追求卓越品质MINITAB实现制造业品质提升Johnson Transformation for MinutesProbability Plot for Original DataSelect a Capability Sixpack of MinutesN21399Johnson Transformation with SB Distribution -Value< + * Log( ( X + ) / ( - X ) )90I ChartCapability HistogramLSL*USL* PUCL=_LSL*=*-2LCL= Range ChartNormal Prob Plot04008001200(P-Value = means <= )AD: , P: Plot for Transformed DataUCL=__N213MR==-202490P-Value for Best Fit: 25 ObservationsTransformed Capa PlotZ for Best Fit: Transformation Type: SB0Prduct function + * Log( ( X + ) / ( - X ) )-2024Worksheet: : TypellleeeaaavvvnininiitititiiigggsssiiirnrnrnOeOeOeSSSAnglessseeeeeerrrgggeeeddd 05049ValuesIndividual ValueMoving RangeMean of WeightPercentPercentP-Value for AD test
内容纲要Minitab--六西格玛品质的实现工具六西格玛概述Minitab的几个分析实例案例分享
Overview总述•Six Sigma is a methodology that drives process improvement. MINITAB statistical software progresses the Six Sigma initiatives through its easy-to-use interface and informative results. This talk to will focus on applications of statistical tools throughout the phases of Six Sigma projects from both the manufacturing and service industries. •六西格玛是推动过程改善的一种方法。MINITAB统计软件通过它易操作的界面与有价格的结果来推动六西格玛运动的发展。这说明贯穿在制造业与服务业的六西格玛项目整个阶段的重点在于统计工具的应用上。
Six Sigma Model六西格玛模式–Define –Determine Process to Improve–定义－确定改善流程–Measure –Current State of Process–测量－当时的流程形态–Analyze –Root Cause Analysis–分析－根源分析–Improve –Change Process–改善－改变流程–Control –Sustain the Gains–控制－维持效果
DMAIC改进流程过程质量D界定D界定C控制M量测D界定C控制D界定M量测I 改进A分析C控制I 改进M量测A分析I 改进A分析螺旋式上升的DMAIC流程
Define定义 MINITAB¾Determine the process to be improved•Pareto帕莱托图¾Identifyyour Customers•Gage R&R¾Identify your Problem•Attribute Agreement ¾确定需要改善的流程属性一致性分析¾鉴别你的客户•Control Charts 控制图¾鉴别你的问题•Capability Analysis Measure测量能力分析Determine the extent of Problem•Graphics 图表Identify your Inputs & Outputs (X’s & Y’s)Measure it确定问题范围鉴别你的输入与输出（X’s & Y’s）测量它
Analyze分析Screen potential causes筛选潜在原因Identify sources or root causes of problem鉴别问题的来源或根源Draw Conclusions引出结论 MINITAB•Regression 回归分析•Gage R&R•Capability Analysis 能力分析•Hypothesis Tests 假设检验•Graphics 图表•ANOVA 方差分析•DOE 试验设计•Correlation 相关性
Improve改善Identify what improvements to implement确定改善的工具Determine optimal settings确定最佳方案Confirm results确定结果 MINITAB•Gage R&R•Graphics 图表•Attribute Agreement •Capability Analysis 能力分析属性一致性分析•Response Optimizer 优化响应•Hypothesis Tests 假设检验•DOE 试验设计•ANOVA 方差分析•Regression 回归分析
Control控制Sustain the gains维持效果Permanent change持久改善MINITAB•Control Charts控制图•Capability Analysis能力分析
示例1某生产摩托车的公司，要降低有缺陷的测速计所产生的质量成本。在检验过程中，有一定数量的测速计遭到拒收，并记录了缺陷的类型。在工作表的“缺陷”列中输入缺陷的名称，并在“计数”列中输入相应的计数。我们都知道，通过集中处理造成大部分拒收的缺陷，可以节省大部分费用。Pareto 图将帮助您确定导致大部分问题的是哪些缺陷。
Define: How can we reduce quality costs associated with defective speedometers?定义：我们如何才能减少由于计速器的缺陷而造成的质量成本Pareto Chart of Speedometer Defect Types100400803006020040Weighted Pareto Chart of Speedometer Defect Types100207000100006000Defect TypettssrgrepenwikialhessPt80C rau 5000OceogGtS nHe iyl % TypettsrgprepenakiairlhssPtc Cau OeSoGtg Henyli % CountPercentPercent
Measure: Can we trust our data?测量：我们可以相信我们的数据吗？Production > > > >Inspection = Defects产品检查缺点15%=10%=
Measure: Attribute Agreement Analysis测量：属性一致性分析Each Appraiser vs Standard 每一位鉴别者VS 标准Appraiser #Inspected #Matched Percent 95 % CI1 20 18 (, )2 20 15 (, )Between Appraisers 鉴别者之间#Inspected #Matched Percent 95 % CI20 17 (, )Date of study: June 15. 2005Assessment AgreementReported by:METName of product:Springfield Plant Appraiser vs % CIPercent80604020012AppraiserPercent
Measure: Capability Analysis测量：能力分析Understanding the Capability of the Process了解流程能力LSLUSLProcessDefectiveDefectiveAcceptable ProductProductProductCustomers Wants
Capability Analysis: Metrics能力分析：评价参数•Sigma -also known as Z Bench or Benchmark Z•西格玛-也称Z Bench 或者Benchmark Z•Cpk -Potential Capability•Cpk-潜在能力•Ppk -Process Potential•Ppk-潜在流程能力
测量：了解当时流程的形态某电缆制造商要评估电缆的直径是否符合规格，线缆直径规格为 + cm 。分析员评估过程的能力以确保其满足客户的要求，即Ppk为。分析员每小时从生产线中取5 根连续的线缆作为一个子组，并记录直径。评估当时的流程能力?
Capability Analysis: Assumptions能力分析：假设•Process is stable over time.•流程在时间跨度方面是稳定的。•The assumed distribution (usually normal) is an adequate fit to the sample data.•假定分布（通常状态）对样本数据是足够的，适合的。
Assessing Stability over Time: Control Charts评估在时间跨度方面的稳定性：控制图Xbar-R Chart of HardnessUCL=__X=====0161116212631364146SampleSample MeanSample Range
Assessing the Distribution: Probability Plots, Histograms, Anderson-Darling test评估分布：概率图，柱状图，Anderson-Darling试验Probability Plot of HardnessNormal for Hardness20Anderson-Darling Normality % Confidence Interval for % Confidence Interval for % Confidence Interval for StDev95% Confidence
能力分析－SIXPACK直径 的 Process Capability SixpackXbar 控制图能力直方图LSLUSLUCL=规格 == 控制图正态 概率图AD:，P: ===最后 20 个子组能力图组内组内整体标准差标准差整体*规格5101520样本值样本极差样本均值
Measure:Capability Analysis测量：能力分析Understanding the Capability of the Process了解过程的能力直径 的过程能力LSLUSL过程数据组内整体目标*潜在（组内）能力样本均值样本 标准差（组内）标准差（整体）整体能力*实测性能预期组内性能预期整体性能PPM < < < > > > 合计 合计 合计
Analyze: Identify Significant Inputs分析：鉴别有意义的输入Process Output流程输出Y = f (x)Process Inputs流程输入
Analyze: Root Cause Analysis分析：根源分析Root Cause Analysis = Effect of Input Variables 根源分析＝输入变量结果Hypothesis Testing假设检验Suppliers供应商Defect Rate缺陷比率Test and CI for Two Proportions 两个比例的检测和置信区间10%Sample X N Sample p=1 137 1392 189 1549 % CI for difference: (, )Test for difference = 0 (vs not = 0): Z = 12%=P-Value =
Improve: Determine Optimal Settings改善：确定最佳方案A manufacturer wants to determine the best settings for the adhesion of automobile paint. Two variables are under consideration:一家工厂想要为汽车附着漆确定最佳方案。需要考虑的两个变量是：Pressure压力(310 kPa vs. 380 kPa)Primer Type初始类型(A vs. B)A full factorial design is run. The goal is to determine the optimal settings for the response.
Improve: Design of Experiments改善：试验设计
DOE: Fit the model试验设计：适合的模型
DOE: Graphical Tools试验设计：图形工具Main Effects Plot (data means) for Plot (data means) for of AdhesionMean
DOE: Response Optimizer试验设计：优化响应
Control: Sustain the Gains控制：维持效果Process Capability Sixpack of QuenchingXbar ChartCapability HistogramLSLTargetUSLUCL=== ChartNormal Prob PlotAD: , P: =_R==016111621263136414639404142Last 25 SubgroupsCapability MeanValuesSample Range
Case Study: Cost Reduction within Human Resources案例分析：在人力资源内部的成本降低The HR manager at a large multi-national investment firm would like to reduce operating costs by 8% over the next 12 months. Departmental expenditures data are collected, and the results are as follows:一个庞大的多国投资公司的人力资源经理想要在接下来的12个月内减少营业成本8%。收集的部门支出数据和计算结果如下：Recruiting:招募：$ 1,100,000Compensation/Benefits:补偿/津贴：$ 3,400,000General Overhead/Supplies:综合费用/供应品: $ 200,000Training:培训：$ 80,000Capital projects:投资项目：$ 500,000Consultants:咨询：$ 160,000
Benchmark the process: Pareto chart流程基准：帕莱托图Pareto Chart of 2002 nlOsRo %
Target: Recruiting costs for entry level positions目标：招募成本水平位置Pareto Chart of 2002 nlOsRo %
semuser gnirotSsemuser weiver ot emiTCost Reduction within HR: Brainstorming results displayed using a Cause and Effect Diagram人力资源内部成本减少：使用因果图显示头脑风暴结果Recruiting for Entry Level positionsMeasurementsMaterialPersonnelPamphletsInterviewInterviewsscoringPromotionalMaterialsTravel ExpensesDisplaysEntry LevelRecruitingCostsSchedulingInterviewWeb ServerCampusTravel processConnectivity ofdatabase(s)WebMethodsHardware
Reducing recruiting costs减少招募成本Due to high costs and significant time needed to screen resumes submitted electronically, the team decides to evaluate the screening process for resumes submitted via the web.由于电子提交的简历筛选需要高成本和长时间，小组决定评估通过网络提交简历的筛选流程。Think about the response. Resumes are evaluated by team members -what type of data is this?考虑输出。经过小组成员评估的简历－这是什么类型的数据？
Attribute Agreement Analysis属性一致性分析Binary response –Pass or Fail. 二元响应－通过或者失败A resume either proceeds to the next level or not.简历到达或未到达下一个水平。4 people are responsible for screening resumes. 4个人负责筛选简历。Each is presented with 50 resumes, randomly selected and already judged to be either a pass or fail.
Attribute Agreement Analysis属性一致性分析Assessment AgreementWithin AppraisersAppraiser vs % % CI100100PercentPercent9595909085858080AlexDavidNinaWendyAlexDavidNinaWendyAppraiserAppraiserPercentPercent
Attribute Agreement Results属性一致性结果
Baseline: Test of One Proportion基线：一个比例的测试2,200 resumes are received and 250 are considered ‘viable’. Determine the % of good resumes coming in via the 个收到的简历中，250个被考虑为“可行的”。Sample X N Sample p 95% CI确认通过网络收到的好的简历所占的百分比。1 250 2200 (, ) It is estimated that between 10 and % of incoming resumes received via the web are viable candidates.经过评估，有10－%通过网络引入的简历被
Making a change to the website:将变化体现到网站：A change is made to the website that should screen candidates before they are able to submit resumes. After 3 months a random sample of 600 resumes is selected and 285 are judged to be viable.在他们能提交简历并将变化体现到网站上之前，应筛选出侯选者。三个月之后，600份简历中的一个随机样本是经挑选的，285个被判断为可行的。
Tests for Two Proportions两个比例测试It is estimated that % of resumes are now viable. The change to the website seems to screen, or dissuade unqualified applicants from submitting their resumes. The test for two proportions demonstrates a statistically significant increase after the change.经评估，简历中的%目前是可行的。这个体现到网站上的变化显示提交的简历似乎是经过筛选的，或者是劝阻了不合格的申请者。两个比例检验证明了变化后在统计意义上的显著增长。Test and CI for Two Proportions Sample X N Sample p1 285 600 250 2200 = p (1) -p (2)Estimate for difference: % CI for difference: (, )Test for difference = 0 (vs not = 0): Z = P-Value =
Considerations for Data Analysis within Six Sigma考虑六西格玛的数据分析•Selecting good candidates is crucial to the success of a Six Sigma implementation-particularly for those who will be responsible for data analysis.•选择好的侯选者对于执行六西格玛并要为数据分析负责的人来说是至关重要的•Proper training in the appropriate area is also important -Service Quality students should receive Transactional/Business training and Manufacturing students should receive manufacturing training, because the tools that each will utilize somewhat differ.•在适当的领域进行适当的培训也是同样重要的，服务质量领域的学生需要接受交易/商务培训，制造领域的学生需要接受生产培训，因为每一个工具都将被不同地利用。•“Which Tool When?”-Knowing which statistical tool to use to solve a problem is sometimes difficult, particularly in the early stages of Six Sigma.•“何时使用哪个工具？”－了解使用哪个统计工具解决问题有时是困难的，这在六西格玛的早期阶段尤其显著。
Thank you for your time.
