Case Study (wileyonlinelibrary.com) DOI: 10.1002/qre.1212 Published online in Wiley Online Library Application of Six Sigma Methodology to Reduce Defects of a Grinding Process E. V. Gijoa, Johny Scariab and Jiju Antonyc∗† Six Sigmaisadata-drivenleadershipapproachusingspecifictoolsandmethodologiesthatleadtofact-baseddecision making. This paper deals with the application of the Six Sigma methodology in reducing defects in a fine grinding process of an automotive company in India. The DMAIC (Define–Measure–Analyse–Improve–Control) approach has been followed here to solve the underlying problem of reducing process variation and improving the process yield. This paper explores how a manufacturing process can use a systematic methodology to move towards world-class quality level. The application of the Six Sigma methodology resulted in reduction of defects in the fine grinding process from 16.6 to 1.19%. The DMAIC methodology has had a significant financial impact on the profitability of the company in terms of reduction in scrap cost, man-hour saving on rework and increased output. A saving of approximately US$2.4 million per annum was reported from this project. Copyright © 2011 John Wiley & Sons, Ltd. Keywords: Six Sigma; Kappa statistic; process capability evaluation; chi-square test; ANOVA; Taguchi methods 1. Introduction S ix Sigma is a well-structured methodology that focuses on reducing variation, measuring defects and improving the quality of products, processes and services. Six Sigma methodology was originally developed by Motorola in 1980s and it targeted adifficult goalof3.4partspermillion defects1.SixSigmahas beenonanincredible runover25years,producing significant savings to the bottom line of many large and small organizations2. Leading organizations with a track record in quality have adopted Six Sigma and claimed that it has transformed their organization3. Six Sigma was initially introduced in manufacturing processes; today, however, marketing, purchasing, billing, invoicing, insurance, human resource and customer call answering functions are also implementing the Six Sigma methodology with the aim of continuously reducing defects throughout the organization’s processes4. AccordingtoHarryandSchroeder5,SixSigmaisapowerfulbreakthroughbusinessimprovementstrategythatenablescompanies to use simple and powerful statistical methods for achieving and sustaining operational excellence. It is a business strategy that allows companies todrastically improve their performance bydesigning andmonitoring everyday businessactivities inwaysthat minimizewasteandresourceswhileincreasing customersatisfaction6.TheSixSigmaapproachstartswithabusinessstrategyand endswithtop-downimplementation,havingasignificantimpactonprofit,ifsuccessfullydeployed3.Numerousbooksandarticles provide the basic concepts and benefits of the Six Sigma methodology. These publications cover topics, such as What is Six Sigma3?Whydoweneed SixSigma7?Six Sigmadeployment8; critical success factors ofthe SixSigma implementation4;Hurdles in the Six Sigma implementation9; the Six Sigma project selection10 and organizational infrastructure required for implementing SixSigma11.Numerousarticlesareavailableindifferent aspectsofSixSigmaoverthepast10years12--17.TheSixSigmaapproach has been widely used to improve performances and reduce costs for several industrial fields18--22. Thispaperpresentsthestep-by-stepapplicationoftheSixSigmaDMAIC(Define–Measure–Analyse–Improve–control) approach to eliminate the defects in a fine grinding process of an automotive company. This has helped to reduce defects in the process andthereby improve productivity andon timedelivery tocustomer. During the measure andanalyse phases of theproject, data were collected fromtheprocesses tounderstand the baseline performance and forvalidation ofcauses. Thesedatawere studied through various graphical and statistical analyses. Chi-square test, ANOVA23, Design of Experiments (DOE) 24, Control Charts25, Taguchi methods26, etc. were used to make meaningful and scientifically proven conclusions about the process and the related causes. aSQC & OR Unit, Indian Statistical Institute, 8th Mile, Mysore Road,Bangalore 560 059, India bDepartment of Statistics, Nirmala College, Muvattupuzha 686 661, India cDepartment of DMEM, University of Strathclyde, Glasgow, Scotland G1 1XJ, U.K. ∗Correspondence to: Jiju Antony, Department of DMEM, University of Strathclyde, Glasgow, Scotland G1 1XJ, U.K. †E-mail: [email protected] Copyright © 2011 John Wiley & Sons, Ltd. Qual.Reliab.Engng.Int.2011 E. V. GIJO, J. SCARIA AND J. ANTONY The structure ofthis article is as follows. The research methodology adoptedforthis studyis explained inSection 2. Section 3 explains an introduction to the case study, Section 3.1 indicates the define phase, Section 3.2 details the measure phase with baseline performance. The Analyse phase is explained in Section 3.3 with details of potential causes and its validation followed by the Improvement phase in Section 3.4 with details of solutions implemented. Section 3.5 explains the controls introduced to ensuresustainabilityoftheresults.Section4providesinformationaboutthelessonslearnedfollowedbySection5,themanagerial implications of the initiative. Section 6 presents the concluding remarks and discusses the benefits and limitations of the study. 2. Research methodology Thissectionexplainsthemethodologyadoptedforthiscasestudy.Scientificinvestigationoninnovatingasystemorimprovement totheexistingoneneedstobeginwithsomestructureandplan.Thisstructureandplanofinvestigationwereconceivedsoasto obtain answers to research questions in the research design27. The researcher worked with the company to provide support for theprojectintheSixSigmatechniques, whilstrecording dataabouttheexercise fromwhichtodevelop acasestudy.Aliterature review wasundertaken with an objective of identifying the pasthistory of various improvement initiatives carried out toaddress process-related problems. The methodology is divided into four major sections namely problem definition, literature survey, case study design and data analysis. Based on the available data on the process, the team studied the baseline status of the process and drafted a project charter, which explains the details of the problem. A detailed literature review was undertaken in Six Sigma with an objective of identifyingthetypeofimprovementscarriedoutbydifferentpeopleinvariousorganizationstoaddressprocess-relatedproblems. A case study entails the detailed and intensive analysis of a single case—a single organization, a single location or a single event28. Yin29 describes a case study as an empirical inquiry that investigates a contemporary phenomenon within its real- life context. According to Lee30, the unit of analysis in a case study is the phenomenon under study and deciding this unit appropriately is central to a research study. In this paper, a case study is designed to study the underlying process problem so that solutions can be implemented for process improvement. The collected data were analysed using descriptive and inferential statistics. Measurement system analysis, chi-square test, ANOVA, DOE with Taguchi methods, etc. were used for analysing the data and inferences were made. Graphical analyses, such as histogram and control chart, were also utilized for summarizing the data and making meaningful conclusions. Minitab statistical software was used to analyse the data collected at different stages in the case study. 3. Case study This casestudydeals withthereduction ofdefects inthefinegrinding processinanautomobilepartmanufacturing companyin India. The company with manpower of approximately 2550 people is manufacturing common rail direct injection (CRDI) system pumps for vehicles. These pumps were used in cars, trucks and buses throughout the world. An injector primarily consists of nozzle and nozzle holder body. A schematic view of fuel injector is given in Figure 1. The components used in fuel injector and Nozzle holder body Spring Distance piece Cap nut Nozzle Distance piece Figure 1. Schematic view of fuel injector Copyright © 2011 John Wiley & Sons, Ltd. Qual.Reliab.Engng.Int.2011 E. V. GIJO, J. SCARIA AND J. ANTONY their functions are as follows. Distance piece aligns the high-pressure fuel lines of nozzle holder body and nozzle. Its both sides are fine ground precisely to ensure sealing of the high-pressure fuel coming from holder body to nozzle. Cap nut retains the nozzle and distance piece with the holder body with sufficient torque to ensure sealing. Spring and pressure bolt ensures the functioning of injector with set opening pressure and timely delivery of fuel. The current project was undertaken in the distance piece fine grinding process, which is done by fine grinding machine. Different types ofdistance pieces were fine ground in this machine. This is asophisticated and very expensiveCNC fine grinding machine. It finishes both faces of distance pieces in batches precisely with sub-micron flatness values. After fine grinding, distance pieces were inspected visually to find various defects. Since the production of distance pieces were in thousands per shift, it was not practically possible to do 100% inspection of these components by objective methods. Hence visual inspection was carried out for all the components with reference to master pieces and visual limit samples. Since therejection level ofdistance piecesafterfinegrinding processwasveryhighandthefunction ofthecomponent intheproduct was highly critical, it was essential to do 100% inspection. Under these circumstances, the project was of highest priority to the management as it was clear that an effective solution to this problem would have a significant impact in reducing rework/ rejection and improving productivity. Also, it was clear to the team members and champion of the project that the elimination of this problem will help the organization to cater to the increasing demand of market. In the past, many attempts were made to solve this problem by using different methodologies, which were unsuccessful. The Six Sigma problem solving methodology (DMAIC) was recommended when the cause of the problem is unclear3. Hence, it was decided to address this problem through the Six Sigma DMAIC methodology. 3.1. Define phase This phase of the DMAIC methodology aims to define the scope and goals of the improvement project in terms of customer requirements and to develop a process that delivers these requirements. The first step towards solving any problem in the Six Sigma methodology is by formulating a team of people associated with the process. The team selected for this project includes theSeniorManager—Manufacturing astheBlackBelt(BB). TheothermembersoftheteamwerePlanning Manager,Maintenance Manager, Quality Control Senior Engineer and one Machine Operator. BB acts as the team leader, and was responsible for the overall success of the project. In this particular project, BB himself was the process owner. The primary responsibility of team members was to support BB in executing the project-related actions. The Head of manufacturing department was identified as the Champion and the Head of Business Excellence department as master black belt (MBB) for this project. The team along with the Champion and MBB developed a project charter (Appendix 1) with all necessary details of the project. This has helped the team members to clearly understand the project objective, project duration, resources available, roles and responsibilities of team members, project scope and boundaries, expected results from the project, etc. This creates a common vision and sense of ownership for the project, so that the entire team is focused on the objectives of the project. The team had several meetings with the Champion and MBB todiscuss various aspectsof the problem, including the internal and customer-related issues arising because of this problem. The team decided to consider the rejection percentage of distance pieces after fine grinding process as the Critical to Quality (CTQ) characteristic for this project. The goal statement was defined as the reduction in rejection of distance pieces by 50% from the existing level, which should result in large cost saving for the company in terms of reduction in rework and scrap cost. Since there was a cross-functional team for executing this project, the team felt that it was necessary to perform a SIPOC (Supplier–Input–Process–Output–Customer) analysis to have a better understanding of the process. This is a method similar to process mapping for defining and understanding process steps, process inputs and process outputs3. The team with the involvement of people working with the process prepared a SIPOC mapping along with a basic flowchart of the process. This SIPOC has given a clear understanding of the process steps needed to create the output of the process. Through this exercise, the team got the clarity of the project in terms of the scope of the project, inputs, outputs, suppliers and customers of the process.The teamfocusedonthe finegrinding process forimprovement thatis defined as thescopeoftheproject. The process mapping along with SIPOC is presented in Appendix 2. 3.2. Measure phase The objective of the measure phase is to understand and establish the baseline performance of the process in terms of process capabilityorsigmarating.TheCTQconsidered inthiscasewastherejection percentageofdistancepieces afterthefinegrinding process.Theserejections weremainlyduetotheoccurrence ofdifferenttypesofdefects,suchasburr,shades,deeplines,patches and damage, on the component after machining. The schematic representation of these defects is presented in Figure 2. These defects create an uneven surface in thecomponent that could lead tofuel leakages in pumps.Aftermachining, the components were visually inspected for these defects. Master samples were provided for identifying each of these defects and inspectors did the inspection. Since there was no instrument involved in the inspection process and only visual inspection was performed, before going ahead with further data collection, the team decided to carry out Attribute Gage Repeatability and Reproducibility (GageR&R)studytovalidatethemeasurement system.Insuchstudies,intra-inspectoragreement measuresrepeatability(within inspector), inter inspector agreement measures the combination of repeatability and reproducibility (between inspectors)31. The non-chance agreement between the two inspectors, denoted by Kappa, defines as Number of observed agreements−Number of expected agreements (cid:2)= . Total number of observations−Number of expected agreements Copyright © 2011 John Wiley & Sons, Ltd. Qual.Reliab.Engng.Int.2011 E. V. GIJO, J. SCARIA AND J. ANTONY Patches Burr Shade Damage Deep lines Figure 2. Schematic representation of defects Table I. Data collection plan Characteristic Data type How measured Sampling notes Related conditions Rejection percentage of Attribute Visual checking by comparing 100% of units in all the Shift wise and defect distance pieces after the with visual limit samples three shiftsfortwomonths type wise fine grinding process For conducting the study, 100 components were selected and they were classified as good or bad independently by two inspectors. From the resulting data, the Kappa value was calculated and was found to be 0.814 with a standard error of 0.0839. Since the Kappa value was more than 0.6, the measurement system was acceptable31. After the measurement system study, a data collection plan was prepared with details of types of data, stratification factors, sampling frequency, method of measurement, etc. for the datatobecollected during the measure phase of this study. The data collection plan thus prepared is presented in Table I. The data were collected as per the plan to understand the baseline status of the process. During the defined period of data collection, 368219 components were inspected and 61198 components were rejected due tovarious defects. Each one of therejected components washaving one ormore defects. The detailed dataon the typeof defects were collected and the same was graphically presented as apareto diagram (Figure 3). The collected data shows thattherejection intheprocesswas166200PPM.Thecorresponding sigmaratingoftheprocesscanbeapproximatedto2.47. For any improvement initiative in this organization, the general goal set by the management was to reduce the rejection by 50% from the existing level. Based on this policy, the target set for the study was to reduce the rejections at the fine grinding process to 83100PPM from the existing level of 166200PPM. 3.3. Analyse phase After mapping the process, the team proceeded to analyse the potential causes of defects. A cause and effect diagram was prepared afterconducting abrain storming sessionwith all theconcerned peoplefrom theprocess along with theproject team, Champion and MBB. The output of the cause and effect diagram depends on a large extent on the quality and creativity of the brain storming session32. Figure 4 illustrates the cause and effect analysis prepared during the brain storming session. Thenext step inthis phasewas togather datafrom theprocess in order toobtainabetterpicture of thepotential causes, so thattherootcause/scanbeidentified. Theteamhaddetaileddiscussionwiththeprocesspersonneltoidentifythepossibledata that can becollected on the potential causes in the cause and effect diagram. Aftergetting the detailed picture of availability of data on causes, the team discussed with MBB to identify the type of analysis possible on these causes. Based on this discussion, a cause validation plan was prepared to detail the type of data to be collected and the type of analysis possible for each of thesecauses.Thepotentialcauses,suchas‘variationininputparts’,‘suppliermaterialvariation’and‘programparametersnotOK’, can be validated by statistical analysis on the data collected from the process. But potential causes, such as ‘improper cleaning after dressing’ and ‘repair batches mix up’, have to be validated only by observing the process (gemba). Hence for few causes, detailed data were collected and statistical analyses were planned, and for the remaining causes gemba was planned to validate the causes. Table II summarizes the potential causes and the type of analysis planned for each cause. Copyright © 2011 John Wiley & Sons, Ltd. Qual.Reliab.Engng.Int.2011 E. V. GIJO, J. SCARIA AND J. ANTONY 200000 100 150000 80 ount 100000 60 rcent C e P 40 50000 20 0 0 Defect Shades Deep lines Patches Others Count 76539 60267 43995 318 Percent 42.3 33.3 24.3 0.2 Cum % 42.3 75.5 99.8 100.0 Figure 3. Pareto diagram for visual defects Measurements Material Man Variation in size of input parts Improper cleaning after Product family to family dressing Inspector to inspector variation variation Presence of sand blasting dust Repair batches mix up Supplier to supplier High variation Rejection in Grinding Process Wheel straightness not OK Loading/unloading system not OK Program parameters not Optimum Improper setting Material removal rate not OK Method Machine Figure 4. Cause and effect diagram for rejection in grinding process As three different suppliers provided the raw material, it was suspected that there was possibility of supplier-to-supplier variation with respect to the thickness of input raw material. To study this variation, the data on the thickness of components from all the three suppliers were collected and ANOVA (A statistical procedure used to determine the significant effect of a variable under study23.) was performed in the data and p-value was observed. P-value is a means for judging the significance of astatisticaltest.Thesmallerthep-value,themoresignificant theresultsare.Typicallyvaluesbelow0.05areconsidered indicative of a significant test outcome25. In this case, p-value was found to be 0.407, ruling out the possibility of significant difference between the suppliers23. Further to validate the potential cause of variation in size of input parts, a batch of 57 distance pieces were selected and thickness measured. The data were subjected to Anderson Darling Normality test, and found to follow normal distribution23. Process capability study was carried out on this data and found to be capable, confirming that the thickness variation in input part wasnot aroot cause33.The Process capability studyis acomparison of the process outputwith customer requirements to determine whether a process is capable of meeting customer expectations25. The Minitab statistical software output of process capability evaluation is presented in Figure 5. Copyright © 2011 John Wiley & Sons, Ltd. Qual.Reliab.Engng.Int.2011 E. V. GIJO, J. SCARIA AND J. ANTONY Table II. Cause validation plan Sl. no Cause Plan for validation 1 Improper cleaning after dressing GEMBA 2 Repair batches mix up GEMBA 3 Variation in size of input parts Process capability analysis 4 Product family to family variation Chi-square test 5 Presence of sand blasting dust GEMBA 6 Supplier to supplier variation ANOVA 7 Material removal rate not OK ANOVA 8 Process parameters not Optimum Design of experiments (DOE) 9 Improper setting DOE 10 Loading/unloading system not OK GEMBA 11 Wheel straightness not OK GEMBA 12 Inspector to inspector variation Gauge R & R Process Capability Analysis for Thickness LSL USL Process Data USL 3.16000 Target * LSL 3.11000 Mean 3.13937 Sample N 57 StDev (Overall) 0.0045027 Overall Capability Pp 1.85 PPU 1.53 PPL 2.17 Ppk 1.53 3.11 3.12 3.13 3.14 3.15 3.16 Cpm * Observed Performance Expected Performance PPM < LSL 0.00 PPM < LSL 0.00 PPM > USL 0.00 PPM > USL 2.30 PPM Total 0.00 PPM Total 2.30 Figure 5. Process capability analysis Sincedifferent families ofproductswereproduced, datawere collected formaterial removalrate(MRR)aswellasdefects with respect to various families of components to test their significance. Data on MRR were collected on three types of components to study the effect of type of component on MRR. Based on the quantity of material removed from the component during machining, the MRR value was calculated by an inbuilt software program in the machine. These MRR data were recorded in millimeter/minute. ANOVAwasperformedonthisdataandp-valuewasfoundtobe0.085,notshowingsignificanceat5%level23. To test whether product family-to-family variation affects the defects, a chi-square test was carried out between defect type and family of components23. For each of the defect types, viz., patches, shades and deep lines, separate chi-square test was done with three different families of components. The details of chi-square test are given in Table III. From Table III, it was clear that except for shades, family-to-family variation does not affect visual defects. The machining program and machine parameters for each family and type were different. The team thought, it was better to have a uniform machining program and parameters for all the family components so that the process can be better managed. Hence for validating the process parameters and identifying theoptimumoperatingconditions,theteamdecidedtoconductaDOEduringtheimprovephase.DOEisatechnique for understanding variability, in which factors are systematically and simultaneously manipulated while the variability in outputs (responses) is studied to determine which factors have the biggest impact24. The other causes listed in the cause and effect diagram were validated by gemba analysis. Some of the details are presented in Table IV. The details of validation of all causes in the cause and effect diagram are summarized in a tabular format and is given in Table V. Copyright © 2011 John Wiley & Sons, Ltd. Qual.Reliab.Engng.Int.2011 E. V. GIJO, J. SCARIA AND J. ANTONY Table III. Test statistic values for Chi-square test Defect type Chi-square statistic Degrees of freedom p-value Patches 2.509 2 0.285 Deep lines 2.398 2 0.301 Shades 452.256 2 0.000 Table IV. Gemba observations Sl. no. Cause Observation/conclusion 1 Improper cleaning after dressing It was observed that cleaning done after dressing. Not a root cause. 2 Repair batches mix up Repair batches were found mixed with other batches during the random visit to the process. Root cause. 3 Presence of sand blasting dust Traces of shot blasting dust found in the input batch during inspection. Root cause. 4 Loading/unloading system not OK Loading table wear out observed at the edges. Root cause. 5 Wheel straightness not OK Wheel straightness found OK. Checking frequency followed as per procedure. Not a root cause. Table V. Summary of validation of causes Sl. no. Cause Tools used for validation Results 1 Improper cleaning after dressing GEMBA Not a root cause 2 Repair batches mix up GEMBA Root cause 3 Variation in size of input parts Process capability analysis Not a root cause 4 Product family to family variation Chi-square test Root cause 5 Presence of sand blasting dust GEMBA Root cause 6 Supplier to supplier variation ANOVA Not a root cause 7 Material removal rate not OK ANOVA Root cause 8 Process parameters not optimum DOE Root cause 9 Improper setting DOE Root cause 10 Loading/unloading system not OK GEMBA Root cause 11 Wheel straightness not OK GEMBA Not a root cause 12 Inspector to inspector variation Gauge R & R Not a root cause 3.4. Improve phase This phase of the Six Sigma project is aimed at identifying solutions for all the root causes identified during the Analyse phase, implementing them after studying the risk involved in implementation and observing the results. At this stage, as decided earlier, a DOE was planned for optimizing the process/machine parameters. The team along with champion, MBB, theproduction supervisorandoperators oftheprocessconducted aseries of brainstorming sessionstoidentify the important parameters for experimentation. The parameters selected through these discussions were load applied, initial load setting, coolant flow rate, upper wheel rpm, lower wheel rpm and cage rpm. Since the relationship between these parameters and MRRwasnotknown,itwasdecidedtoexperimentalltheseparametersatthreelevels26.Theexistingoperatinglevelwasselected asonelevelforexperimentation.Theteambasedonvariousoperationalfeasibilitiesselectedtheothertwolevels.Theparameters and levels selected for experimentation are presented in Table VI.Also, the team felt there is a possibilityof interaction between load applied with upper wheel rpm, load applied with lower wheel rpm and load applied with cage rpm. Hence it was decided to estimatetheeffectofthesethreeinteractionsalso.Sixparametersatthreelevelsandthreeinteractions withreplicationsrequirea hugenumber ofcomponents forconducting afullfactorial experiment, which wouldbeacostlyandtime-consuming exercise24. It was possible to estimate the effect of these selected parameters and interactions using the 27 experiments with the help of OrthogonalArray(OA).Hence forconducting anexperiment withsixparametersandthreeinteractions, L27(313)orthogonalarray was selected34. As the name suggests, the columns of this array are mutually orthogonal. Also, experiments using orthogonal arrays play a crucial role in achieving additivity of the model effects34. The design layout prepared as per L27(313) orthogonal array is given in Table VII. The response of the experiment was decided as material removal rate (MRR). As per the design layout given in Table VII, the experiments were conducted after randomizing the sequence of experiments, and the data were collected32. The experimental data were analysed by Taguchi’s Signal-to-Noise (S/N) ratio method35. The S/N ratio is advocated in the Taguchi method to maximize the performance of a system or product by minimizing the effect of noise36.TheS/Nratiocanbetreatedasaresponse(output)oftheexperiment,whichisameasureofvariationwhenuncontrolled Copyright © 2011 John Wiley & Sons, Ltd. Qual.Reliab.Engng.Int.2011 E. V. GIJO, J. SCARIA AND J. ANTONY Table VI. Process parameters and their levels Sl. no. Factor Levels ∗ 1 Load applied 170 200 230 ∗ 2 Initial load setting Minimum Medium Maximum ∗ 3 Coolant flow rate 8 LPM 12 LPM 16 LPM ∗ 4 Upper wheel RPM 70 90 110 ∗ 5 Lower wheel RPM 50 60 70 ∗ 6 Cage RPM 20 30 40 ∗ Existing levels. Table VII. Design layout for experimentation Exp. no. Load Setting Coolant UW RPM LW RPM C RPM 1 170 Min. 8 LPM 70 50 20 2 170 Min. 8 LPM 90 60 30 3 170 Min. 8 LPM 110 70 40 4 170 Med. 12 LPM 70 60 40 5 170 Med. 12 LPM 90 70 20 6 170 Med. 12 LPM 110 50 30 7 170 Max. 16 LPM 70 70 30 8 170 Max. 16 LPM 90 50 40 9 170 Max. 16 LPM 110 60 20 10 200 Med. 16 LPM 70 50 20 11 200 Med. 16 LPM 90 60 30 12 200 Med. 16 LPM 110 70 40 13 200 Max. 8 LPM 70 60 40 14 200 Max. 8 LPM 90 70 20 15 200 Max. 8 LPM 110 50 30 16 200 Min. 12 LPM 70 70 30 17 200 Min. 12 LPM 90 50 40 18 200 Min. 12 LPM 110 60 20 19 230 Max. 12 LPM 70 50 20 20 230 Max. 12 LPM 90 60 30 21 230 Max. 12 LPM 110 70 40 22 230 Min. 16 LPM 70 60 40 23 230 Min. 16 LPM 90 70 20 24 230 Min. 16 LPM 110 50 30 25 230 Med. 8 LPM 70 70 30 26 230 Med. 8 LPM 90 50 40 27 230 Med. 8 LPM 110 60 20 noise factors are present in the system34. Since the requirement of this process was to remove materials in a uniform rate from distance pieces to achieve specified dimension, the S/N ratio of nominal-the-best type was selected for analysis36. The S/N ratio for nominal the best type characteristic was defined as 10∗log((Y¯2)/s2), where Y¯ is the average and s, the standard deviation for each experiment26. The S/N ratio values were estimated for all the 27 experiments and ANOVA was performed on these S/N values to identify the significant parameters and interactions. From the ANOVA table (Table VIII), it was clear that the interaction effect between load applied and upper wheel rpm was significant at 5% level of significance. Also, parameters setting and lower wheel rpm were significant at 10% level of significance. The main effect (the change in average response produced by a change in the level of the factor24) and interaction (a measure of the degree to which the effect on the response of one factor is dependent upon the settings of one or more other factors24) plots of the S/N ratio values were made with the help of Minitab statistical software and are presented in Figures 6 and 7. From these plots, the best levels for parameters were identified as the level corresponding to highest value of S/N ratio35. Thus, the best levels for load applied and upper wheel rpm were selected from the interaction plot and the best levels for the other parameters were selected from the main effect plot. The optimum combination for process parameters thus arrived at is given in Table IX. One important point to be noted here is that for the factor load applied, the best level from the main effect plot was 230daN and that from the interaction plot was 170daN. This shows the importance of estimating interaction effect during the DOE study. Copyright © 2011 John Wiley & Sons, Ltd. Qual.Reliab.Engng.Int.2011 E. V. GIJO, J. SCARIA AND J. ANTONY Table VIII. ANOVA table for S/N ratio Source DF Seq. SS Adj. SS Adj. MS F p-value Load 2 0.0031630 0.0031630 0.0015815 07.00 0.125 Setting 2 0.0073185 0.0073185 0.0036593 16.20 0.058 Coolant 2 0.0020963 0.0020963 0.0010481 04.64 0.177 UW RPM 2 0.0005852 0.0005852 0.0002926 01.30 0.436 LW RPM 2 0.0050296 0.0050296 0.0025148 11.13 0.082 C RPM 2 0.0018296 0.0018296 0.0009148 04.05 0.198 Load*UW RPM 4 0.0198815 0.0198815 0.0049704 22.00 0.044 Load*LW RPM 4 0.0057037 0.0057037 0.0014259 06.31 0.141 Load*C RPM 4 0.0062370 0.0062370 0.0015593 06.90 0.131 Error 2 0.0004519 0.0004519 0.0002259 Total 26 0.0522963 Load Setting Coolant 7.8 7.6 7.4 7.2 s o ti a 7.0 r N S 170 200 230 Min. Med. Max. 8LPM 12LPM 16LPM f o n UW RPM LW RPM C RPM a 7.8 e M 7.6 7.4 7.2 7.0 70 90 110 50 60 70 20 30 40 Figure 6. Main effects plot (data means) for S/N ratios TheseoptimumlevelsinTableIXweretakenassolutionsforthecausesrelatedtoprocessparameters.Finally,thelistofselected solutions is presented in Table X. A risk analysis was carried out to identify any possible negative side effects of the solutions during implementation. The team concluded from the risk analysis that there were no significant negative impacts associated with any of the selected solutions. Hence, an implementation plan was prepared for the above solutions with responsibility and target date for completion for each solution. A time frame of two weeks was provided for implementing these solutions. All the solutionswereimplemented aspertheplanandtheresultswereobserved.Agraphicalpresentationofthecomparisonofresults before and after the project is provided in Figure 8. 3.5. Control phase The real challenge of the Six Sigma implementation is the sustainability of the achieved results. Due to variety of reasons, such as people changing the job, promotion/ transfer of persons working on the process, changing focus of the individual to other process-relatedissueselsewhere intheorganization andlackofownership ofnew peopleintheprocess,quiteoftenmaintaining Copyright © 2011 John Wiley & Sons, Ltd. Qual.Reliab.Engng.Int.2011 E. V. GIJO, J. SCARIA AND J. ANTONY 70 90 110 20 30 40 8.0 Load 170 200 Load 7.2 230 6.4 8.0 UW RPM 70 7.2 UW RPM 90 110 6.4 8.0 LW RPM 50 LW RPM 7.2 60 70 6.4 8.0 C RPM 20 30 7.2 C RPM 40 6.4 170 200 230 50 60 70 Figure 7. Interaction plot (data means) for S/N ratios Table IX. Optimum combination for process parameters Sl. no. Factor Optimum level 1 Load applied 170 2 Initial load setting Medium 3 Coolant flow rate 12 4 Upper wheel RPM 90 5 Lower wheel RPM 60 6 Cage RPM 30 Table X. Cause–Solution matrix Sl. no. Cause Solution 1 Repair batches mix up New storage system for repair parts introduced in the process 2 Product family to family variation Process parameters were optimized as per result of DOE 3 Presence of sand blasting dust Cleaning method after sand blasting introduced 4 Material removal rate not OK Reference table prepared for adjusting load 5 Process parameters not OK Optimum factor level combination from DOE 6 Improper setting Optimum factor level combination from DOE 7 Loading/unloading system not OK Conditioning of grinding wheel-loading table is done. theresultsareextremelydifficult9.Sustainabilityoftheresultsrequiresstandardizationoftheimprovedmethodsandintroduction of monitoring mechanisms forthe keyresults achieved. It alsorequires bringing awareness among thepersonnel performing the activities. Standardization of the solutions was ensured by affecting necessary changes in the process procedures that was a part of the quality management system of the organization. The quality plans and control plans were revised as per the solutions Copyright © 2011 John Wiley & Sons, Ltd. Qual.Reliab.Engng.Int.2011