Optimization of Multi-Response Mixture Experiments
to Achieve Six-Sigma Quality Level
Optimization of Multi-Response Mixture Experiments to Achieve
Six-Sigma Quality Level
Student
Bo-Ling Wang
Advisor
Lee-Ing Tong
Chung-Ho Wang
!
"
#
A ThesisSubmitted to Department of Industrial Engineering and Management
College of Management National Chiao Tung University In Partial Fulfillment of the Requirements
For the Degree of Master of Science In Industrial Engineering June 2005 Hsin-Chu, Taiwan Republic of China
!
(Design of Experiments, DOE) !" # $ % & ' () * + ,- ./012 345 6 47 8 $ 9 : ;< => (Mixture Experiments)? @, 5 A B CD E F G H I ,.1> 9 J K L M NOP (Q R> S (component) TRH I (T S )U VW ,XY 0> ,Z [F G H I \ ] ^ _.`a 1b c d e f g [h i j k lm n o p (q rs ,t uv v k * w x ;y zZ({ | 5 } t ~ o O (desirability function) > ,; z Z{ | 5 (A B. Ribardo[ Allen[14]o w x [P , ( u \ (desirability function for
achieving six-sigma quality)! o - > ,;y zZ 5 ( ` u \ NO u \ , . ¡ S ¢ £ ¤ ¥ ¦ § o ¨ ,© ª « ¬ ® ,Y ¯ o ,° ± Z[ Z.
Optimization of Multi-Response Mixture Experiments
To Achieve Six-Sigma Quality Level
Student
Bo-Ling Wang Adviser
Lee-Ing Tong,
Chung-Ho Wang
Department of Industrial Engineering and Management
National Chiao Tung University
Abstract
Design of experiments (DOE) is useful in finding the optimal parameter-setting efficiently with low experimental cost. However, in some specific areas, engineers often misuse factorial design on mixture experiments. In mixture experiments, the factor (which is also called component) affects the response by its proportion in all components, but not the actual volume. This is the main difference between factorial design and mixture experiments. In addition, simultaneously optimization of multiple responses is increasingly essential. Therefore, this study proposes a procedure to optimize multi-response in mixture experiments. The desirability function of six-sigma quality introduced by Ribardo and Allen [14] is employed in the proposed procedure. The desirability function considers the estimated mean and variance of response
variable and estimate the response under assumptions required in the six-sigma quality. A rubber bowl’s experiment with four components and eleven response variables are utilized to demonstrate the effectiveness of the proposed procedure.·
·
Key Words: Mixture Experiments, component, multi-response, desirability function for achieving six-sigma quality
4.3.3 R ,V ...24 4.3.4 V ,NO f ...24 4.3.5 e...28 4.3.6 ð Z ...28 4.3.7 ...29 4.3.8 Y¯ ¼ ...30 xy ¼ 5.1 ¼ [^ ...31 5.2 ...32 w v
`2-1 2×2³ H I ...3 `2-2 RH I ,XY...3 `2-3 u \ ,´ `kµ …...12 `4-1 © ª « ¬ ® , u ...20 `4-2 « ¬ ® ,> `...20 `4-3 NOP H2[H3, A ...21 `4-4 NOP H1, [P ¶ ...22 `4-5 NOP H11R s, ...23 `4-6 À· NOP 1R s, ...23 `4-7 R s,V ...24 `4-8 V Û ¸ ¹NO f ,ANOVA`...25 `4-9 V º 5 ¹NO f ,ANOVA`...25 `4-10 e¯ ° `...28 `4-11 e A ...30 `4-12 e [¯ ° X» `...30
1.1
1% & b¼( $ 9 [ !o p $ % & ' ,) * + . `a 1b c d e f g [h i j k lm n o p (q rs ,½ ¾ ú , ¿ ,o « !{ |Ú;y zZÀ h i j ,k l w x ; , 5 ½ Á  [ " # $ % & ' .¢a r  (off-line quality control)Ãí ¢ o p H ` g ± e Ä T!¢a r  ,) * Å B.G !Æ Ç ((Design of Experiments, DOE)[N O f (Response Surface Method, RSM) ° ¹ º 6 ¸ ì 1½ ¾ ( 9 È É " # $ % & ' .Ê õ $ ËÌ Í O [NO f ÀÎ ÏÐ RËÑ .
/0 1: ;2 345 6 7 8 $ 9 D E F G H I l e }CL K ߯ Ò , ÀÓ H !» Ô $ À , [F G H I , d { !> (Mixture Experiments),? @.Õ > Q Å,NOP (response variables, T zZ)Ö J K Ð H I ,¦W ;× L M 0Q [RH I U Ø V¦W ,XY³ .1 > 9 æ ,H I XY!t ( Ø H I 1> 9Ù ! S (component) 0å ò=F G ,H I 1> 9 ( Q R S U VW , S XY.> ,Z [F G H I \ ] å æ _ ¼ z] ñ ò> g ± . `a i= Ú ; zZA B(k l_!" # Ê õ Û /Ü Ý : ; ; zZA B, }1> ( 9 Þ ;Q ¼ - zZ,A B Dabbas8¸ [10]1 ¼ ß> Z ,÷ Æ à á 6 ¥( 5 - O (desirability function) ; zZ,A B } !å òÅu0Ö â òÅu L J K !$ Ë1 -, ã ¼ ò`o g e.
ä Ü u \ (six-sigma)å G 1D æ ç è é ê (Motorola) Î ë ± : ; ì $ Tí l u \ , !uî o p $ % & ' ï`a 1[ð ñ 4 ò ó c 8 $ 9 1ô ³ 8H õ w W ,s ö X- ÷ $ * l« : ; 0 u \ å G 1`Ô $ 9Ïô ø ) * .H ` ¼ ùRibardo[Allen[14] o , u \ (desirability function for achieving six-sigma
quality)O=; zZ> 5 ,A B- NO f l S XY o ¨ ß u \ Z â òÅu, e.
1.2
ú =t ~ û L Û ¸ ,; zZ> 5 ã ¼ , Ã* t (!ñ ò> A B Ü Ý - ; zZ, 5 .Ø !ß u \ ,; zZNO f ¯ ° 0Ø l ,> e° - ß u \ Z,â òÅu ° o ¨ 6 è !Ào ü , w .
1.3
¼ u \ ùR s(;y zZý ° ß u \ ,V ùR , u \ V Ó > ,NO f .¡NO f e þØ eùß- y u \ ,V ° = u \ , .
1.4
¼ S y x- y!z ¼ ¼ , } ~[t (ïx y ¼ ,å ³ > 4; zZA B u \ ,å ³ ïx4y ¼ Ø ,; zZ> 5 § ï x6 y!Y¯ © ª ª « ¬ ® , Y ¯ ¼ o ,e ° ± ïxy¥!¼ , ¼ [Ø Ù.
y > Àæ N aå ³ K s Q ; zZ ,å ³ Ú, u \ ¡¥Q ; zZ> ,å ³ .
2.1
[9]
@{ x- y { | [} ~ 9¥ H I À > ,zZ | kT» P ´ `(Q S XY ¥OÚ> 0Ö F G , H I . `2-1`2-2!Y[9] A4BÑP !ß> S zZ,P j Ú³ H I @`2-1 ù`2-1£ ÑP ,X Y @`2-2.¥À9³ H I 9, (A=25ïB=225)Ð(A=30ïB=270) , S XY -Q å { ( Àß> S ,zZ d { XY,ÑP S ½Q L M NOP (Ó H ¥OÚ> 0Ö ³ H I . A B 25 30 A B 25 30 225 225 1.0: 9.0 1.2: 8.8 270 270 0.85: 9.15 1.0: 9.0 ä P ß> S zZ ¥ µ xi(i=1 ,2 , ,q)!Ø S U > VW , XY xißszZ q i xi 1, 1 ,2 , , 0≤ ≤ = (2.1) 0 . 1 2 1 1 = + + + = = q q i i x x x x (2.2) i=(2.2),¡Â Úä P ,° ± - y ú .4 S ,> (q=3)!Y w2-1(e Q F G 4H I ( 0> ` 2-1 2×2³ H I ` 2-2 RH I ,XY1" # ¡Â x1+x2 +x3 =1.0(q rs £ ¤ w9e 9( 4 f @w2-2. 1> 9 q D Q Ø R S =NOP ,NO f ! e,e.¥ñ ò- µ , S P Ø (NO f s 4 * (1) * æ ,NO f ï(2) Ø ,ð Zï(3)À ¡ , 3 æ , Ø À ,ð Z[9]. 1` µ NO f @s NO ) , , , (x1 x2 xn φ η= (2.3) 1NO f 9 Ú, !xi(¢Z;, @(2.4)(2.5) = + = q i i i x 1 0 β β η (2.4) ≤ = + + = q j i q j ij i j q i i i x x x β β β η 1 0 (2.5) i=(2.2),¡Â ° ù- _ 9! " 5 ! 0 @ + + + = ≤ ≤ ≤ ≤ = q k j i q k j q k ijk i j k q j i q j ij i j q i i i x x x x x x β β β η 1 (2.6) `a ÀÏ° # 1¦xi −xj(i≠ j)(P å .V0$ , (2.3)° k l% æ , . (1,0,0) (0,0,1) (0,1,0) (1,0,0) (0,0,1) (0,1,0)
w 2-1 4 S > [4H I [9] w2-2 4 S > [9]s° ± & ' ,Û ¸ [( S Ñ) q ¨ @* æ ,NO f . Û ¸ ,° ± & ' ¥ÚD + ,
{ }
q , m [16](SimplexLattice Designïm Åxi S XY,S V )q, S 9 [17](Simplex
Centroid Design) .w2-3Q ù - S ² = S ,° ± & ' ( ? . N ïw2-4¥Q 1 S ,° ± & ' (/ [ 9 -0 1 À å òO,NO f .1 æ 3l Rβ 2 ¡ T° K ß- > S ,NO f . `a xi,? . 1¥ z3 4 sÖ Ü 0ß1 0Q 5 (2.7), q i U x Li i i 1, 1 ,2 , , 0≤ ≤ ≤ ≤ = (2.7) (2.7)9,LiÐUiS ] !xi,s¡[-¡ Àq D Q j 6 $ ã ä â Ú ( æ ¥ S XYxi7 10~LiUi ~1,|d L g ± 8q r à 89 Ú(2.7) !xi(° ± ? . . `| ° : * i x 4xi#!xi¢Z£ ¤ ¡,£ P . 0≤Li ≤xi ≤1 T S X Ys¡,¡Â ¥: L L x x i i i − − = 1 * ° Ú0≤ * ≤1 i x À9 = = q i i L L 1 @w2-5ï 0≤xi ≤Ui ≤1 T S XY-¡,¡Â þ min 1 1 ≤ − = i i q i i U U ¥: 1 # − − = U x U x i i i ° Ú0≤xi# ≤1 À9 = = q i i U U 1 . w 2-4 3- S 9 [8] 1 x (1,0,0) 3 x (0,0,1) 2 x (0,1,0) ( 3 2 ,0, 3 1 ) ( 3 1, 3 1, 3 1) 1 x (1,0,0) 3 x (0,0,1) 2 x (0,1,0) (0, 2 1, 2 1) ( 2 1 ,0, 2 1 ) w 2-3 {3,3} [8]
/0 1L g ± ¢Z£ ¤ 0ÚÀ£ S ,? . !0ß 1| Ú° ± 5 d ¥; 0Ö Û ¸ @w2-6ïQ # 1À· ¡Â @ 0 1 ≥ = q i i x H | À9H∈ q.¥`| 9 L Ú 8 9 < = > º (Computer-Aided Design) . ? @ (< = > º Q 1 NO f ¡ 1° ± ¥ Ó ¥A æ , . ¥ > NO f !y =x + À9 ) , ( ~ 2 n n n iid N 0 σ I ¥iF e , _ 2 ,¶ ˆÏB Ü D ® ) ) ( , ( ~ ˆ 2 1 1 − + σ x'x p N (p≥q) 0yˆ =xˆ ~ N(x ,σ2(x'(x'x)−1x)) D ® d ¥ˆÐyˆ,P C D Þ !σ2(x'x)−1Ðσ2x'(x'x)−1x.D + ,< = > º 6 ) 1. F 5 ˆP C D σ2(x'x)−1,± !t (,D- 5 ÏT F 5 det(x'x)−1. \ À D ® ¥`Ï° F 5 _ 2 ˆ ,E F G & ( H . 1 x 2 x x 3 1 * 1 = x 1 * 2 = x * 1 1 = x w 2-5 s¡ S ,¢Z£ ¤ 1 x 2 x x 3 w 2-6 d ¥,° ±
2. F 5 ˆP C D σ2(x'x)−1,ò ¢Ð!t (,A- 5 ÏT F 5 tr((x'x)−1).Àt (1=F 5 _ 2 0 ˆ b = , P . 3. F 5 yˆ,_P !t (,G- 5 : x x x' x' 1 2 ( )− σ = = pp p p c c c c c c c C 0 11 10 0 01 00 cii∈ i=0~p ¥T!F 5 max(cii i 0, ,p) i = . 4. F 5 cii!t (,V- 5 . - 6 ) D- 5 D Ð Ú.} -1{ - NO f s À9- ) , A 1À· 4) 9 Ï Ê I =Ø 4) , A [9]. > ,) 1=> S , }d `P ß> S zZ.1> 9 Ï° # 1- d Q > S 01Àd { sÞ ò NOP J L M ,P `Ô P Ù ,!§ P (Process Variables).
K ß`Ô A B| ° J K òÓ > g ± J !- y £ k T1Ó ,q S > 9# L n y § P .q=34n=2!Y À9 1 0≤ xi ≤ i=1 ,2 ,3 (2.8) 1 ± = l z l =1 ,2 (2.9) 9 !Y ° K (2.10), 3 2 1 123 3 2 23 3 1 13 2 1 12 3 3 2 2 1 1x x x x x x x x x x x x SC β β β β β β β η = + + + + + + (2.10) 0w x ѧ P ¥° K (2.11), 2 1 12 2 2 1 1 0 z z z z PV α α α α η = + + + (2.11)
* > Ñy £ ( _ 2 γτλ =αλβτ À9λ∈
{
1 ,2 ,12}
Ð{
1 ,2 ,3 ,12 ,13 ,23 ,123}
∈ τ S ] òO§ P η Ð 9 PV η ,2SC .¥Ø £ @s < = = = < = + + + + + = + + = 3 3 2 1 12 2 1 0 3 1 1 2 12 2 1 0 3 2 1 123 3 3 3 1 ) ( ) ( ) ( ) , ( j i j l l ij i j l ij ij i l l i i l i i j i j ij i j i i i x x z z z x z z z x x x z x x z x z z x γ γ γ γ γ γ β β β η 12 1 2 1 2 3 123 2 1 123 0 123 z z z xx x l l l + + + = γ γ γ (2.12) À° ± @w2-7 1(2.12)9,¸ Ï° s ` = = < < = + + + + + = + + + = 2 1 123 1 2 3 3 3 1 3 2 1 0 123 3 0 3 1 0 2 1 12 2 2 1 1 0 ) ( ) ( ) ( ) ( ) , ( l l l j i i j l ij i i l i j i ij i j i i i z x x x x x x x x x x x x z z x z x z x x z x γ γ γ γ γ γ α α α α η 2 1 3 2 1 12 123 3 12 3 1 12 x x x x x x z z j i ij i j i i i + + + < = γ γ γ (2.13) +1 +1 -1 -1 z1 z2 w2-7 (q=3, n=2)¦§ P ,> ( 9 )À9(2.12)TòOw2-7,M w 0(2.13)òOw2-7,N w Ñj (kµ O å { .P ° ± & ' d ¥ Ï° < = > º =¦§ P ,> .
2.2
Î Q R [1] ò§ S 9T U V ú [À~ W (VOCs[ NOx),³ 2 g ± T U X W Á  Y Z , ÀÚ> ,VW L M NOP , Cornell[9] (> -VW .ÀNOP !T U V ú ! - zZ,> A B. [ \ º [5] 4) ~] ª ò< ^ _ < [0 < É ,L M i=» 4) ~ ] ª ,XY!L M _ < [0 < É ,H õ ã À% > e.ÀÏ< - zZ,> A B. C ` a [4] @* ùß> zZ,P b L (2k)c é J p ` ÀÚXVERT ¡y , Ü 9A- 4D- 4 SNX-" d A 0 L J p `, e f !¡,J p `.Àt ( 1ùß> S ,P 0 L c é J p ` g ± ¡n S ¢ . : g h [6] @* 1ßd Z,Scheffé¢Z4 ¹L p i ,4¹ 9 g ± j k ,D- .l m | [3] @* 1ßd Z,Scheffé¢ Z ¹ 9 g ± j k , A- .À 1 @* " # < = > º , j k Z. Guo8¸ [12] n o 8> S p n q r Z, A À ¦§ P ] s [( t u ÀÛ /{ | y NOP , 5 }Q i v (Trade-Off),eg ± ¼ (d - Q , e. Sandoval-Castro8¸ [15]> 4) w 3x 3,d { XYy X z { | @* L M h 5 ( A (| W .ÀNOP Ïd } - }À t (Ö ` e 0Q iRNOP ,NO f Rw 3 x 3J ,> A . æ aÞ : ;> å ³ , };d Q ; zZ,A B 1`d ~ .
2.3
[2][13]
1 ; zZ,A B- D + Q 6 $ k+ ! 1L J K e f ,Á Â H I 9 i v ,e e./0 1NOP ;þÐ Á Â H I ,³ 2 ½ ¾ | i v ,ed þÛ /Q 6 $ k+ ¸ 1 Ú Y | Þ / : ;d Z Æ ¹ , e d { .æ NOP ; ¸ { | , Ï¡ , e[ ð , e,\ " q D ^ _. H ` ; zZA B, ¿ , .» J ý ,e!Ö× y P - y ) ` ) - y VNO.1 » y VNO| Ú Ó A B,¦ Ö , ) NO /0 1» ½ ¾ (A B- Ö æ , ) d Q (ó .
1: ;;P W eÐ Ü Ý ¡ Ã Ì Í Ð O1; zZ,A B-.@Ã S S ¢ (Principal Component Analysis, PCA)4 3 (Data Envelopment
Analysis, DEA)4 ë ¼ (Fuzzy Logic)4o ³ ES ¢ (Grey Relational
Analysis)4 Ô I ú ¾ Ä : (TOPSIS)4Ô â (Neural Networks)4 8 ÀÓ ;!1Ö× ) ) ¡ æ , l ° J K e f ,VNO.
% Ribardo[Allen[14]o , u \ ` g ,Åu i å X» , o ¨ - À u \ ,Åu Ú c [ ' å ß u \ ,; zZ, S XY.s!Ø , .
2.4
[7][14]
Q zZNOòÀt u( ú Ø , Q ÅN O1° K ,? . N K Ê t u, ú [7].0 u \ !Ê õ r ,) * 1 - Q ùNO, P L w W ÚÀV ß u \ , ° J K i u \ V Ø ;¹) ½ ± ¡,j k Z.Àe @s
Ribardo[ Allen[14]o , u \ zZP ,Z ° S ! Ë¡[L Ë¡Ñ) sS ] òÀg ± 1. zZP # 1 , Ë¡ zZP # 1 , -s¡ ¶ Ø P NO f , µ [P σ ¡ ¥1 τ =sσ (q rs µ u \ d(µ ,σ,τ)@s
[
( , , ), ( , , )]
min ) , , (µ σ τ = Y µ σ +τ Y µ σ −τ d (2.14) À9Y(•)`1 % Ò j ,q rs : W τ |,ö À¦ @s + − Φ + − Φ − + − Φ − + − Φ = σ σ µ σ σ µ σ σ µ σ σ µ τ σ µ ) ( ) ( 1 ) ( ) ( ) , , ( s U s L s L s U Y (2.15) (2.15)9,Φ(•)!u D ® S ,¡ H ~ U[L¥S ] ! zZ P , -4s¡. * ° NO u \ , Àkµ T!1 j | 1.5¢ u \ !° K ,E \ s( ö . 2. zZP L , Ë¡ æ zZP d # 1 , -s¡ 0t u[\ ° K !£ W NO: ; u | µ u \ @s) 5 . 1 ( ) (Y =Φ zeff − d (2.16) (2.16),Φ(•)!u D ® S ,¡ H ~ 0 − − + = M T M Y zeff 1.5 3 À 9Y! zZP ,NO4M ! zZP ,\ ° K 4T! zZP ,t u. , UL# 1 , Ud # 1 , Ld # 1
¥- ¦ ,R zZP ( ¥V D(x) !
[
]
[
]
{
w}
S m m w r r w r r r d r d m d d D(x) 1(Y1(x)) (Y (x)) 1(Y 1(x)) +1 (Y (x)) 1 + + = (2.17) À9wi! ) .1D(x)91ßr!# 1 Ë¡, zZP r+1ß xm !Ãý Ö t u, zZP .1 u \ τ =1.5 (q rs À ,w¨ @w2-8.`a Ribardo[Alleno , u \ ,´ `kµ @`2-3.
kµ ¤ 0.9999966~1.00 a ß u \ , L 9 g ± 0.9938~0.9999966 a ß6 ß u \ , f ö }Ö ¥ , 0.9332~0.9938 a ß4ß6 u \ , ° K }Þ 0.69~0.9332 a ßÑß4u \ , d ° K , 0.00~0.69 Ñu \ s, Û ³ d ° K
2.5
¦§ ³ =; zZ> ,å ³ å æ ¨ Dabbas 8¸ [10]O > =÷ Æ à ,á 6 ¥ 5 Àt (1= æ XY, © á 6 ¥.Ø Ú ù;y NOP - Åu NO
`2-3 u \ ,´ `kµ [14] w2-8 u \ w¨ 0 0.2 0.4 0.6 0.8 1 0.5 1 1.5 1.5 1 0.5 3 1 = σ U L i µ ) , (µ σ d 6 1 = σ
f e.Ø w x § P þÀÚ,Åu!å òÅ u d @ u \ ,â òÅu.
Smith[Cornell[18]Ú w(Biplot) { |`NOP [> S P ,³ 2 . wQ Ý zZ,å ³ ZRH I òRNOP , L M A ª ° ¹ º ¸ ì « ¬ Rä P [OP (³ 2 } B e,l .
ßt ~ !® > ,A BL Ø - Û ¸ , ; zZA B( þH O u \ r ,8* Z ¼ o - ° N O u \ ,; zZ> ( 5 . Ø ,; z Z> 5 § ¦4y ¯ ° .yRF ùÄ R¯ ° , ± ± ². g ± > , ï u \ S ¢ , eï¡g ± ð Z ¯ Z e ,° ± Z.
3.1
F ùÜ 8* ,P ³ ´ ß æ ,> .
3.1.1
6 $ ¸ ì ,â Ú R> S ,° ± & ' § P , NO P ,t ut u& ' 4-s¡° K ? . 4[å ò ) À· å ³ .3.1.2
ä P ,Z æ ,> 1. S P ßÛ ¸ ,° ± & ' ¥ ° K ,;¹Ú 9 æ , η=φ (x1 ,x2 , ,xn) . 2. S P L â û P £ ¤ ¦Û ¸ ° ± & ' ¥ ° K , ;¹F 9 ÚD- 5 æ , ) , , , (x1 x2 xn φ η = .
3.2
F @* u \ S ¢ , l eÀV .3.2.1
3.1.2 , ß Ë¡, zZP ( P À¶ @s ) ( ) ( ˆij xi =E ηij (3.1) ) ( ) ( ˆ2 ij i ij Varη σ x = (3.2) À9ηij =φ(xi) xi!xi ¹, e ηij!xi ¹, eòO=xjy N OP ý ° .
3.2.2
NOP # 1 , Ë¡ þ 1 sj¢ u \ (q rs 3.2.1 ,µˆij(xi) ˆ ( ) ˆ2( ) i ij i ij x σ x σ = ° RNOP 1R s, dij(xi)
[
(ˆ ( ), ˆ ( ), ), (ˆ ( ), ˆ ( ), )]
min ) ( i ij ij i ij i j ij ij i ij i j ij Y s Y s d x == µ x σ x + µ x σ x − (3.3) :µˆij =µˆij(xi)σˆij =σˆij(xi) ¥À9 + − Φ + − Φ − + − Φ − + − Φ = ij ij j ij j ij i j ij j ij ij j ij j ij ij j ij j j ij ij ij s U s L s L s U s Y σ σ µ σ σ µ σ σ µ σ σ µ σ µ ˆ ) ˆ ˆ ( ˆ ) ˆ ( 1 ˆ ) ˆ ˆ ( ˆ ) ˆ ˆ ( ) , ˆ , ˆ ( (3.4) , UiLi# 1 , Uid # 1 , Lid # 1(3.4)9,Φ(•)!u D ® S ,¡ H ~ Uj4LjS ] !xjy zZ P ,-4s¡. æ NOP d ß , Ë¡ # 1t uTj[\ ° K Mj ¥ RNOP 1R s, dij(xi)! ) 5 . 1 ( ) ( ij =Φ eff ,i ,j − ij Y z d (3.5) (3.5),Φ(•)!u D ® S ,¡ H ~ 0 − − + = j j j ij j i eff T M M Y z , , 1.5 3 .· * ß u \ , :sj=1.5 ¥Ø ° u \ , @`2-3.
3.2.3
3.1.1 ,xjy NOP å ò ) wj3.2.2, ° R ,V Di(xi)
[
]
[
]
{
w}
S i ij im w i ij r i w i ir i i i i d d r d Y r d Y m D(x ) 1(x ) (x ) ( 1)( (x )) +1 ( (x )) 1 + = (3.6) À9 = = m j j w S 1 ð 5 V .3.2.4
ù3.2.3 K ,R ,V Di(xi) =3.1.2 , ) (x φ η = Ø ,NO f .
3.2.5
i3.2.4 ,NO f ; zZ> , e ° K - y u \ ,â òÅu.3.3
F òS ¢ A g ± ð Z ¯ Z Q ( Î .
3.3.1
3.1.2 , g ± ,ð Z .ð Z µ ° S !Ñy ¶ S ,· æ Z , Z.~ j ¦NO f \ ,D ® 48P ¸ Z ý ¹ \ ,D ® ~w[¯ ° ò \ ,± ² wQ ( D ; ® g ± .¡j ¥Q 1 ¯ ° Ã 89 ? Z, À A g ± ú (lack-of-fit test). q û ð Z ¥k g ± 3£ ¤ ) £ ï N, ¥g ± s- ± ².3.3.2
! K , e1 º |ß Z 89 g ± . ,g ± ¹ù 8H õ Ú , A [¯ ° å \ ñ ; ¥O) £ æ ,ä P [NOP ïN, ¥Ø eß Z ¥Ø e ° ± . , e§ @w3-1.yS ¢ £ ¤ ¥ ¦ § o ¨ ,© ª ª ® (> Y x 4yo ,; zZ> 5 ( § ¯ À Z.
4.1
« ¬ ® Q © ª ª å G 9() * » i=1Úû 9 â D ßê ¼ [½ ¾ 8a' L M þ% Ò [ ª å G 9,¿ À Á 8Â 3K Ã ÚÀÄ Z_Å . Æ « ¬ ® , e ÚÀ1- Ç È É 9 % | ð D Ü Ê Î . 6 þ â L M « ¬ ® ,Ë ú [Ä Z8Z ,Ã* H õ ! û 9# L (ç Ì ª 4§ } º ª [ ¬ ª 85 Ý 3) Ô [Í W .i= ó Î â æ (5 Ý 3) Ô ã Y,t (1= « ¬ ® û 9 R5 Ý 3( Ï W XY. « ¬ ® , - ³ ´ 89 ù æ XY,« ¬ Ó 3[5 Ý eÏ L « ¬ ~(@ v w- ) T° z e,« ¬ 1 (@ v w ) z ~ Ð ½ ¡ « ¬ ® , . ¡ * Ø e,« ¬ ® Q ( 1 ª å G 9% |ð D º 89 â û - å (° ê .YÑ ê ~(@ v w4)g ± « ¬ 1 ,R) ° `a ! Ò % |Ú1 ª å G 9Ú¡(È É S ] ç 5 Ó (@ v w6 )[Ä Á Ó (@ v w)g ± # É ç 5 Ä [Ä Á # É ç 5 Ä Ñ ê ~ ° Ø « ¬ ® 1% | =Ç 6 È É 9 Q ( Þ ð D º . â Ô ¶ u Õ ,© ª Á ½ ª Ö ,« ¬ ® ° u Y , zZ× ú 4Ø ê Ë ú 4Ù % 4½ ¾ î Ú P ¨ H P 5 8 D ® 4ç 5 ZÄ Á Z8d { q r - zZ° S !R - y NOP RNOP Ë¡[t u@`4-1.NOP t P Û Ù s¡ t u -¡ \ ° K × ú (Hs) H1 65 70 75 — Ø ê Ë ú (kgf cm2 ) T1 — _ — 100 D ® Ù % (%) E1 — _ — 200 × ú P 5 (Hs) H2 -5 0 5 — Ø ê Ë ú P 5 (%) T2 — 0 — 40 ç 5 Z Ù % P 5 (%) E2 — 0 — 40 ½ ¾ î Ú P ¨ ½ ¾ î Ú P ¨ (%) C — 0 — 30 × ú P 5 (Hs) H3 -15 0 0 — Ø ê Ë ú P 5 (%) T3 — 0 — 40 Ù % P 5 (%) E3 — 0 — 40 Ä Á Z H P 5 (%) V — 0 — 20 YÆ R5 Ý 3, Ï W i=L M « ¬ ® zZ(Q R 5 Ý 3,Ï W XY 0Ö R5 Ý 3,Ï L W H `À > ,z Z OØ > s!Y , .
4.2
° òO¦ e,¯ ° - . 6 þ â e > , S ç Ì ª (x1)4§ } º ª A(x2)4§ } º ª B(x3) ¬ ª (x4) Ø 6 ) 5 Ý 3U « ¬ ® V) W (16.7%.`a 6 þ Ü â e - 6 ) S ,XY ¡Â S ] !6.7%~16.7%413.3%~25%413.3%~25%433.3%~66.7% i= S P ,° ± & ' d Û ¸ Y% D- 5 g ± @` 4-2. RUN ç Ì ª § } º ª A § } º ª B ¬ ª 1 0.167 0.133 0.133 0.567 2 0.1135 0.25 0.1915 0.445 3 0.06 0.1915 0.25 0.4985 4 0.06 0.133 0.195 0.612 5 0.167 0.25 0.133 0.45 ` 4-2 « ¬ ® ,> ` 4-1 © ª « ¬ ® , u6 0.167 0.25 0.25 0.333 7 0.167 0.1915 0.1915 0.45 8 0.167 0.133 0.25 0.45 9 0.06 0.25 0.133 0.557 10 0.167 0.133 0.1915 0.5085 11 0.067 0.133 0.133 0.667 12 0.1042 0.1812 0.133 0.5816 13 0.06 0.25 0.133 0.557 14 0.167 0.25 0.133 0.45 15 0.06 0.1915 0.25 0.4985 -`,R ¹9 À9ѹQ ! g ± ¡n , ú 04¹ ) ½ ¥Q ¶ ,Ý E \ ÀI R ¹!D- 5 Ó ¥ ( .1` RNO f ηj =φj(x1,x2,x3,x4).
4.3
° òO¦ e,¯ ° [¯ ° 4.4.3.1¦4.3.5RF ,S ¢ Ä l S < =¯ ° ,§ ï4.3.6 [4.3.7F ¥! ð Z [ Z¯ , A ¥< =¯ ° 4,§ .4.3.1
!
!
!
!
"
"
"
"
#
#
#
#
$
$
$
$
%
%
%
%
!
!
!
!
&
&
&
&
'
'
'
'
!
!
!
!
i`4-1° ã NOP ,9H14H2[H3 , Ë¡ 3.2.1 89 ¶ À1R esNO, [P .µ ý ¹ A ° Ü H2[H3,NO L ø > \ þO1 Ë¡N @`4-3. Run H2 H3 1 2 -8 2 0 -9 3 0 -7 4 0 -7 5 0 -8 6 1 -8 7 0 -10 8 0 -8 `4-3 NOP H2[H3, A
9 2 -7 10 1 -9 11 0 -8 12 1 -6 13 0 -7 14 2 -8 15 2 -8 i`4-3° + â û 6 þ â e , 1NOP H2[ H3( ` Î Þ =j ã 1Y,S ¢ 9 d ñ ò» Ñy P g ± S ¢ .1` k ¶ NOP H11R s, [P .
ÚDesign Expert NOP H1, NO f A @s
) 8706 . 0 ( 12975 . 80 94636 . 65 42458 . 71 38715 . 34 ˆ 2 4 3 2 1 1 = + + + = R x x x x H η (4.1) (4.1)ø i=p i òNOP ,L M d ø > ã Design Expert e , NO f !¢Z ¥ (4.1)° ¶ R , [P @s
4.3.2
!
!
!
!
"
"
"
"
#
#
#
#
14.3.19¶ ß Ë¡,NOP ( [P ¡ g ± 3.2.2,± ² S ] R s, . Run µˆH1(ηˆH1) σˆH1(ηˆH1) 1 69.44656 0.877965 2 70.04555 0.982610 3 72.17231 0.898376 14 68.42805 0.920586 15 72.17231 0.898376 `4-4 NOP H1, [P ¶ 1. H1, i=H1# 1 -4s¡ ã ´ L (3.4) × + − Φ − × + − Φ = 1 , 1 , 1 1 , 1 , 1 , 1 1 , 1 , ˆ ) ˆ ˆ ( 65 ˆ ) ˆ ˆ ( 75 H i H i H H i H i H i H H i H i s s Y σ σ µ σ σ µ (4.2) i= Æ ß u \ , ã :sH1=1.5 S ] Yi,H1(µˆi,H1 ,σˆi,H1 ,1.5)[Yi,H1(µˆi,H1 ,σˆi,H1 ,−1.5) ¥° K ßNOP H11 R s, di,H1(xi) @`4-5. 2. ? H14H2[H3a ÀI S y NOP d ß , Ë¡ # 1 t uTj[\ ° K Mj H `(3.5) À1R s, dij(xi) A @`4-6. Run Yi,H1(µˆi,H1 ,σˆi,H1 ,1.5) Yi,H1(µˆi,H1 ,σˆi,H1 ,−1.5) di,H1(xi) 1 0.999999301 0.999817806 0.999817806 2 0.999801547 0.999860928 0.999801547 3 0.950278792 0.999998321 0.950278792 14 0.999999904 0.986918080 0.986918080 15 0.950278792 0.999998321 0.950278792 Run di,E1(xi) di,T1(xi) di,E2(xi) di,T2(xi) di,C(xi) di,E3(xi) di,T3(xi) di,V(xi) 1 0.539431 0.960626 0.997495 0.998005 0.820341 0.919648 0.99214 0.868643 2 0.705401 0.962134 0.686472 0.991388 0.935325 0.947124 0.985286 0.830211 3 0.266314 0.633825 0.620349 0.536477 0.954486 0.910332 0.54425 0.758036 14 0.815141 0.976483 0.990825 0.975851 0.938625 0.948317 0.994466 0.932543 15 0.266314 0.81273 0.950855 0.99441 0.959941 0.977555 0.950855 0.525913 4-5 H1 4-6
4.3.3
4.3.2 (3.6) Di(xi) 4-7 ! " Run Di(xi) 1 0.885679 2 0.885472 3 0.644311 4 0.655295 5 0.935072 6 0.862478 7 0.955683 8 0.955507 9 0.749178 10 0.922825 11 0.608523 12 0.681324 13 0.698211 14 0.949453 15 0.766436
4.3.4
#4.3.3 $ % & ' ( ) * + Design Expert , - . / 0 1 2 3 45 6 7 ANOVA 4-8 ! "
8 2 3 45 6 7 R2=0.93909: 3 R2=0.8291; < ANOVA= > ? @ A 45 B C % + DE F G H I J KL M A B C % + D N = O P Q 45 . / 6 7 4-9 ! " 8 P Q 6 7 R S G H x1x2Tx3x4B C % + DU V 5 DTW G H 4 5 B C % + DXY Z [ D\ & 6 7 R2T: 3 R20 ] ^ 0.9271T 0.8866_ ` R2a b c d e: 3 R2f g 5.5%h ijk + P Q N . / 6 7 . / l m n 4 3 2 1 4 3 2 1 4 3 2 1 83808 . 7 09111 . 9 52044 . 0 65374 . 2 49026 . 2 90092 . 4 ) , , , ( x x x x x x x x x x x x + − − − + = η (4.3) 4-8 2 3 45 . / 6 7 ANOVA 4-9 P Q 45 . / 6 7 ANOVA
o4-1 p q r Q s . / t V u o 2 x 4 x 3 x 2 x 4 x 3 x o4-19o4-29o4-3vo4-40 ] ^ p q r Q s =16.70%9w x y s A =25.00%9w x y s B=18.68%vz { s =48.65% . / t V u o" o 4-2 p q w x y s A . / t V u o 1 x 1 x 3 x 3 x 4 x 4 x
o4-4 p q z { s . / t V u o 1 x 2 x x1 3 x 2 x 3 x o 4-3 p q w x y s B . / t V u o 1 x 1 x 2 x 2 x 4 x 4 x
4.3.5
* + Design Expert| } 4.3.4 \ ~ . / 6 7 l 4-10 ! " r s w x y s Aw x y s B z { s | 0.1670 0.2500 0.1554 0.4276 0.9474 0.1670 0.1332 0.2147 0.4851 0.9420 0.1670 0.2257 0.1719 0.4355 0.9417 0.1670 0.1330 0.2233 0.4767 0.9414 0.1670 0.2453 0.1330 0.4547 0.9413 0.1670 0.2500 0.1844 0.3986 0.9408 0.1670 0.1630 0.1993 0.4707 0.9383 4-10= #(x1 ,x2 ,x3 ,x4)=(0.167 ,0.25 ,0.1554 ,0.4276) | 0.9474^ V ^ ij l l X 0.9332# " 4.3.2 \ q X 1.5 [ h 4-10 ^ ~ + [ #RibardoT Allen[14] [ 8 l T ¡ l X¢£ ¤¥¦ [ "4.3.6
(4.3) O . / 6 7 § [ ¨ © ª « o(o4-5)T| § [ ¬ o(o4-6) ! 8 4oE ® ¯ ° G ± ¨ © ² 9t ³ v´ µ ² t ¶ "·4-9\ G ! 8 6 7 ~ ¸ ¹ º q p=0.8136» 8 ® ¯ ¼ ½ G ! & 6 7 W ~ " 6 7 ¾ ¿ ² º q ¾ ¨ , - ) À Á Â ? ² ¼ " 4-10 l |4.3.7
Ã Ä 4.3.5 Å q l (x1 ,x2 ,x3 ,x4)=(0.167 ,0.25 ,0.1554 ,0.4276), - ¿ Æ 4-11 ! " Ä 4-1\ Ç È É Ê & l XË Ç È É Ê ÌG ! & l ¯ Í Î Ï Ð º Ñ Ç q º "Ò Ã 4.3 6 7 & l Ó ) ( i D x =0.9682¢£ ¤¥¦ [ "· & l | o4-5 § [ ¨ © ª « o o 4-6 | § [ ¬ ov 95%Ô Õ Ö × 4-12 ! " H1 68 T1 271.80 E1 165.20 H2 0 T2 0.07% E2 5.08% C 3.23% H3 -8 T3 0.15% E3 5.21% V 2.34% | ¼ 95%Ô Õ Ö × Ê 95%Ô Õ Ö × Ê 0.9474 0.9682 0.8884 1.0000 4-12G ! ¼ T| [ ³ W Ø Ù Ë | 95%Ô Õ Ö × ÌG ! 8 Ú 0  ? ² Û Ü Ý Þ Ã 8 l ß à & á â ã â + ä å æ ç "
4.3.8
!
!
!
!
"
"
"
"
#
#
#
#
Î iè é ê ¤ë ì í ² ' Q î w m 0 1 ij Ú 0 ^ 16.7%r s 925%w x y s A915.54%w x y s BT42.76% z { s " ¼  ? ² Û Ü ï à 8 l ð ñ ò q ó ô ¢¤¥¦ [ " 4-12 l ¿ Æ T| õ S 4-11 l ¿ Æ5.1
ö ÷ 9Q ø ù ú t ô û ü ý ì þ ' q Q " ¨ Z + » l Ç E l ? » 8 ¨ F O £ | ñ Ú » ^ ' i T » W Ç ï F £ | ñ Ú " _ ` Ì X¯ ' ! " eì Ê # í ² " $ % W & ' , T( ) * + , - . / 0 ô è > * ù ß à 1 2 2 3 4 ì 5 í ² 6 47 8 ( ) * + "9 _ ` > : ý ì ì î e ; < ì í ² Q ' $ F 4 Q ì 5 í ² 4 Ì ! " = #C > l , - l · ? ¯ @ A B + ' ì í ² 4 Q î CDabbas t D [10]E + F G ì í ² " ` · ^ ² C6 + õ S ·F Ó H G "·Ribardo TAllen[14] # [ î ^ I ¯ [ ² "h iè é U J + ' Ý K ) J 2 3 = - ' ì í ² Q î "iè é #0 1 L M @ N O á â ã â + ä å æ ç Ó P j Q ° iè é Ç w m T0 1 À Á E ¼ ¯ ² " ? U iè é R " S 3 n 1. iè é Ý K ' w m = #F G I W Ç T = - ; Ö U ' = # 4ü Ú 0 Tw m ' , - Ç » 8 6 F G Ø ì ' " 2. iè é k + . / V B Ú iW X ù 4× Y Z t 5 ¯ Ê / 0 6 [ + í \ T Ý K 6 7 E ] l " 3. iè é B + [ î l 6 [ V $ ß m ò q ó ô 4 ó ô Û « "5.2
iè é ' ì í ² Q î ^ ¯ # _ ` a × n
1. iè é k + RibardoTAllen[14] [ î F ° ¿ Ç È É Ê l / $ Cü b ·F [ = Ò c 4 d E ® ¯ e·$ b ^ Ö × · f q 4F I [ "» 8 iè é Ý Þ g = #h i W j b > : W [ " 2. $ l XÒ 9 [ (V k Ò 9 1)4 [ ³ S l m . / 6 7 \ × ® ¯ G H [ ³ n #, - 0 1 " Ç È É Ê o p ù [ = Ò c q r b s 4t ¯ = 6 > ó u / 0 "» 8 iè é Ý Þ g è é = _ Û [ î N B O Ò 9 [ 4^ ` = #¾ ¨ A B + . / 6 7 , - 0 1 "
[1] v w x 2001y 0 1 z { T| } H ~ ' H 6 µ Ø w - è é ! " [2] 2003y ì b x 6 ; è é # l ß m ^ j µ Ú Ø ß à ø m è é ! " [3] 2002y ' Scheffé6 7 ò A ~ µ \ Ø + @ è é ! " [4] 2002y ¯ ' » B è é Ø ø û ø m T G @ ! " [5] y 2002y ; ¡ l ¢ £ ² 6 è é µ Ú Ø Q ø m @ ! " [6] ý ¤ ¥ 2002y ' Scheffé6 7 ò D ~ µ \ Ø + @ è é ! " [7] ¦ § ¨ 2004y [ ì í ² 4 Q î µ B © Ø ø û ø m G @ ! "
[8] Cornell J. A., “A Comparison between Two Ten-Point Designs for Studying
Three-Component Mixture Systems,” Journal of Quality Technology, Vol. 18, pp. 1-15, 1986.
[9] Cornell J. A., Experiments with Mixtures: Designs, Models, and the Analysis of
Mixture Data, Prentice Hall, Upper Saddle River, New Jersey, 2002.
[10] Dabbas R. M., Fowler J. W., Rollier D. A., and McCarville D., “Multiple
Response Optimization Using Mixture-Designed Experiments and Desirability Functions in Semiconductor Scheduling,” International Journal of Production
Research, Vol. 41, No. 5, pp. 939-961, 2003.
[11] Derringer G. C. and Suich R., “Simultaneous Optimization of Several Response
Variables,” Journal of Quality Technology, Vol. 12, No. 4, pp. 214-219, 1980. [12] Guo T., Geaghan J. P., and Rusch K. A., “Determination of Optimum Ingredients
for Phosphogypsum Composite Stability under Marine Conditions-Response Surface Analysis with Process Variables,” Journal of Environmental Engineering, Vol. 129, No. 4, pp. 358-365, April 2003.
[13] Jeyapaul R., Shahabudeen P., Krishnaiah K., “Quality Management Research by
Considering Multi-Response Problems in the Taguchi Method – A Review,” The
International Journal of Advanced Manufacturing Technology, Jan, 2005.
[14] Ribardo C. and Allen T., “An Alternative Desirability Function for Achieving 'Six
Sigma' Quality,” Quality and Reliability Engineering International, Vol. 19, pp. 227-240, 2003.
[15] Sandoval-Castro C. Capetillo-Leal A., Cetina-Góngora C., Ramirez-Avilés R., L.,
“A Mixture Simplex Design to Study Associative Effects with An in Vitro Gas Production Technique,” Animal Feed Science and Technology, Vol. 101, pp. 191-200, Oct. 2002.
[16] Scheffé H., “Experiments with Mixtures,” Journal of the Royal Statistical Society,
B, Vol. 20, No. 2, pp. 344-360, 1958.
[17] Scheffé H., “Simplex-Centroid Design for Experiments with Mixtures,” Journal of
the Royal Statistical Society, B, Vol. 25, No. 2, pp. 235-263, 1963.
[18] Smith W. F. and Cornell J. A., “Biplot Display for Looking at Multiple Response Data in Mixture Experiments,” Technometrics, Vol. 35, No. 4, pp. 337-350, 1993.
o¤9ª 6 « ª