2k实验设计探讨.ppt

第 5部分: 第5部分: 2k实验 23 和 24 全析因实验定义： 例1: 清洁过程 DOE术语 -‘ 正交’’ 分区 分区 (续) 随机与隐藏变量 重复和反复 重复还是反复？取决于不同的情况 DOE图1: 主要影响图 2个水平时独立变量间的交互作用 2个水平独立变量的交互作用 DOE图2: 交互作用图 现在，我们看一下温度和浓度的交互作用 时间和浓度交互作用图 主要影响和交互作用 24实验立方图 25实验立方图 DOE和样本容量 DOE和样本容量 (续) 基本样本容量表仅适用于连续数据 数据窗口输出结果 进行实验: DOE图1: 主要影响图 图1:残值平均值和标准变差的主要影响图 平均值残渣值的交互作用图 残渣值的标准差交互作用图 DOE图3: 平均值和标准差的立方图 残渣值平均值的立方图 残渣值标准差的立方图 既然我们观察了所有图表，让我们用数学方法分析实验设计… 会话窗口中的分析结果 课堂练习: 使用因素实验减小变差 使用因素实验减小变差 (续) 主要概念 - 第5部分:2k实验 In an orthogonal array, the dot product sum of columns is 0. Define Orthogonal array - the dot product ? = 0 Equal # low levels and 高 levels. Thus the ??= 0. Balance - same sample size all calculations. Full factorial - every possible combination of variables. A Replicate is a set of tests. If you do two sets, you do two replicates. This requires redoing settings. Repetition - No presetting. Just duplicate run. If process has little variation, replication may not be required. If cost is 高, and variability is low, you can avoid replication. Replication is prescribed if Gauge R R is marginal or 高. Why take 2 data points Variability Repeatability Always try to plan for more than 1 data point. When choosing levels, choose them at some level of significance to create a change (you could look at the process capability to “bookend” it, also). Summation effect This graph is generated by graph on last page. With HOT, run can run fast. With WARM, you have to run slower. ?Is there an interaction. What would you conclude Ans: At colder temp (warm) time has an effect on residue. There is interaction. At hot, time doesn’t matter. What does this graphic say about concentration and temp. What about concentration and time. All columns are orthogonal - multiply together and add up. The answer will be “0”. i.e. tempxtime mults temp + time for that run -1*-1=+1 2-way interactions are common; 3-way interactions are rare Use cube (tinker toy) and explain points of cube 6 -1, -1, -1 Each point in this cube is generated by 1