【導(dǎo)讀】Identifyrootcauses. Atheorythathasbeen. GreenBeltMaterials—ProcessAnalysis. ActivityFlowCharts. DeploymentFlow. Charts. ValueStreamMaps. DataDoor:. ScatterPlots. Multi-VariAnalysis. ProcessorDataDoor?Toimprovetheunder-. Totacklecycletime. WhatistheTool?WhyIsItUseful?WhenIsitused. WhenShouldItBeUsed?

　　

【正文】 50 1 2 3 4 5 6 7 8 9 10 11 121 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 41 0 . 1 51 0 . 2 51 0 . 3 51 0 . 4 5C o i lThicknessLMRM u l t i V a r i C h a r t f o r T h i c k n e s s B y W i d t h C o i lL e n g t hW i d t hExercise 4: Answers, cont. 51 Using MultiVari with Upstream Variables Using upstream variables (X39。s) as well as output measures (Y39。s): ?Gather information on relevant continuous X39。s at the same time you gather product samples and measure output characteristics ?Associate changes in X39。s with patterns of variation in the Y39。s. 52 Illustration (1) Suppose the team in the paint thickness example had been able to record a pair of continuous X39。s: ? line pressure into the paint booth ? paint booth temperature. Neither of these factors apparently will account for the differences between front and back panels. Is there any other potential relationship? (Remember that the day to day variation is smaller than the variation within 3 consecutive units.) Take a few minutes to review the main effects plot on the next page and note down some ideas. 53 Illustration (1), cont. P a n e l I DP o s i t i o nA M / P MD a t e121110 9 8 7 6 5 4 3 2 1FrontBackPMAM4/19/994/18/990 . 0 6 50 . 0 6 00 . 0 5 50 . 0 5 00 . 0 4 5ThicknessM a i n Ef f e ct s Pl o t D a t a M e a n s f o r T h i ckn e ssDate T i m e A c tua l ti m e Li n e P r es s ur e ( ps i ) P ai nt B oo th T ( deg F)4/18/ 99 AM 9:30 4/18/ 99 PM 14:0 0 4/19/ 99 AM 7:45 4/19/ 99 PM 13:3 0 54 Answer ?Line pressure moves from low to high on both days, in line with the pattern with the morning to afternoon difference on the main effects plot. ?Higher line pressures might account for more paint delivered to the surfaces in the afternoon vs. the morning. ?The booth temperature does not look like it has any relationship with morning to afternoon differences. 55 Using MultiVari with upstream variables: a second approach ?Use information about the X39。s to guide your sampling approach ?Seek to observe the process when the X39。s are at or near their extreme values You can guide your data collection if you can identify when the X’s are likely to be at or near extreme values: V ari a ble Level W he n y o u are l i kel y t o obse rv e t he l ev elX1 LowX1 Hi g hX2 LowX2 Hi g h. . . . . .56 Illustration (2) ?Your improvement team is studying sealing strength of containers. ?Your preliminary work has identified press temperature and pressure as two continuous variables known to affect sealing strength in other situations. ?In addition, you have identified persistent differences between two presses, A and B with respect to sealing strength, with A producing consistently stronger seals over the past month (on average 4% stronger.). ?Press B has produced more defective seals than A. ?Discussion with operators reveals that press B will run at 175 to 200 degrees (F) and pressure will vary between 36 to 44 psi. These variables are used to verify operation and have not been entered into any database in the past. 57 Illustration (2), cont. The team proposes to sample four containers from press B at four press conditions and then test the strength of the seals (a total of 16 tests). ? Low Temperature: 175 to 180 deg F ? High Temperature: 195 to 200 deg F ? Low Pressure: 3637 psi ? High Pressure: 4344 psi What remendations would you give to the team before they carry out their proposal? 58 Answer 1. It is reasonable to focus on press B since a persistent difference has already been observed between presses A and B. Given limited resources, it makes sense to focus on the variation within press B39。s output. 2. The team should record the time of day and information about press conditions (including shift and operators) that might provide additional clues. 3. When the team grabs samples at the low or high temperatures, it should also record the pressure at the same time. Similarly, when they grab samples at the low or high pressures, they should record the temperature. 4. Jumping right to observations on temperature and pressure may miss other signals that might be revealed by a more plete multivari analysis that considered within unit variation, unit to unit variation, subgroup to subgroup variation and longterm variation. Also, there may be other families of variation (related to geometry of units within the press, for example) relevant to the study. 5. Has the team verified the performance of the temperature and pressure gauges? 59 Summary Multivari analysis is a method that helps you depose total variation into ponent families. By means of a multivari graph, you seek to match patterns in the variation with potential causes. You use the method to narrow down potential causes and find clues as you build a useful causal theory. 60 Reflection Consider how you might apply Multivari analysis to your project. What questions do you have now about this method?