PHILIPS绿带训练教材-MSA测量系统分析研究报告.ppt

小结 重复性 (= 一人重复) 可再现性 (= 多人重复) 两个因子引起测量变差 RR 应低于允许公差的 30% ，最好低于 10%. 将结果绘制成图表可能会比文字分析更有效 文字分析有时是引起变差的最大原因！ 补 充 内 容 测量系统评估 初试 复试 P M P / M 8 其它方法: ISO 图 同一评估员！ 破坏性测量的MSA 使用规格极其接近的部件，可以当作同一部件 同一时间，地点，同批出产 用同一批次内的变差值估算重复性 如果批量很小，只有一个操作员可以测量，则使用环(nested)量具 RR 如果批量很大，每位操作员可以多次测试本批产品，则使用交叉(cross)测量 RR（批次和操作员交叉） Further reading in 1 03e M MeasSysAnalysis and Destructive Testing.pdf 属性测量的MSA 首先将属性数据转化为变量数据 如不可行，采取简易方法: 选择20 个部件，某些略高或略低于规定要求 两位操作员两次测量 (不分次序) 如果四次测量相吻合，量具可以接受 * 时间: 2 小时 Assignment can be done with line measures (file 1 03b M gauge experiment.ppt) or with catapult. This still has to be developed. 工具: 文档 1 03b M gauge experiment.ppt or Catapults 每3个学生用1 台带Minitab的电脑 * * Ask: what is the objective of measuring? When looking at ice scating or gymnastics we find it normal that we need 3 or 4 jury members to give their score. We are also ok with the fact that they score not all the same, because we have the feeling that is the best wat possible. For a business environment this is of course not acceptable. When we measure something we expected it be the exact value It should be accurate (the real value, no bias) with a resolution fit for the purpose (so no km counter to measure the size of a building Also when you repeat the measurement you should get the same value And also when somebody else does the measurement you expect that he/she reproduces the same value as well * * * * * * When only graph appears ask: how do you think about the resolution of this measurement? Answer: its too low. There are too many samples exactly on the lsl or just one digit above or beneath, so a big change on things approved/disapproved where it should be the other way around * One cannot answer the questions with 100% certainty because you do not know the real value of the test piece. Operator Brown is the most consistent and, given the available info, the best operator. Corrective action would be a training in how to me