Hemant Urdhwareshe Director, Instinwte of Quality and Reliability, Pune (India) py As we convert a practical prob- Jem in to a statistical model using & process map. Ie takes form of a mathematical equation: Y=f0C1,X23.... Xa) Where ¥ isthe key process output (KPOV) and Xs are the key process inputs (KPIV) Six Sigma We then evaluate the process data ro build (model) rela tionship of KPOV and KPIVs. This is easier said than done. It isin this context that Six Sigma implementation requires application of many advanced statistical tools. These tools involve complex calculations, Such calculations need not be done manually in twenty-first century. There are good softwares that can make life of black and green belts much easier. Six Sigma is a data based approach for taking deck sions to improve processes. Belts frequently need to analyze large amount of data in different ways so that it can pro- vide strong clues to the solutions. With ISO 9000, QS- 9000 and ISO/TS 16949, many companies do a good job of collecting data, This is an impgrtent step for making improvement. However, in absence Of good understand- ing of tools to analyze data, it simply is an addition of cost, for the company. It is extremely important for companies to train their key people in analyzing data and interpret la cee [— Ae wait Taupoing the results of analysis. Some of the tools that help analysis of data are: + Pareto Analysis for prioritizing improvement areas, projects + Statistical Process Control (SPC) for assessing stabil- ity of processes and identifying presence of special or assignable causes + Measurement Systems Analysis (MSA) to assess whether it is adequate for measuring process output + Graphical Analysis of data using various toots such as: © Box Plots and Dot Plots o Histograms o Scatter Plots Stem and Leaf Plots Pie Charts ‘o Multi Vari Charts + Correlation and Regression Analysis + Multiple regression ‘+ Hypothesis Tests such as test, F test, Chi-Square Test ‘Multi-Vari Chart for Strength by Time - Current ‘+ Analysis of Variance (ANOVA) + Planning and analyzing experiments for screening and characterization + Response Surface Methods to optimize process + Advanced control charts such as Cusum and Exponentially Weighted Moving Average (EWMA) Charts ‘This lis is not all inclusive but these are the most com= monly used tools in Six Sigma Projects. Often, process data is not normally distributed making the analysis more complex. Moreover, when data is analyzed by belts, it may not give the desired result in the first attempt. In case project requires analysis of field failures, it needs to be analyzed using different distributions such as Weibull, and Lognormal. We now realize why softwares are nec- essary! In absence of softwares, belts will retire analyzing the data! Some examples of data analysis are show here to illus- trate how simple graphical tools can be very useful in providing a clue to solve a problem or improve a process. ‘Figure 1 shows multi-vari chart of how strength of welding 10 20 [ "3000 Panel variable: Current 4500, Tae depends upon three factors Cur- 280 2 "| rent, Fore and time. The chart shows variation in strength with ~ all combinations 2407 ‘Figure 2 shows Scatter-plotof marks i ai ‘obtained by students and study time. ‘As we can see, ‘We can clearly see a relationship 2x0. Stengh renses becween the two. / ‘tiger erent 180. higher forceand: ‘Figure 3 shows box plot of how | oe higher time paint gloss (Shine of the paint) is cater) a affected by paint source, gun type ecm and distance of paint ose Seater ot ars ve Stine Soxplt le va Pit sou, an yp, Distance eo} ban ore Sue pave shoe? “B : : bo a| @ 7 te 4 ll a a ao" = i fe 2 ae rE 2 0 ome: 50D no © ® 7 ome a 2 1. 2 Sree sity vane 1 sey QaaltyWorld | August 2006Figure 4 shows matrix ia Figure 4 scatter plot of units con , Temperature, Production, Fest Hours, Days sumed in a factory and factors such es tempers- ture, working days, pro- duction. We are not able to see strong relation- ships between individual Matrix Plot of Us E | 8 f § factors. However, Multi ple regression can show ah ycte how they are related po, lieu Figure 5 shows a dot plot pee generated in SigmaXL. for ftp 4° distortion in plastic ec itg a Hle moulds as a function of 7 : injection pressure _ "| ae support analysis needs of even avery highly complex nature, It is therefore an obvious af oe choice over SigmaX1. if company is launch- neg att + ++ | ing a massive companywide Six Sigma Pro- ] “os a0 10 150 ‘gram with a pool of fall time black belts | Distortion - A Pressure: + and green belts. However, there are many companies, who launch Six Sigma program to assess its value. There are other compa- nies who may not need many of the ad- vanced tools sch as response surface meth- ‘ods (RSM). This could be true for service companies. Why would a company prefer lower end software tothe best in class? Fre- ‘quently reason is cost. Cost of a Minitab ‘There are many good types of software for analysisof such Sinsle wer license (about USS1295) may be about six times data. But choosing the right software is important while q ‘A of @ SigmaXT license (about USS199). With the re- company launches Six Sigma Program, Fequenty,choice C8 Version 5 of SigmaXL, its capabilities are enhanced ‘of software depends upon recommendation ofthe consult- #4 users need to carefully evaluate their choice, ant, Most commonly used softwares are: o Minitab Statgraphics, 0 Statistica ouMP 0 SigmaXt, Thave not used al but used Minitab when I worked with ‘Cummins India and later for Black Belt Training in vari= ous companies. T have also used SigmaXL for Six Sigma Green Bele Training. Both are good softwares. However, SigmaXL is Excel based user friendly software that can support green belts where analysis usually does not re- quire very complex tools. Minitab on the other hand can Qusitywrrta | August 2006 9