【導讀】指所有的有控制的輸入及有計劃的結(jié)果分析的試驗。原因與結(jié)果矩陣/失效模式。Blocking:將干擾最小化的方法。Interaction：兩個因素合起來對總輸出的影響將高于兩個單獨的因素造。世界,而不是試驗試驗數(shù)據(jù)。確定限制試驗可采用方案數(shù)與試驗次數(shù)的限制條件。決定你將做多少次試驗。結(jié)合你的試驗目的，選擇最佳試驗設(shè)計。那一個因子比較重要？的距離)的廣告很有興趣。你不確定這些球棒和球如何能提高球的距離，但。A草地四面環(huán)山，幾乎。沒有什幺風；B草地位于草原上，因此有很大的風。你知道該怎幺做來提高你的成績?提高我的Golf分數(shù)。增加擊球的距離。游戲結(jié)束的最終分數(shù)。

　　

【正文】 rst k1 factors from the base k個因素 是 k1個 因素的相互作用。 This interaction is found by multiplying the factor level settings together for the first k1 factors from the base design. 用基礎(chǔ)設(shè)計的前 K1個因子的因子水平設(shè)置相乘可獲得它們的相互作用 實驗次數(shù)計算 Number of runs = N = 2k1 45 Exercise Half Fraction of a 25 Factorial 5個因子的半因子實驗設(shè)計 Generate a half fraction experiment for k = 5 factors, A, B, C, D, and E. A B C D E Factors 46 Answers Half Fraction of a 25 Factorial – + – + – + – + – + – + – + – + A – – + + – – + + – – + + – – + + B – – – – + + + + – – – – + + + + C – – – – – – – – + + + + + + + + D + – – + – + + – – + + – + – – + E 47 TradeOffs Between Full Factorial and Half Fraction Designs 比較全因子與半因子 Number of Effects Computed 可計算的實驗效果的數(shù)量 Effects Mean Main Factors 2 Factor Int. 3 Factor Int. 4 Factor Int. 5 Factor Int. Total Effects Computed計算的總效果 Full Factorial Half Fraction 1 5 10 10 5 1 32 1 5 10 — — — 16 也是實驗的次數(shù) ? Are the additional runs worth it? 額外的實驗次數(shù)是否值得 ? ? What happens to the higher order interactions? 高次的相互作用如何 ? 48 A 2Factor Experiment with Confounded Effects 2因子實驗中的混淆 – – – – + + + + Factor A 1 2 3 4 5 6 7 8 Run – – – – + + + + Factor B The effects of Factors A and B are confounded. A因素和 B因素互相混淆 130 125 133 130 50 85 79 93 Response Confounding is the bining of the effects of two or more factors into one resulting number such that the magnitude of the effects of the individual factors cannot be separated. 混淆是指兩個或以上因素的作用組合成一個結(jié)果，因此單個因素的作用無法分開。 49 Confounding in the Half Fraction半因子實驗中的混淆 A B C D AB AC AD mean = = = = = = = = BCD ACD ABD ABC CD BD BC ABCD For the half fraction, each letter must be present on one side of the equal sign or the , 等式左右的因子作用混淆了 , 所以可用一邊的字母代替另一邊字母的作用 – + – + – + – + A – – + + – – + + B – – – – + + + + C – + + – + – – + D + – – + + – – + AB + – – + + – – + CD 50 Effects Plot for the Full Factorial 全因子的效果圖 A B AB BC BD ABCD AC AD ABC ABD C D CD ACD BCD 0 5 10 15 Magnitude of Effect Effects from the Full Factorial 51 Effects Plot for the Half Fraction半因子試驗的 效果圖 0 5 10 15 A B AB AC C AD D BC BD BCD ABC ABD CD ACD + + + + + + + Fraction Effects from the Half Magnitude of effect 52 Summary of the Half Fraction半因子試驗的總結(jié) The half fraction of a full factorial can often provide the same information as the full factorial, with only half the number of 的信息 ， 卻只需 進行 一半的試驗 ? Fewer runs saves time and ? More plicated to analyze (must understand confounding).分析更復雜 ， 必需理解存在混合效應(yīng) ? In designs with few runs, important effects (such as 2way interactions) are 中 , 有一些因素的效應(yīng)混合了 Benefits Costs Note: Some texts (and Minitab) call confounding aliasing. 53 54 DOE 基礎(chǔ) — 小測驗 1) T/F DOE is an iterative process....one DOE generally leads to another. DOE是一個反復的過程 … 一個 DOE通常引導出另一個。 2) T/F The experimental design should be determined by the objective and constraints. 試驗設(shè)計應(yīng)該由目的和限制條件所決定。 3) The following is are conditions for a DOE: 下列是一個 DOE的條件： Account Manager 客戶經(jīng)理 Joe, Sue, Kelly, amp。 Don Sales Region銷售區(qū)域 North, South, East, West Product Type產(chǎn)品類型 Fluorescent, CFL How many factors are there? 有多少個因素？ What are the factors? 什么因素？ How many levels are there for each factor? 每個因素有多少個水平？ What is the total of treatment binations? 要試驗多少次？ 4) T/F In order to run a DOE the response must be variables type data. 為了執(zhí)行 DOE， 輸出結(jié)果必須是連續(xù)變量。