第4期课件：6SIGMA简介及导入.pptVIP

Definition A well-designed experiment eliminates the effect of all possible xs except the ones that are changed. If the output or “response variable” changes significantly, it can be tied directly to an input x variable that was changed rather than some other x variable that was not changed. Reference: Six Sigma Memory Jogger II, page 105 Generate Improvement Alternatives Pilot Validate Improvement Create “Should Be/Future State” Process Map Perform Cost/Benefit Analysis Improve Phase FMEA Experiments Compared to Other Techniques By comparison, other techniques are not as effective as statistical experimentation: Observation takes a long time, costs a lot of money, has a small probability of detecting optimum levels, has a low confidence in results, and offers no ability to identify independent main and interaction effects. Guessing is only slightly better than observation, and still offers no ability to identify independent main and interaction effects. Expert judgment further improves, but still offers no ability to identify independent main and interaction effects. Conducting one-factor-at-a-time experiments improves the chances of finding optimum levels, but increases financial costs and time spent, and still fails to identify independent main and interaction effects. Generate Improvement Alternatives Pilot Validate Improvement Create “Should Be/Future State” Process Map Perform Cost/Benefit Analysis Improve Phase FMEA Strategy for Experimentation As with brainstorming, the Project Team will review the Problem Statement and Project Goal in the Project Charter to ensure that everyone on the team has a clear view of the improvement criteria. With an understanding of the improvement criteria, the team will follow the strategy outlined above to generate and evaluate potential improvements. When determining factors to use in experimentation, the Y=f(x) relationship is examined: The response variables are the customer CTQs identified in the Project Charter