6分部试验设计.docx
《6分部试验设计.docx》由会员分享,可在线阅读,更多相关《6分部试验设计.docx(16页珍藏版)》请在冰豆网上搜索。
6分部试验设计
第六章分部试验设计法
第一节分部试验设计概述
第二节四因素二水平4级分部试验设计
第三节六因素二水平3级分析试验设计
第一节分部实验设计概述
一.分部实验设计的特点
1.比全因子试验设计需要的试验组合最少少一半以上。
2.同时可评估许多因素,因为大大减少了试验组合,就意味着同时评估的因素大增。
3.被选中的试验组合是从相对的全因子试验组合中仔细选择的有代表性的子集。
4.分部试验设计可用于为后续试验设计确定关键的影响因素。
5.信息少,精度差
二.因素间的交错或混淆
ABCABBCABC
121222第一组实验黄色区ABC的交互作用一样,无法区分开来
122211
221121白色区域A与BC无法区分是那个因子造成的影响,故白色区域
112122
111111中是A+BC共同造成的影响。
211212
222112
212221
三.分部试验设计的分辨率
分辨率
主要特点
3
主要影响与2因素交互作用混淆
2因素交互作用与2因素交互作用混淆
4
主要影响与2因素交互作用没有混淆(如果要区分交互作用影响与主要因素选择四级分辨率)
主要影响与3因素交互作用混淆
2因素交互作用与2因素交互作用混
5
主要影响与2因素交互作用没有混淆
2因素与2因素交互作用没有混淆
2因素与3因素交互作用混淆
主要影响与4因素交互作用混淆
第二节四因素二水平分辨率4级分部试验设计
1.试验目标:
取得塑胶轮的最大机械强度
2.试验计划:
轮的强度,》100N,测力计
3.影响因素:
A.模腔温度,B喷嘴温度,C.射出口径,D.注塑机吨位
4.水平设置:
因素
-1
+1
模腔温度A
120
180
B喷嘴温度
40
60
C.射出口径
5.8
10
D.注塑机吨位
750
950
考虑有交互作用,4因素2水平,5级分辨率要做16次试验,成本太大,4级分辨率做8次试验,可区分主要因素与两因素交互作用的影响.
FractionalFactorialDesign
Factors:
4BaseDesign:
4,8Resolution:
IV
Runs:
24Replicates:
3Fraction:
1/2
Blocks:
1Centerpts(total):
0
DesignGenerators:
D=ABC
DefiningRelation:
I=ABCD
AliasStructure
I+ABCD
A+BCD
B+ACD
C+ABD
D+ABC
AB+CD
AC+BD
AD+BC
DesignTable
ABCDY
-1-1-1-1123
1-1-11143
-11-11121
11-1-156
-1-111127
1-11-1156
-111-1134
1111156
-1-1-1-1134
1-1-11165
-11-1176
11-1-176
-1-111123
1-11-1122
-111-187
1111112
-1-1-1-1134
1-1-1189
-11-11112
11-1-198
-1-111125
1-11-1125
-111-1134
1111112
从上图可看出,拟合值较小时,残值也较小.
FactorialFit:
YversusA,B,C,D
EstimatedEffectsandCoefficientsforY(codedunits)
TermEffectCoefSECoefTP
Constant118.334.74224.950.000>0.05,影响显著
A-1.67-0.834.742-0.180.863
B-24.33-12.174.742-2.570.021>0.05,影响显著
C15.507.754.7421.630.122
D6.833.424.7420.720.482
A*B-7.33-3.674.742-0.770.451
A*C10.505.254.7421.110.285
A*D17.178.584.7421.810.089
S=23.2325R-Sq=48.22%R-Sq(adj)=25.57%方程模拟试验结果不好
AnalysisofVarianceforY(codedunits)
SourceDFSeqSSAdjSSAdjMSFP
MainEffects4529152911322.82.450.088
2-WayInteractions327522752917.41.700.207交互作用不显著
ResidualError1686368636539.8
PureError1686368636539.8
Total2316679
UnusualObservationsforY
ObsStdOrderYFitSEFitResidualStResid
181889.000132.33313.413-43.333-2.28R
Rdenotesanobservationwithalargestandardizedresidual.
AliasStructure
I+A*B*C*D
A+B*C*D
B+A*C*D
C+A*B*D
D+A*B*C
A*B+C*D
A*C+B*D
A*D+B*C
ResidualsvsFitsforY
Y最大,A-,B-,C+,D+
从上图可看出,AB,AC,AD有交互作用.
ResponseSurfaceRegression:
YversusA,B,C,D
Thefollowingtermscannotbeestimated,andwereremoved.
B*C
B*D
C*D
Theanalysiswasdoneusingcodedunits.
EstimatedRegressionCoefficientsforY
TermCoefSECoefTP
Constant118.3334.74224.9530.000
A-0.8334.742-0.1760.863
B-12.1674.742-2.5660.021显著因子
C7.7504.7421.6340.122
D3.4174.7420.7200.482
A*B-3.6674.742-0.7730.451
A*C5.2504.7421.1070.285不是显著因素
A*D8.5834.7421.8100.089
S=23.23R-Sq=48.2%R-Sq(adj)=25.6%方程模拟不良
AnalysisofVarianceforY
SourceDFSeqSSAdjSSAdjMSFP
Regression78043.38043.331149.052.130.100不显著
Linear45291.05291.001322.752.450.088
Interaction32752.32752.33917.441.700.207
ResidualError168636.08636.00539.75
PureError168636.08636.00539.75
Total2316679.3
UnusualObservationsforY
ObsStdOrderYFitSEFitResidualStResid
181889.000132.33313.413-43.333-2.28R
Rdenotesanobservationwithalargestandardizedresidual.
EstimatedRegressionCoefficientsforYusingdatainuncodedunits
TermCoef
Constant118.3333
A-0.8333
B-12.1667
C7.7500
D3.4167
A*B-3.6667
A*C5.2500
A*D8.5833
Y=118.333-12.17B
Y的最佳值Y=118.333+12.17=130.17
对偏差进行分析
先转化为8个组合
GeneralLinearModel:
sigmaversusA,B,C,D
FactorTypeLevelsValues
Afixed2-1,1
Bfixed2-1,1
Cfixed2-1,1
Dfixed2-1,1
AnalysisofVarianceforsigma,usingAdjustedSSforTests
SourceDFSeqSSAdjSSAdjMSFP
A11.21.21.20.010.915,四因素对偏差无显著影响
B153.553.553.50.500.487
C1159.2159.2159.21.500.236
D114.814.814.80.140.714
Error192018.52018.5106.2
Total232247.2
S=10.3072R-Sq=10.18%R-Sq(adj)=0.00%方程拟合不良
UnusualObservationsforsigma
ObssigmaFitSEFitResidualStResid
1533.234014.42124.704618.81282.05R
Rdenotesanobservationwithalargestandardizedresidual.
ResponseSurfaceRegression:
sigmaversusA,B,C,D
Thefollowingtermscannotbeestimated,andwereremoved.
B*C
B*D
C*D
Theanalysiswasdoneusingcodedunits.
EstimatedRegressionCoefficientsforsigma
TermCoefSECoefTP
Constant14.49281.9137.5750.000
A-0.22771.913-0.1190.907
B1.49241.9130.7800.447
C-2.57571.913-1.3460.197
D-0.78411.913-0.4100.687
A*B-3.43271.913-1.7940.092所有因素对偏差无显著影响
A*C-2.80341.913-1.4650.162
A*D2.42881.9131.2700.222
S=9.373R-Sq=37.5%R-Sq(adj)=10.1%
AnalysisofVarianceforsigma
SourceDFSeqSSAdjSSAdjMSFP
Regression7841.68841.68120.2411.370.284
Linear4228.67228.6757.1680.650.635
Interaction3613.01613.01204.3372.330.113
ResidualError161405.521405.5287.845
PureError161405.521405.5287.845
Total232247.20
EstimatedRegressionCoefficientsforsigmausingdatainuncodedunits
TermCoef
Constant14.4928
A-0.2277
B1.4924
C-2.5757
D-0.7841
A*B-3.4327
A*C-2.8034
A*D2.4288
试验结论:
A-,B-,C+,D+
第三节:
六因素两水平分辨率3级分部试验
1.试验目标:
金属部件两部分最大粘接力
2.测量指标:
两部件的粘接力
3.影响因素X’S
紫外线光亮,保压时间,固定夹具,胶水型号,点胶位置,点胶量
4.实验水平:
因素
水平-1
+1
A
600
1000
B
5
10
C
A
B
D
L1
359T
E
3
6
F
100
150
FactorialFit:
YversusA,B,C,D,E,F
EstimatedEffectsandCoefficientsforY(codedunits)
TermEffectCoefSECoefTP
Constant149.7511.1013.490.000显著
A2.751.3811.100.120.904
B31.2515.6311.101.410.197
C7.753.8711.100.350.736
D6.003.0011.100.270.794
E-55.50-27.7511.10-2.500.037显著
F3.501.7511.100.160.879
A*F33.7516.8711.101.520.167
S=44.4016R-Sq=57.39%R-Sq(adj)=20.12%方程模拟不良
AnalysisofVarianceforY(codedunits)
SourceDFSeqSSAdjSSAdjMSFP
MainEffects616690.816690.827821.410.318不显著
2-WayInteractions14556.24556.245562.310.167不显著
ResidualError815772.015772.01972
PureError815772.015772.01972
Total1537019.0
UnusualObservationsforY
ObsStdOrderYFitSEFitResidualStResid
42240.000175.00031.39765.0002.07R
1310110.000175.00031.397-65.000-2.07R
Rdenotesanobservationwithalargestandardizedresidual.
AliasStructure
I+A*B*D+A*C*E+B*C*F+D*E*F+A*B*E*F+A*C*D*F+B*C*D*E
A+B*D+C*E+B*E*F+C*D*F+A*B*C*F+A*D*E*F+A*B*C*D*E
B+A*D+C*F+A*E*F+C*D*E+A*B*C*E+B*D*E*F+A*B*C*D*F
C+A*E+B*F+A*D*F+B*D*E+A*B*C*D+C*D*E*F+A*B*C*E*F
D+A*B+E*F+A*C*F+B*C*E+A*C*D*E+B*C*D*F+A*B*D*E*F
E+A*C+D*F+A*B*F+B*C*D+A*B*D*E+B*C*E*F+A*C*D*E*F
F+B*C+D*E+A*B*E+A*C*D+A*B*D*F+A*C*E*F+B*C*D*E*F
A*F+B*E+C*D+A*B*C+A*D*E+B*D*F+C*E*F+A*B*C*D*E*F
E为主要因子
拟合值越小,残值越小.
Y最大,A+,B+,C+,D+,E-,F+
FactorialFit:
YversusA,B,C,D,E,F
EstimatedEffectsandCoefficientsforY(codedunits)
TermEffectCoefSECoefTP
Constant149.7511.1013.490.000
A2.751.3811.100.120.904
B31.2515.6311.101.410.197
C7.753.8711.100.350.736
D6.003.0011.100.270.794
E-55.50-27.7511.10-2.500.037
F3.501.7511.100.160.879
A*F33.7516.8711.101.520.167
S=44.4016R-Sq=57.39%R-Sq(adj)=20.12%
AnalysisofVarianceforY(codedunits)
SourceDFSeqSSAdjSSAdjMSFP
MainEffects616690.816690.827821.410.318
2-WayInteractions14556.24556.245562.310.167
ResidualError815772.015772.01972
PureError815772.015772.01972
Total1537019.0
UnusualObservationsforY
ObsStdOrderYFitSEFitResidualStResid
42240.000175.00031.39765.0002.07R
1310110.000175.00031.397-65.000-2.07R
Rdenotesanobservationwithalargestandardizedresidual.
PredictedResponseforNewDesignPointsUsingModelforY
PointFitSEFit95%CI95%PI
1175.00031.397(102.599,247.401)(49.598,300.402)
290.00031.397(17.599,162.401)(-35.402,215.402)
389.00031.397(16.599,161.401)(-36.402,214.402)
4175.00031.397(102.599,247.401)(49.598,300.402)
5177.50031.397(105.099,249.901)(52.098,302.902)
689.00031.397(16.599,161.401)(-36.402,214.402)
790.00031.397(17.599,162.401)(-35.402,215.402)
8175.00031.397(102.599,247.401)(49.598,300.402)
9182.50031.397(110.099,254.901)(57.098,307.902)最大,选择
10144.50031.397(72.099,216.901)(19.098,269.902)
11182.50031.397(110.099,254.901)(57.098,307.902)
12144.50031.397(72.099,216.901)(19.098,269.902)
13175.00031.397(102.599,247.401)(49.598,300.402)
14164.50031.397(92.099,236.901)(39.098,289.902)
15164.50031.397(92.099,236.901)(39.098,289.902)
16177.50031.397(105.099,249.901)(52.098,302.902)
预测方程.
Y=149.75-27.75E
ACDF设定后对BE进行全因子试验.因为交互作用无法识别.