furnaces can decrease % of Si in
O. Rama MohanaRao, K.VenkataSubbaiah, K.NarayanaRao, T. SrinivasaRao / InternationalJournal of Engineering Research and Applications (IJERA) ISSN: 2248-9622 www.ijera.comVol. 3, Issue 3, May-Jun 2013, pp.635-641641 | P a g ehot metal. An increase in the Fe content of sintermay optimizes the Carbon/ Sulpher ratio anddecrease in Al2O3 content in hot metal. Manganesereaction is always accompanied by silica reaction.By adding additional SiO2 can reduce % Mncontent in hot metal. By implementing these stepsmay lead to reduce the defectives in the output andimproves the quality of hot metalConclusionThis paper explores monitoring ofvariables that effects hot metal making in anintegrated steel plant. In the first phase criticalprocess variables that affect the quality of hot metalare identified through regression analysis. From thestudy the variables namely, Blast Volume, BlastPressure, Steam,% Enrichment, Oxygen, %FeO,%MgO, %Mn, SiO2, and %Al2O3 are identified ascritical process variables. Subsequently and R-Charts are drawn to monitor these critical processvariables.When the more number of variables arecorrelated with each other, univariate control chartsare difficult to manage and analyze because of thelarge numbers of control charts of each processvariable. An alternative approach is to construct asingle multivariate T2control chart that minimizesthe occurrence of false process alarms. Hence thisstudy may be extended to multivariate controlcharts that monitor the relationship between thevariables and identifies real process changes whichare not detectable through univariate charts.References Juran, J. M.., Editor, & Gryna, F. M.,Associate Editor. (1998). Juran’s QualityControl Handbook, 4th ed, McGraw-HillBook Company, New York. Montgomery. D.C, and George C. Runger,(2003), Applied Statistics and Probabilityfor engineers, (John Wiley and sons). Samanta . B., A. Bhattacherjee (2004),Problem of Non-normality in StatisticalQuality Control- A case study in SurfaceMine, The Journal of the South AfricanInstitute of Minining and Metallurgy,257-264. Woodall, W. H. and Faltin, F.W., (1993),Auto correlated Data and SPC, ASQC,Statistics division News Letter, 13, 4, 18-21. Weller. E.F., (2000), Practical applicationsof Statistical Process Control, IEEESoftware, V.17, 3,48.55 Mohammed MA (2004), Using statisticalprocess control to improve the quality ofhealth care. QualSaf Health Care, 13,243–5. M A Mohammed,1 P Worthington,2 W HWoodall, (2008), Plotting basic controlcharts: tutorial notes for healthcarepractitioners, QualSaf HealthCare;17:137–145. Grigg, N.P.; Daly, J. & Stewart, M.(1998). Case study: the use of statisticalprocess control in fish product packaging.Food Control,Vol. 9, 289-297 Srikaeo, K., &Hourigan, J.A. (2002). Theuse of statistical process control (SPC) toenhance the validation of critical controlpoints (CCPs) in shell egg washing. FoodControl, 263-273. Rashed, H. I. (2005). A PerformanceAnalysis of Univeriate and MultivariateQuality Control Charts for OptimalProcess Control. M.Sc. Thesis, Faculty ofEngineering, Menoufia University, Egypt,Under Publication. "Dr.AbasMukbelbaarz ,AllaTalalYassin ,SahabDia Mohamed,(2012),"" Assessingthe quality of Product using statisticalquality control maps: a case study"",Diyala Journal for pure sciences.Vol: 8No: 1, January 2012 ISSN: 2222-8373 M. A. Sharaf El-Din1, H. I. Rashed2 andM. M. El-Khabeery (2006), StatisticalProcess Control Charts Applied toSteelmaking Quality Improvement,Quality Technology & QuantitativeManagement, 3, 4, 473-491.