




版权说明:本文档由用户提供并上传,收益归属内容提供方,若内容存在侵权,请进行举报或认领
文档简介
1、 DOEDesign of Experiment- Screening DOE (Part1)1 DOEObjectivesBy the end of this course, the participants should be able to:Identify tools used to quantitatively segregate the vital xs. Understand the structure of DOE.Design, Perform and Analyze a simple 2-level experiment.Use Minitab to assist in t
2、he design and analysis of experiments.2 DOEIntroductionDuring design a new process(product) or process improvement, sometimes several possible factors (input, x) are identified impacting response (output, y). To get desirable y, Owners must control xs. But sometimes it is not possible to control all
3、 xs. So we need explore and uncover the vital few xs from possible many, and need to identify how much impact each x has on the response y.Experimental design (DOE) usually being used for to explore and confirm the effect of xs on the ys. By experimentally manipulating the independent variables, DOE
4、 provides an efficient and economical meansfor reaching valid and relevant conclusions about a process.3 DOEExample Lifetime of Cutting toolsThe life of a quite expensive cutting tool of a stamping process depends on the factors of the cutting speed, tool angle, temperature, humidity and raw-materia
5、l of component basing on team brainstorming result.To get the longest lifetime of this cutting tool, a team hasbeing commissioned.1.2.3.Which factors are key to the life time of cutting tools?How to set up key factors to achieve longest lifetime? What is the longest lifetime?4How will you get answer
6、s to above three questions? DOEExample HPC Product DesignDG HPC team is designing a new type of high performance cable, key customer requirement on this product is the impedance of cable.Team brainstorming and benchmarking result show cable core material, cable length, soldering quality (depends on
7、ultrasonic intensity, and soldering time) and operators assembly skill possible factors. A team being commissioned for below task:1.2.3.Which are key factors to HPC cable impedance? How to set key factors to achieve lowest impedance? What is the lowest impedance?5How will you get answers to above th
8、ree questions? DOEDefinition of DOEExperimental Design is a structured proactive process for investigating the relationship between input and output factors. Multiple input factors are considered and controlled simultaneously to ensure that the effects on the (multiple) output responses are causal a
9、nd statistically significant.6 DOEHistory of DOE1920s with Sir. R. Fisher; Origination is agriculturally-based (e.g. treatments); 1950s with technique development; 1980s withdevelopmentthrough computer assistance.7 DOEWhere DOE used Process out of control Process not capable Process in control, but
10、level not satisfactory Process may not be operating at its best Process is new Product is new8 DOERecap: Other quality tools used to identify vital factors Fishbone (Cause-Effect Chart) FMEA Pareto Chart Fault Tree9 DOEUsing Fishbone to Identify Vital FactorsOperator(30%) & Method(40%) are vital fac
11、tors! ?10 DOEUsing FMEA to identify vital factorsMold temperature, Pack pressure and Material are vital factors!?11Ke y Pro ce ss I np utPo te nti al Fa ilu re M od ePo te nti al Fa ilu re EffectsS E VPo te nti al Ca use sO C CCu rren t C on tro lsD E TR P NM art er ia l us edProc ess in g w in dow
12、s of di ffe rent raw m aterail are di ffe rentO perating ranges m ay be too narrow8Inharent properties and lo t- to- lo t variatio n of the raw m aterial us ed8M ateral properties chec ked by QA5320M eltTem perat ureSetting tem perature is to high/ lo wO ver pac k in g / Hig h s tres s7bigger di ffe
13、rent betw een ac tual/ s etting t em p.5t herm om eter7245P ack in g P resurePac king P res sure is toohigh or wrong lo c atio n of p- vO ver pac k in g/s hrin k age8ex perie nce/s k il l of m o ld in g8none ( dat a sheet )8512In je c tio n SpeedToo h ig h/lowO rent ia tion of m at er ia l/ st ress
14、dist rubr iti on5ex perie nce/s k il l of m o ld in g7none ( dat a sheet )7245B arrel Tem p.im proper s et tin gO rent ia tion of m at er ia l/ st ress dist rubr iti on8does nt c heck i t8none ( dat a sheet )8512H o ld in g Tim eToo longO ver pac k in g/ H ig h s tres s5ex perie nce/s k il l of m o
15、ld in g8none ( dat a sheet )7280V ent in gpoor venti la tionju ic in g ( chem ic al A ttac k )8conc ept of too li ng des ig n8N on-venting around P/L4256C ool in g s ys temUn-evenU n-even t em perature / s hrink age rate / s tres s5conc ept of too li ng des ig n1none210 G at e Loc atio nLong fl ow l
16、enghts tres s d is tr ib utio n4conc ept of too li ng des ig n1none28 DOEUsing Pareto Chart to identify vital factorsCount98100%94.7%84.2%7654321073.7%Factor1Factor2Factor3Factor4Factor5Factor6So Factor1& Factor2 are vital Factors!?121223557.9%631.6% DOEDOE compare with Fishbone, FMEA, Pareto Chart
17、DOE VS Fishbone DOE VS FMEA DOE VS Pareto Chart13Above 4 quality tools except DOE all could not give transfer function DOEThe Output of Screening DOEBrainstormingFMEAVital xsExplorationTrivial xsof the y-x relationshipFishbonePareto Chart14Screening out Vital Xs, and produce prediction equationx7 =
18、40%x6 = 25%x2 = 10%x9 = 4%x10= 4%x5 = 2%x1 x3 x4x8= 15%x11 x12x1 x2 x3 x4 x5 x6 x7 x8 x9 x10 x11 x12 DOECharacteristic & Benefit of DOETo identify the relationship between y and xs, a tool is need which :is Structured;is Proactive toward Data Collection; is Statistically Sound;Quantifies the y=f(x)
19、Relationship; andSegregates the Vital Few Xs.Identify optimized setting of Vital Few Xs Identify optimized response15 DOESteps of DOE1. Define the Problem2. Establish the Objective3. Select the Response4. Select the Factors5. Choose the Factor Levels6. Select the Experimental Design7. Collect the Da
20、ta8. Analyze the Data9. Draw Conclusions10. Run Additional Experiments, if necessary11. Achieve the Objective16 DOE“DOE” TerminologyExperimental Plan Factors (xs) Response (y) DesignLevel17 DOEKey ConceptsExperimental Plan: The formal plan for conducting the experiment is called the “experimental de
21、sign” (also the “experimental pattern”). It includes the choices of the responses, factors, levels, blocks, and treatments and the use of certain tools called planned grouping, randomization, replication.Factors: A factor is one of the controlled or uncontrolled inputs whose influence upon the respo
22、nse is being studied in the experiment. Factors are also known as the xs. A factor may be quantitative, e.g., temperature in degrees, time seconds. A factor may also be qualitative, e.g., different machines, different operator, clean or not clean.Response: The measured characteristic used to quantif
23、y the result of a combination of factors at given levels. The response will be one of the ys.Design (Layout): Complete specification of experimental test runs including factor-levels and order.Level: The “levels” of a factor are the values of the factor being examined in the experiment. For quantita
24、tive factors, each chosen value becomes a level, e.g., if the experiment is to be conducted at two different temperatures, then the factor temperature has two “levels”. In a qualitative factor, the single factor “cleanliness” has two levels: clean and not clean.18 DOETypes of Experimental DesignsScr
25、eeningFull factorial2kfactorialFractional factorialOptimizationFull factorialMulti-level experiments Composite designsRobust designsConfirmation19 DOEKey ConceptsScreening experiments can be used to quantitatively separate the vital few xs from the trivial many xs. These types of experiments are als
26、o known as exploratory experiments.The second phase of screening experiments might be to form a mathematical equation relating the y to the x. This step is called characterization.Optimization experiments seek the level(s) of the vital few xs that will optimize the performance of the y, with respect
27、 to targeting, variability reduction, or both.Confirmation experiments validate the transfer function within the range of experimentation. Further insight is also given on the noise estimate due to uncontrolled factors.20 DOEDOE compare with One-Factor-at-a-time methodExample - Consider a process in
28、 which the yield is a function of the temperature and the reaction time. From historical information, it is known that the process produces an acceptable, yet hopefully improvable, yield at a temperature of 155 degrees and a reaction time of 1 1/2 hours. An experiment is run to investigate the effec
29、ts of reaction time on yield using the traditional one-factor-at-a-time approach. Five runs are made producing yields as follows:Temperature155155155155155Time0.51.01.52.02.5Yield45%65%77%71%48%21Which reaction time would you choose? DOEOne-Factor-At-A-Time (continued)A second experiment is run with
30、 the reaction time held at1 hour and40 minutes. The one-factor-at-a-time approach is used as temperatures are varied to produce the following yields:Temperature140150160170180Time1.671.671.671.671.67Yield59%74%78%73%66%1. What temperature would you choose?2. Where would you run the process according
31、 to your experimental data?3. What is the best yield the process can get?22 DOEThe Result Get from DOE23TempContour Plot of Yield200190901801708516080150757014065601300.00.51.01.52.02.5Timethe same process, the same condition but different result! Why?60 65 707580 859095 DOEComparison of Experiment
32、TypesOne-at-a-TimePlannedNo guarantee of finding optimumInteraction of variables can produce incorrect conclusionsTime-consuming and inefficientOptimum is methodically sought, statistically verified and documentedInteractions of variables are incorporated into the designEfficient and effective24 DOE
33、Pitfall of one-factor-at-a-timeLarge number of observation Cant detect interactionEffects of factors cant be independently estimatedNot possible to test the significance of individual effectsProcess optimization is difficult25In what cases could use one-factor-at-a-timemethod? DOE“Wrong” Progress of
34、 an ExperimentE F F O R TL E V E LPlanningWork Like HellAnalysis26 DOERole of Experimental DesignThere is no unique route to the solution of a problem. Two equally competent investigators might usegreatly different routes when confronted by a problem, and yet reach the same answer.The qualities need
35、ed to play the experimental “game” well are:Subject matter knowledge Knowledge of strategyIt is possible for a scientist or engineer to conduct an experiment without statistics, but his or her intellect and subject matter knowledge are best used when effective statistical tools are available and use
36、d. The tools are:Efficient methods of experimental design, which allow the investigator to obtain answers which areunequivocal and minimally affected by experimental error. Sensitive methods of data analysis. These should make it clear what conclusions can legitimately be drawn from the data.Of the
37、two tools, design is the more important. From a good design, one can extract valid conclusions with simple and straightforward analysis. Often all the important conclusions can be eyeballed directly from the data. If the design is poorly chosen, even the most sophisticated methods of statistical ana
38、lysis will simply report that the experiment was lacking in information.27DOEIdeal Progress of an ExperimentPlanningPhase 1E F F O R TExperimentationAnalysisTimePhase 2PlanningE F F O R TExperimentationAnalysisTimeThe analysis of Phase 1 leads to (overlaps) the planning stage of Phase 2.28 DOEKey Co
39、nceptsIt is critical that the Team:Clearly define the objectives and scope of the experiment; Gather information needed to design the experiment;Ensure that resources will be available to complete the experiment(s).Use ample planning and sequential experimentation.29 DOEPitfalls of Experiments No Ob
40、jectiveIf you dont know where youre going youll probably wind up there.Out of control System - Process not StabilizedCannot measure Cannot control factors Too ComplicatedKISS(Keep it statistically simple) Too LongWhat else is changing while you do the experiment?30 DOEKey ConceptsNo objective: A sur
41、prising number of programs fail to observe one of the “laws” of human nature:If you dont know where youre going, youll probably wind up there.A well-defined experimental objective is similar to a good problem statement. It is developed using standard problem solving tools such as problem statement,
42、cause and effect diagrams and root cause analysis.A good objective:Avoids looking to the solution prematurely; helps to solve the right problem; focuses efforts?In any case, an objective should be:Unbiased (if you already know the answer, why are you running an experiment?); specific and measurable
43、(it should be clear when the objectives have been met).Above all, the objective should relate to some business action. (If not, why are you wasting your time running an experiment?)In some cases the experiment is being done to solve an existing problem; in other cases it is directed toward new oppor
44、tunities. The exact content of the objective will differ in those two cases.Out of control system: One of the purposes of the trial run phase of experimentation is making sure that the process and the measurement system are in some kind of control. Ideally, both the process and the measurements shou
45、ld be in statistical control as measured by a functioning SPC system. At the very least, process settings should be reasonably reproducible and the variability of the measurements should be known.31 DOEScreening DOE32 DOEKey PointsScreening experimental designs are used to segregate the vital fewxs.
46、 Properties of this type of design are:1. Little knowledge of the xs is required2. All possible factors (xs) are included in the initial screening design.3. Only two levels (high/low) are assigned for each factor.4. Resolution III and IV designs are used in order to capture main effects and interact
47、ion effects.5. Screening designs lead directly into optimization designs, concentrate on the best settings for the Vital Few Factors.which33 DOEFactorial Experiments1.Full factorial2.3.2k factorialFractional factorial34 DOEObjectiveFactorial designs form the foundation of most experiments. To be eff
48、ective, the user should:1.2.Understand the advantages of factorial experiments.Be able to consider all facets of the process in setting up an effective experiment.Understand the concept of statistical interaction. Be able to analyze general factorial experiments.Use diagnostic techniques to evaluate
49、 the “goodness of fit” ofthe statistical model.Be able to suggest any needed further experimentation.3.4.5.6.35 DOEKeyConceptsFactorial designs represent most experiments run.1.2.While all experiments are not factorial designs, understanding of these designs will assist in forming other designs that
50、 require special conditions.36DOEVarious Factor-Level Designs2 Factors2 Levels of Each4 Levels of Factor A 3 Levels of Factor B3 Levels of Factor A 2 Levels of Factor B 2 Levels of Factor C37What is common among these three designs?ABCABAB DOEBalance and OrthogonalityHighBLow(1)aLowHighASX=0 for eac
51、h factor sumiBalancedThis feature helps to simplify the analysis.SXX=0 for all dot product pairsijOrthogonalThis feature ensures the effects are independent.38(L,H)(H,H)(-1+1)(+1, +1)bab(L,L)(H,L)(-1,-1)(+1, -1) DOEKey ConceptsWithout balance and orthogonality, we lose the ease in analyzing our fact
52、ors.These properties are desirable but not necessary.Three ways of describing a given trial setting are indicated within the square1.2.39 DOECharacteristic of Factorial Design1.Any number of levels can be chosen for any number of factors.2.All combinations of the levels of each factor are run.3.Allow for the presence of interaction effects.40 DOEAn Incomplete Factorial De
温馨提示
- 1. 本站所有资源如无特殊说明,都需要本地电脑安装OFFICE2007和PDF阅读器。图纸软件为CAD,CAXA,PROE,UG,SolidWorks等.压缩文件请下载最新的WinRAR软件解压。
- 2. 本站的文档不包含任何第三方提供的附件图纸等,如果需要附件,请联系上传者。文件的所有权益归上传用户所有。
- 3. 本站RAR压缩包中若带图纸,网页内容里面会有图纸预览,若没有图纸预览就没有图纸。
- 4. 未经权益所有人同意不得将文件中的内容挪作商业或盈利用途。
- 5. 人人文库网仅提供信息存储空间,仅对用户上传内容的表现方式做保护处理,对用户上传分享的文档内容本身不做任何修改或编辑,并不能对任何下载内容负责。
- 6. 下载文件中如有侵权或不适当内容,请与我们联系,我们立即纠正。
- 7. 本站不保证下载资源的准确性、安全性和完整性, 同时也不承担用户因使用这些下载资源对自己和他人造成任何形式的伤害或损失。
最新文档
- 培训课件的互动游戏
- 人类活动与环境问题高中地理湘教版(2019)选择性必修3
- 接受双膦酸盐治疗患者拔牙围手术期处理 专家共识
- 刑侦学考试试题及答案
- 脑机接口康复训练系统设备研发进展与市场应用前景
- 《白杨礼赞》教学课件
- 电商平台安全管理体系与交易保障措施
- 煤矿章程备案管理办法
- 物业关于租户管理办法
- 物业废品处理管理办法
- 北京八中分班数学试卷
- 培训课件:血糖监测
- 康复医学科关于无效中止康复训练的制度与流程
- 工伤保险待遇申请表
- 甘肃省建筑安全员A证考试题库及答案
- 【艺恩】JELLYCAT品牌洞察报告
- 2025年中考物理终极押题猜想(广东省卷专用)(原卷版)
- DB36-T 2037-2024 地质灾害治理工程施工监理规范
- 《公路建设项目文件管理规程》
- 腰麻课件教学课件
- 2024年《治安管理处罚法》多项选择题题库及答案(共193题)
评论
0/150
提交评论