【正文】 t to know ? Statistical Problem Understood ? Link Process Map to Fishbone w/C, N, X Labeling ? Ensure Data is Collected for All Xs ? Drill Down to Vital Few Xs via ANOVA, GLM, Ttest, Ftest ? List of Vital Few X’s ? Quantified Financial Opportunities What did we learn ……… What were the causal Xs ? How did we identify them ? PM/CE/CNX/SOP Process Mapping Why? * Process visual foundation for current situation and analysis ? Aids in identifying bottlenecks, redundancies and waste ? Look for Non Value Add ? Look for Variables (CNX) How? ? Determine beginning and end of process to be mapped ? Involve people knowledgeable about the process ? Brainstorm steps group in major process areas ? Layout activities in sequence ? Validate by physically walking through process PM/CE/CNX/SOP CE: Cause Effect Sources of Variation Output(s) Specs People Material Machine X C C N C N X X Measurement Method Environment Every variable on the diagram should be labeled as either: C = Constant N = Noise X = Experimental variable or Factor PM/CE/CNX/SOP SOP: Standard Operating Procedures Standard Operating Procedure are rules that we define to ensure that we have Consistent processes in everything we do. ? Based on good judgment ? Common sense ? Engineering knowledge Remember ISO 9000! Make sure we have defined processes and that the rules are being obeyed by all. 數(shù)據(jù)分析 size risk ? Is the risk of say there is a difference when there really isn’t one(生產(chǎn)者冒險(xiǎn)率 ) ? Is the risk of not finding a difference when there really is one (消費(fèi)者冒險(xiǎn)率 ) ?/? is he magnitude or size of the difference been tested. This is sometimes called the test sensitivity. αδ / σ 20% 10% 5% 20% 10% 5% 20% 10% 5% 20% 10% 5% β0 . 2 225 328 428 309 428 541 392 525 650 584 744 8910 . 3 100 146 190 174 234 2890 . 4 56 82 107 98 131 1620 . 5 36 53 69 63 84 1040 . 6 25 36 48 44 58 720 . 7 18 27 35 32 43 530 . 8 14 21 27 25 33 410 . 9 11 16 21 19 26 321 9 13 17 略 16 21 26 略1 . 1 7 11 14 13 17 211 . 2 6 9 12 11 15 181 . 3 5 8 10 9 12 151 . 4 5 7 9 8 11 131 . 5 4 6 8 7 9 121 . 6 4 5 7 6 8 101 . 7 3 5 6 5 7 91 . 8 3 4 5 5 6 81 . 9 2 4 5 4 6 72 2 4 4 4 5 62 . 1 2 3 4 4 5 62 . 2 2 3 4 3 4 52 . 3 2 3 3 3 4 52 . 4 2 3 3 3 4 52 . 5 1 2 3 3 3 42 . 6 1 2 3 2 3 42 . 7 1 2 3 2 3 42 . 8 1 2 3 2 3 32 . 9 1 2 2 2 2 310%20% 5% 1%A Basic Sample Size Table Applies to Continuous Data Only δ/σ For example : The 1st adhesive has an average of and a standard deviation of The practical significance is such that any alternative Adhesive must have an average strength of 20 or more to make the change worth while. Q: For α=5% β=10% what should be the sample size in each level of the experiment? Ans: How big of a change is important? δ/σ=()/= α=5% β=10% So follow the table we find the sample size is 5 : Comparing Means : Comparing Variances (Between 2 Groups) :(Analysis Of Variance) (Between multigroups) Improve Statistical Control ? PokaYoke Plan (if Applicable) ? Confirmed Vital X’s ……Are They Statistically Significant ? ? Sample Size Calculation For Confirmation Run ? DOE Plan (if Applicable) ? Regression Equation Linked to statistical Problem ? SOP Changes and/or Optimal Solution Identified ? Plan to Implement SOP Changes and/or Optimal Solution What contribution did each vital X have to the Y ? How can we control the Xs ? What do we want to know ? How do we determine a model Correlation Regression Correlation tells if you have a relationship between two variables(Y and X, or two Xs) Regression is used to identify the nature of the relationship, and be able to predict Y while better understanding Y and possibly improving the controllability of Y 10987654152151150p u l ltempCorrelation Correlations (Pearson) Correlation of pull and temp = 210120 . 50 . 0 0 . 5 1 . 0N o r m a l S c o r eResidualN o rm a l Pro b a b i l i t y Pl o t o f t h e R e s i d u a l s(res po ns e is pu ll)Regression Analysis (Minitab software) The regression equation is pull = 353 temp Predictor Coef StDev T P Constant temp S = RSq = % RSq(adj) = % Analysis of Variance Source DF SS MS F P Regression 1 Error 30 Total 31 Unusual Observations Obs temp pull Fit StDev Fit Residual St Resid 12 150 31 151 DOE (Design of experiment) 找尋最佳之生產(chǎn)條件 (製程參數(shù) ) Design of Experiments 6? Overview SCREENING OPTIMIZATION CHARACTERIZATION ? For Experiments Involving a Large Number of Factors ? Useful in Isolating the “Vital Few “ from the “Trivial Many” ? For Experiments Involving a Relatively Small Number of Factors ? Useful When Studying Relatively Unplicated Effects Interactions ? For Experiments Involving Only 2 or 3 Factors ? Useful When Studying Highly Complicated Effects Relationships DOE is More Effective Than Testing One Factor at a Time Control What do we want to know ? Statistical Control ? PokaYoke Plan (if Applicable) ? Measurement of Final Capability Using Confirmation Run ? Comparison of Before and After Distributions ? How Do You Control or PokaYoke the Vital Xs ? ? Is the Learn