• 沒有找到結果。

在使用估計參數時,RSS 管制圖的傳統管制界限之整體假警報率和整體 ARL2 0 無法符合我們所期望的值和1 / ,而本文所提供的修正管制界限使得整體 ARL0能夠

真正符合我們所期望的值1/ ,使得我們不必再將就於傳統管制界限因使用估計參數而

造成的錯誤。有些人會認為它們和傳統管制界限的差別不大,因此仍然使用傳統的管制 界限。不過,它們在使用上的便利性就跟傳統管制界限差不多,只有在一開始建構修正 管制界限比較花時間;既然如此,那麼我們為何不選擇具備正確統計檢定性質的修正管 制界限呢?然而,不管是使用傳統或是修正後的管制界限,在RSS 管制圖之中,2

我們都較建議使用S 管制圖,因為它在大部分情況下的表現都優於另外兩者。 2

另一方面,因為本文討論的是RSS 管制圖,所以對於階段一的 m 筆管制下2 子群,只假設它們的標準差都是0,而每個子群i 的平均i不一定要相等;若是經由其

他管制圖發現製程平均也在管制下,那麼我們就可以改變參數0的估計式ˆ0為直接混

和階段一n m 筆資料的樣本全距和樣本標準差,這樣的估計式會使用到更多的自由度,

更為準確。所以我們也可以研究在這種估計參數下的修正管制界限,甚至可允許子群有 不相同之樣本數。除此之外,Zhang et al.(2005)提供的 SAC 和 TPC 兩種準則也可以 和我們的想法做結合,由此建構出參數估計時的「最佳化」修正管制界限。

在 SPC 的範疇上,還有很多種管制圖沒有經過修正,而本文提供的概念與演算法還

可以套用在很多管制圖上,例如說檢定力更強的CUSUM 和 EWMA 管制圖,甚至可以

考慮把我們的想法套用在合併管制圖(combined chart)。

           

附錄 A

對於 S 和S 管制圖,2 ˆtSt,所以式(A.1)變為

 

附錄 B

傳統 S 管制圖之 R 程式碼 n=5

m=25

a=0.0027 #alpha ARL0=1/a

c4=sqrt(2/(n-1))*gamma(0.5*n)/gamma(0.5*(n-1)) M2=(1-c4*c4)/(m*c4*c4)

r=1/(-2+2*sqrt(1+2*M2)) t=M2+1/(16*r*r*r) v=1/(-2+2*sqrt(1+2*t))

c=1+1/(4*v)+1/(32*v*v)-5/(128*v*v*v)

h<-function(u) {dchisq(v*u*u/(c*c), df=v)*2*u*v/(c*c)}

b=1 #rho

Ln=sqrt(qchisq(0.5*a,df=(n-1))/(n-1)) Un=sqrt(qchisq(1-0.5*a,df=(n-1))/(n-1)) L<-function(u)

{pchisq((n-1)*Ln*Ln*u*u/(b*b),df=n-1)+1-pchisq((n-1)*Un*Un*u*u/(b*b),df=n-1)}

g<-function(u) {h(u)/L(u)}

ARL=integrate(g,lower=0,upper=Inf) #overall ARL g2<-function(u) {L(u)*h(u)}

avg_alpha=integrate(g2,0,Inf, rel.tol=1e-10) #overall alarm rate Ln

Un ARL avg_alpha

調整

的修正 S 管制圖之 R 程式碼 n=5

m=25 a=0.0027 ARL0=1/a

c4=sqrt(2/(n-1))*gamma(0.5*n)/gamma(0.5*(n-1)) M2=(1-c4*c4)/(m*c4*c4)

r=1/(-2+2*sqrt(1+2*M2)) t=M2+1/(16*r*r*r) v=1/(-2+2*sqrt(1+2*t))

c=1+1/(4*v)+1/(32*v*v)-5/(128*v*v*v)

h<-function(u) {dchisq(v*u*u/(c*c), df=v)*2*u*v/(c*c)}

aL=a aU=2*a

a1=0.5*(aL+aU) b=1 #rho

Ln=sqrt( qf(0.5*a1, n-1, v) )/c Un=sqrt( qf(1-0.5*a1, n-1, v) )/c L<-function(u)

{pchisq((n-1)*Ln*Ln*u*u/(b*b),df=n-1)+1-pchisq((n-1)*Un*Un*u*u/(b*b),df=n-1)}

g<-function(u) {h(u)/L(u)}

ARL=integrate(g,lower=0,upper=Inf) #overall ARL while(abs(ARL0-ARL$value)>0.001)

{

if(ARL$value>ARL0) { aL=a1

}else {aU=a1}

a1=0.5*(aL+aU)

Ln=sqrt( qf(0.5*a1, n-1, v) )/c Un=sqrt( qf(1-0.5*a1, n-1, v) )/c

ARL=integrate(g,lower=0,upper=Inf) }

a1 Ln Un ARL

整體 ARL 不偏的修正 S 管制圖之 R 程式碼 n=5

m=25 a=0.0027 ARL0=1/a

c4=sqrt(2/(n-1))*gamma(0.5*n)/gamma(0.5*(n-1)) M2=(1-c4*c4)/(m*c4*c4)

r=1/(-2+2*sqrt(1+2*M2)) t=M2+1/(16*r*r*r) v=1/(-2+2*sqrt(1+2*t))

c=1+1/(4*v)+1/(32*v*v)-5/(128*v*v*v)

h<-function(u) {dchisq(v*u*u/(c*c), df=v)*2*u*v/(c*c)}

b=1 #rho a2L=0.5*a a2U=a

a2=0.5*(a2L+a2U)

L2<-function(u) {pchisq(qchisq(a2,df=n-1)*u*u/(b*b),df=n-1)}

a3L=0 a3U=0.5*a

a3=0.5*(a3L+a3U)

L3<-function(u) {1-pchisq(qchisq(1-a3,df=n-1)*u*u/(b*b),df=n-1)}

g<-function(u) {h(u)/(L2(u)+L3(u))}

ARL=integrate(g,lower=0,upper=Inf) #overall ARL

while(abs(ARL0-ARL$value)>0.001)

dL2=dchisq( qchisq(a2,df=n-1)*(u/b)^2, df=n-1 ) dL2=dL2*qchisq(a2,df=n-1)*u*u*(-2)/(b^3) dL3=dchisq( qchisq(1-a3,df=n-1)*(u/b)^2, df=n-1 ) dL3=dL3*qchisq(1-a3,df=n-1)*u*u*2/(b^3) return(dL2+dL3)

}

gg<-function(u) {-dL(u)*h(u)/(L2(u)+L3(u))} #式(3.10)中待機分之函數(integrand) dg=integrate(gg,0,Inf) #ARL’(b) (目前 b=1)

ARL=integrate(g,0,Inf)

while(abs(ARL0-ARL$value)>0.001) {

if(ARL$value>ARL0) { a3L=a3

}else {a3U=a3}

a3=0.5*(a3L+a3U) ARL=integrate(g,0,Inf) }

dg=integrate(gg,0,Inf) }

Ln=sqrt(qchisq(a2,df=n-1)/(n-1)) Un=sqrt(qchisq(1-a3,df=n-1)/(n-1)) a2

a3 Ln Un

傳統

S

2管制圖之 R 程式碼 n=5

m=25 a=0.0027 ARL0=1/a k=m*(n-1)

h<-function(u) {dchisq(k*u*u, df=k)*2*u*k}

b=1 #rho

Ln=sqrt(qchisq(0.5*a,df=n-1)/(n-1)) Un=sqrt(qchisq(1-0.5*a,df=n-1)/(n-1)) L<-function(u)

{pchisq((n-1)*Ln*Ln*u*u/(b*b),df=n-1)+1-pchisq((n-1)*Un*Un*u*u/(b*b),df=n-1)}

g<-function(u) {h(u)/L(u)}

ARL=integrate(g,0,Inf) #overall ARL g2<-function(u) {L(u)*h(u)}

avg_alpha=integrate(g2,0,Inf, rel.tol=1e-10) #overall alarm rate Ln

Un ARL avg_alpha

調整

的修正

S

2管制圖之 R 程式碼 n=5

m=25 a=0.0027 ARL0=1/a k=m*(n-1)

h<-function(u) {dchisq(k*u*u, df=k)*2*u*k}

aL=a aU=2*a

a1=0.5*(aL+aU) b=1 #rho

Ln=sqrt( qf(0.5*a1, n-1, k) ) Un=sqrt( qf(1-0.5*a1, n-1, k) ) L<-function(u)

{pchisq((n-1)*Ln*Ln*u*u/(b*b),df=n-1)+1-pchisq((n-1)*Un*Un*u*u/(b*b),df=n-1)}

g<-function(u) {h(u)/L(u)}

ARL=integrate(g,lower=0,upper=Inf) #overall ARL while(abs(ARL0-ARL$value)>0.001)

{

if(ARL$value>ARL0) { aL=a1

}else {aU=a1}

a1=0.5*(aL+aU)

Ln=sqrt( qf(0.5*a1, n-1, k) ) Un=sqrt( qf(1-0.5*a1, n-1, k) ) ARL=integrate(g,lower=0,upper=Inf) }

a1 Ln Un ARL

整體 ARL 不偏的修正

S

2管制圖之 R 程式碼 n=5

m=25 a=0.0027 ARL0=1/a k=m*(n-1)

h<-function(u) {dchisq(k*u*u, df=k)*2*u*k}

b=1 #rho a2L=0.5*a a2U=a

a2=0.5*(a2L+a2U)

Ln=sqrt(qchisq(a2,df=n-1)/(n-1))

L2<-function(u) {pchisq(qchisq(a2,df=n-1)*u*u/(b*b),df=n-1)}

a3L=0 a3U=0.5*a

a3=0.5*(a3L+a3U)

Un=sqrt(qchisq(1-a3,df=n-1)/(n-1))

L3<-function(u) {1-pchisq(qchisq(1-a3,df=n-1)*u*u/(b*b),df=n-1)}

g<-function(u) {h(u)/(L2(u)+L3(u))}

ARL=integrate(g,0,Inf) #overall ARL while(abs(ARL0-ARL$value)>0.001)

{

dL2=dchisq( qchisq(a2,df=n-1)*(u/b)^2, df=n-1 ) dL2=dL2*qchisq(a2,df=n-1)*u*u*(-2)/(b^3) dL3=dchisq( qchisq(1-a3,df=n-1)*(u/b)^2, df=n-1 ) dL3=dL3*qchisq(1-a3,df=n-1)*u*u*2/(b^3) return(dL2+dL3)

}

gg<-function(u) {-dL(u)*h(u)/(L2(u)+L3(u))} #式(3.10)中待機分之函數(integrand) dg=integrate(gg,0,Inf) #ARL’(b) (目前 b=1)

while(abs(ARL0-ARL$value)>0.001)

{

if(ARL$value>ARL0) { a3L=a3

}else {a3U=a3}

a3=0.5*(a3L+a3U) ARL=integrate(g,0,Inf) }

dg=integrate(gg,0,Inf) }

Ln=sqrt(qchisq(a2,df=n-1)/(n-1)) Un=sqrt(qchisq(1-a3,df=n-1)/(n-1)) a2

a3 Ln Un

傳統 R 管制圖之 matlab 程式碼 n=5

m=25 a=0.0027;

ARL0=1/a;

f=@(r,s) n*(n-1)*(normcdf(s+r)-normcdf(s)).^(n-2).*normpdf(s).*normpdf(s+r); %式(3.17) R1=@(r,s) r.*f(r,s);

d2=dblquad(R1,0,20,-8,4,1E-8);

R2=@(r,s) r.*r.*f(r,s);

v2=dblquad(R2,0,20,-8,4,1E-8);

v2=v2-d2*d2;

M1=v2/(m*d2*d2);

r=1/(-2+2*sqrt(1+2*M1));

t=M1+1/(16*r*r*r);

v=1/(-2+2*sqrt(1+2*t));

c=1+1/(4*v)+1/(32*v*v)-5/(128*v*v*v);

b1=0; %二分法搜尋Ln的下界

tmp=dblquad(f,0,b3,-8,4,1E-8);

end

tmp=dblquad(f,0,c3,-8,4,1E-8);

end Un=c3;

b=1; %rho

L=@(u) dblquad(f,0,Ln*u/b,-8,4,1E-8)+1-dblquad(f,0,Un*u/b,-8,4,1E-8);

g=@(u) 2*u*v/(c*c).*chi2pdf(v*u.*u/(c.*c),v)./L(u); %h(u)/L(u) dx=0.01; %黎曼和的partition大小

ARL=ARL+avg;

end

g2=@(u) 2*u*v/(c*c).*chi2pdf(v*u.*u/(c*c),v).*L(u); %h(u)*L(u) avg_alpha=0;

right= g2(0.75);

for i = 1:k %此for迴圈目的在做黎曼和 left=right;

right=g2(0.75+i*dx);

avg=0.5*(left+right)*dx;

avg_alpha=avg_alpha+avg;

end Ln Un

ARL %overall ARL avg_alpha %overall alarm rate

調整

的修正 R 管制圖之 matlab 程式碼 n=5

m=25 a=0.0027;

ARL0=1/a;

f=@(r,s) n*(n-1)*(normcdf(s+r)-normcdf(s)).^(n-2).*normpdf(s).*normpdf(s+r); %式(3.17) R1=@(r,s) r.*f(r,s);

d2=dblquad(R1,0,20,-8,4,1E-8);

R2=@(r,s) r.*r.*f(r,s);

v2=dblquad(R2,0,20,-8,4,1E-8);

v2=v2-d2*d2;

M1=v2/(m*d2*d2);

r=1/(-2+2*sqrt(1+2*M1));

t=M1+1/(16*r*r*r);

v=1/(-2+2*sqrt(1+2*t));

c=1+1/(4*v)+1/(32*v*v)-5/(128*v*v*v);

aL=0.5*a;

tmp=dblquad(f,0,b3,-8,4,1E-8);

end

tmp=dblquad(f,0,c3,-8,4,1E-8);

end Un=c3;

b=1; %rho

L=@(u) dblquad(f,0,Ln*u/b,-8,4,1E-8)+1-dblquad(f,0,Un*u/b,-8,4,1E-8);

g=@(u) 2*u*v/(c*c).*chi2pdf(v*u.*u/(c.*c),v)./L(u); %h(u)/L(u)

dx=0.01; %黎曼和的partition大小 k=0.5/dx; % k為partition的數量

ARL=0;

tmp=dblquad(f,0,b3,-8,4,1E-8);

while( abs(tmp-0.5*a1) > a1/500)

tmp=dblquad(f,0,c3,-8,4,1E-8);

while( abs(tmp-1+0.5*a1) > a1/500) if(tmp>1-0.5*a1) c2=c3;

else c1=c3;

end

c3=0.5*(c1+c2);

tmp=dblquad(f,0,c3,-8,4,1E-8);

end Un=c3;

L=@(u) dblquad(f,0,Ln*u/b,-8,4,1E-8)+1-dblquad(f,0,Un*u/b,-8,4,1E-8);

g=@(u) 2*u*v/(c*c).*chi2pdf(v*u.*u/(c.*c),v)./L(u);

ARL=0;

right=g(0.75);

for i = 1:k left=right;

right=g(0.75+i*dx);

avg=0.5*(left+right)*dx;

ARL=ARL+avg;

end end a1 Ln Un ARL

整體 ARL 不偏的修正 R 管制圖之 matlab 程式碼 n=5

m=25 a=0.0027;

ARL0=1/a;

f=@(r,s) n*(n-1)*(normcdf(s+r)-normcdf(s)).^(n-2).*normpdf(s).*normpdf(s+r); %式(3.17)

R1=@(r,s) r.*f(r,s);

tmp=dblquad(f,0,b3,-8,4,1E-8);

end

while( abs(tmp-1+a3) > a3/500) if(tmp>1-a3) c2=c3;

else c1=c3;

end

c3=0.5*(c1+c2);

tmp=dblquad(f,0,c3,-8,4,1E-8);

end

tmp=dblquad(f,0,c3,-8,4,1E-8);

while( abs(tmp-1+a3) > a3/500)

%積分對象( Ln*u*R(Ln*u) - Un*u*R(Un*u) ).*h(u)./L(u) dg=0;

k1=0.5/dx;

for i=1:k1

dg=dg+dx*( Ln*(0.75+(i-0.5)*dx)*R(Ln*(0.75+(i-0.5)*dx)) -

Un*(0.75+(i-0.5)*dx)*R(Un*(0.75+(i-0.5)*dx)) ).*h((0.75+(i-0.5)*dx))./L((0.75+(i-0.5)*dx))

; end

while(abs(dg)>0.1 && a2U-a2L>1E-8) if(dg<0)

a2L=a2;

tmp=dblquad(f,0,b3,-8,4,1E-8);

while( abs(tmp-a2) > a2/500)

tmp=dblquad(f,0,c3,-8,4,1E-8);

while( abs(tmp-1+a3) > a3/500) if(tmp>1-a3) c2=c3;

else c1=c3;

end

c3=0.5*(c1+c2);

tmp=dblquad(f,0,c3,-8,4,1E-8);

while(abs(ARL-ARL0)>0.01 && a3U-a3L>1E-8) if(ARL>ARL0)

Un=c3;

L3=@(u) 1-dblquad(f,0,Un*u/b,-8,4,1E-8);

L=@(u) L2(u)+L3(u);

g=@(u) h(u)./L(u);

ARL=0;

right=g(0.75);

for i = 1:k left=right;

right=g(0.75+i*dx);

avg=0.5*(left+right)*dx;

ARL=ARL+avg;

end

end dg=0;

for i=1:k1

dg=dg+dx*( Ln*(0.75+(i-0.5)*dx)*R(Ln*(0.75+(i-0.5)*dx)) -

Un*(0.75+(i-0.5)*dx)*R(Un*(0.75+(i-0.5)*dx)) ).*h((0.75+(i-0.5)*dx))./L((0.75+(i-0.5)*dx))

;

end end a2 a3 Ln Un ARL

參考文獻

Champ, C. W., Jones-Farmer, L. A., and Rigdon, S. E. (2005). Properties of the T control 2 chart when parameters are estimated. Technometrics 47, 437-445.

Chen, G. (1997). The mean and standard deviation of the run length distribution of X charts when control limits are estimated. Statistica Sinica 7, 789-798.

Chen, G. (1998). The run length distribution of the R, s and s control charts when 2  is estimated. Canadian Journal of Statistics 26, 311-322.

Jensen, W. A., Jones-Farmer, L. A., Champ, C. W., and Woodall, W. H. (2006). Effects of parameter estimation on control chart properties: a literature review. Journal of Quality Technology 38, 349-364.

Maravelakis, P. E., Panaretos, J., and Psarakis, S. (2002). Effect of estimation of the process parameters on the control limits of the univariate control charts for process dispersion.

Communications in Statistics-Simulation and Computation 31, 443-461.

Montgomery, D. C. (2009). Introduction to Statistical Quality Control. 6th edition. Wiley, New York.

Patnaik, P.B. (1950). The use of mean range as an estimator of variance in statistical tests.

Biometrika 37, 78-87.

Pignatiello, J. J. Jr., Acosta-Mejia C. A., and Rao, B. V. (1995). The performance of control charts for monitoring process dispersion. Proc. 4th Indust. Engi. Res. Confe. 320-328.

Zhang, L., Bebbington, M. S., Lai, C. D., and Govindaraju, K. (2005). On statistical design of the S chart. Communications in Statistics-Theory and Methods 34, 229-244. 2

         

表2.1:三種管制圖中各符號所對應的式子。

表2.2:校正係數d 和2 c 。 4

n d2 c4 n d2 c4

2 1.1284 0.7979 32 4.1393 0.9920 3 1.6926 0.8862 33 4.1648 0.9922 4 2.0588 0.9213 34 4.1894 0.9925 5 2.3259 0.9400 35 4.2132 0.9927 6 2.5344 0.9515 36 4.2362 0.9929 7 2.7044 0.9594 37 4.2586 0.9931 8 2.8472 0.9650 38 4.2802 0.9933 9 2.9700 0.9693 39 4.3012 0.9934 10 3.0775 0.9727 40 4.3216 0.9936 11 3.1729 0.9754 41 4.3414 0.9938 12 3.2585 0.9776 42 4.3606 0.9939 13 3.3360 0.9794 43 4.3794 0.9941 14 3.4068 0.9810 44 4.3976 0.9942 15 3.4718 0.9823 45 4.4154 0.9943 16 3.5320 0.9835 46 4.4328 0.9945 17 3.5879 0.9845 47 4.4497 0.9946 18 3.6401 0.9854 48 4.4662 0.9947 19 3.6890 0.9862 49 4.4824 0.9948 20 3.7350 0.9869 50 4.4981 0.9949 21 3.7783 0.9876 51 4.5136 0.9950 22 3.8194 0.9882 52 4.5286 0.9951 23 3.8583 0.9887 53 4.5434 0.9952 24 3.8953 0.9892 54 4.5578 0.9953 25 3.9306 0.9896 55 4.5720 0.9954 26 3.9643 0.9901 56 4.5858 0.9955 27 3.9965 0.9904 57 4.5994 0.9955 28 4.0274 0.9908 58 4.6127 0.9956 29 4.0570 0.9911 59 4.6258 0.9957 30 4.0855 0.9914 60 4.6386 0.9958 31 4.1129 0.9917

2.3:當0.0027,三種傳統管制圖在不同n m之下的整體ARL,其中m 代表參數已知的情況。

整體ARL

n m 管制圖 ρ=0.4 0.6 0.8 0.9 0.95 1 1.05 1.1 1.2 1.7 2.5

5 25 R 24.04 107.96 313.14 395.81 384.01 333.75 262.45 190.99 90.11 6.53 1.85 S 23.83 107.41 314.21 400.83 388.14 334.06 258.15 183.60 82.56 5.76 1.74

23.98 108.17 316.74 403.22 388.57 331.87 254.09 179.11 79.73 5.67 1.73

50 R 23.46 105.14 312.2 416.73 408.75 349.33 263.85 182.99 81.1 6.26 1.84 S 23.28 104.73 312.78 422.12 413.71 349.55 258.37 174.38 73.60 5.53 1.73

23.35 105.09 314.01 423.76 414.28 348.30 255.86 171.77 72.22 5.49 1.72

75 R 23.27 104.22 311.14 424.43 418.99 355.52 263.66 179.27 77.95 6.17 1.83 S 23.10 103.86 311.57 429.92 424.35 355.71 257.63 170.19 70.52 5.46 1.72

23.15 104.10 312.38 431.13 424.87 354.84 255.80 168.36 69.63 5.43 1.72

100 R 23.18 103.77 310.47 428.38 424.63 358.87 263.35 177.16 76.37 6.13 1.83 S 23.01 103.43 310.84 433.91 430.24 359.05 257.00 167.83 68.99 5.43 1.72

23.05 103.61 311.43 434.86 430.70 358.37 255.56 166.42 68.33 5.41 1.72

R 22.91 102.44 307.93 440.20 444.32 370.37 261.09 169.63 71.63 6.01 1.82 S 22.75 102.16 308.15 445.75 450.97 370.37 253.53 159.56 64.45 5.33 1.71 22.75 102.16 308.15 445.75 450.97 370.37 253.53 159.56 64.45 5.33 1.71

10 25 R 2.62 20.45 146.09 302.18 349.38 327.27 250.38 165.18 61.63 3.49 1.23 S 2.33 18.92 141.25 306.29 358.49 327.24 233.83 140.64 44.90 2.56 1.13

2.34 19.02 142.17 308.19 359.70 326.39 231.44 138.34 44.05 2.55 1.13

50 R 2.58 19.76 140.61 307.82 369.6 345.87 253.03 157.95 56.38 3.41 1.23 S 2.30 18.35 136.29 310.81 379.53 345.44 233.12 131.59 40.71 2.52 1.13

2.30 18.40 136.73 311.85 380.38 344.97 231.68 130.35 40.32 2.51 1.13

75 R 2.56 19.54 138.73 309.06 377.4 353.34 253.39 154.84 54.67 3.38 1.23 S 2.29 18.17 134.63 311.57 387.64 352.75 232.07 127.97 39.38 2.51 1.13

2.3(續):當0.0027,三種傳統管制圖在不同nm之下的整體ARL,其中m 代表參數已知的情況。

20 25 R 1.1 4.42 55.73 200.47 301.15 322.42 241.31 141.62 42.32 2.12 1.04 S 1.02 3.23 46.20 189.45 309.32 323.50 202.68 93.78 20.91 1.38 1.01

2.4:當0.0027,三種傳統管制圖在不同n m之下的整體警報率,其中m 代表參數已知的情況。

整體警報率

n m 管制圖 ρ=0.4 0.6 0.8 0.9 0.95 1 1.05 1.1 1.2 1.7 2.5

5 25 R 0.04465 0.01004 0.00339 0.00261 0.00284 0.00367 0.00531 0.00802 0.01764 0.17380 0.54914 S 0.04492 0.01006 0.00338 0.00259 0.00283 0.00374 0.00557 0.00862 0.01958 0.19481 0.58323 0.04448 0.00995 0.00334 0.00256 0.00282 0.00375 0.00560 0.00870 0.01982 0.19663 0.58565

50 R 0.04415 0.00990 0.00332 0.00242 0.00252 0.00315 0.00453 0.00692 0.01576 0.17015 0.54900 S 0.04444 0.00992 0.00331 0.00239 0.00249 0.00318 0.00471 0.00740 0.01751 0.19131 0.58341 0.04422 0.00987 0.00329 0.00238 0.00249 0.00318 0.00472 0.00743 0.01762 0.19222 0.58463

75 R 0.04398 0.00985 0.00329 0.00237 0.00242 0.00299 0.00429 0.00657 0.01515 0.16892 0.54895 S 0.04428 0.00988 0.00329 0.00234 0.00240 0.00301 0.00445 0.00701 0.01684 0.19013 0.58347 0.04413 0.00984 0.00327 0.00233 0.00239 0.00301 0.00445 0.00703 0.01691 0.19073 0.58429

100 R 0.04390 0.00983 0.00328 0.00234 0.00238 0.00292 0.00417 0.00639 0.01485 0.16829 0.54893 S 0.04420 0.00986 0.00328 0.00231 0.00235 0.00293 0.00432 0.00682 0.01650 0.18953 0.58350 0.04409 0.00983 0.00327 0.00231 0.00234 0.00293 0.00432 0.00683 0.01655 0.18998 0.58412

R 0.04365 0.00976 0.00325 0.00227 0.00225 0.00270 0.00383 0.00590 0.01396 0.16641 0.54886 S 0.04396 0.00979 0.00325 0.00224 0.00222 0.00270 0.00394 0.00627 0.01551 0.18773 0.58360 0.04396 0.00979 0.00325 0.00224 0.00222 0.00270 0.00394 0.00627 0.01551 0.18773 0.58360

10 25 R 0.39507 0.05423 0.00786 0.00356 0.00297 0.00338 0.00507 0.00849 0.02278 0.30464 0.81407 S 0.44170 0.05797 0.00802 0.00351 0.00290 0.00347 0.00575 0.01053 0.03137 0.40597 0.88538 0.43996 0.05753 0.00794 0.00348 0.00288 0.00348 0.00580 0.01063 0.03167 0.40771 0.88616

50 R 0.39463 0.05329 0.00763 0.00339 0.00274 0.00302 0.00450 0.00761 0.02111 0.30223 0.81555 S 0.44149 0.05705 0.00781 0.00335 0.00267 0.00307 0.00505 0.00941 0.02924 0.40472 0.88669 0.44061 0.05683 0.00778 0.00334 0.00266 0.00307 0.00506 0.00946 0.02938 0.40560 0.88708

75 R 0.39448 0.05297 0.00756 0.00333 0.00267 0.00291 0.00431 0.00733 0.02056 0.30142 0.81605 S 0.44142 0.05675 0.00775 0.00330 0.00260 0.00294 0.00482 0.00905 0.02853 0.40430 0.88713 0.44083 0.05660 0.00772 0.00329 0.00259 0.00294 0.00483 0.00908 0.02862 0.40488 0.88739

100 R 0.39441 0.05281 0.00752 0.00331 0.00263 0.00285 0.00422 0.00719 0.02029 0.30101 0.81630 S 0.44139 0.05659 0.00771 0.00328 0.00256 0.00288 0.00471 0.00887 0.02817 0.40408 0.88736 0.44094 0.05648 0.00769 0.00327 0.00256 0.00288 0.00472 0.00889 0.02824 0.40452 0.88755

R 0.39419 0.05234 0.00741 0.00323 0.00253 0.00270 0.00396 0.00678 0.01948 0.29976 0.81705 S 0.44128 0.05613 0.00761 0.00321 0.00247 0.00270 0.00439 0.00834 0.02712 0.40344 0.88802 0.44128 0.05613 0.00761 0.00321 0.00247 0.00270 0.00439 0.00834 0.02712 0.40344 0.88802

2.4(續):當0.0027,三種傳統管制圖在不同nm之下的整體警報率,其中m 代表參數已知的情況。

整體警報率

n m 管制圖 ρ=0.4 0.6 0.8 0.9 0.95 1 1.05 1.1 1.2 1.7 2.5

15 25 R 0.75538 0.13786 0.01400 0.00469 0.00326 0.00331 0.00505 0.00902 0.02698 0.40488 0.91931 S 0.86476 0.16624 0.01514 0.00473 0.00315 0.00340 0.00613 0.01274 0.04487 0.58937 0.97351 0.86396 0.16539 0.01502 0.00469 0.00314 0.00341 0.00618 0.01285 0.04520 0.59076 0.97370

50 R 0.75815 0.13622 0.01355 0.00446 0.00303 0.00299 0.00453 0.00821 0.02535 0.40343 0.92070 S 0.86787 0.16461 0.01473 0.00454 0.00294 0.00303 0.00545 0.01159 0.04255 0.58992 0.97414 0.86747 0.16418 0.01467 0.00452 0.00293 0.00304 0.00547 0.01164 0.04271 0.59062 0.97423

75 R 0.75909 0.13566 0.01340 0.00439 0.00295 0.00289 0.00437 0.00795 0.02481 0.40294 0.92116 S 0.86892 0.16406 0.01459 0.00448 0.00287 0.00292 0.00524 0.01122 0.04178 0.59010 0.97435 0.86866 0.16377 0.01456 0.00446 0.00287 0.00292 0.00525 0.01125 0.04188 0.59057 0.97441

100 R 0.75957 0.13538 0.01332 0.00435 0.00292 0.00284 0.00429 0.00782 0.02455 0.40269 0.92139 S 0.86945 0.16378 0.01452 0.00445 0.00284 0.00286 0.00513 0.01103 0.04139 0.59020 0.97445 0.86925 0.16357 0.01450 0.00444 0.00284 0.00286 0.00514 0.01105 0.04147 0.59055 0.97450

R 0.76100 0.13455 0.01310 0.00425 0.00282 0.00270 0.00405 0.00744 0.02375 0.40193 0.92210 S 0.87104 0.16295 0.01432 0.00436 0.00275 0.00270 0.00482 0.01048 0.04023 0.59048 0.97477 0.87104 0.16295 0.01432 0.00436 0.00275 0.00270 0.00482 0.01048 0.04023 0.59048 0.97477

20 25 R 0.91376 0.24111 0.02137 0.00587 0.00356 0.00329 0.00505 0.00947 0.03051 0.48441 0.96352 S 0.98458 0.32334 0.02489 0.00612 0.00345 0.00336 0.00656 0.01512 0.05985 0.73000 0.99453 0.98446 0.32224 0.02473 0.00608 0.00343 0.00337 0.00661 0.01523 0.06022 0.73101 0.99457

50 R 0.91639 0.23954 0.02068 0.00558 0.00332 0.00298 0.00457 0.00869 0.02888 0.48373 0.96451 S 0.98581 0.32211 0.02423 0.00588 0.00324 0.00302 0.00588 0.01391 0.05737 0.73152 0.99473 0.98575 0.32155 0.02414 0.00586 0.00323 0.00302 0.00590 0.01396 0.05755 0.73203 0.99475

75 R 0.91727 0.23901 0.02044 0.00549 0.00324 0.00288 0.00441 0.00844 0.02834 0.48349 0.96484 S 0.98621 0.32169 0.02400 0.00580 0.00317 0.00291 0.00567 0.01351 0.05654 0.73203 0.99479 0.98617 0.32132 0.02395 0.00579 0.00316 0.00291 0.00568 0.01354 0.05666 0.73237 0.99481

100 R 0.91771 0.23874 0.02033 0.00544 0.00320 0.00284 0.00434 0.00832 0.02808 0.48338 0.96501 S 0.98641 0.32148 0.02388 0.00576 0.00314 0.00286 0.00556 0.01331 0.05612 0.73229 0.99483 0.98638 0.32120 0.02384 0.00575 0.00313 0.00286 0.00557 0.01334 0.05621 0.73255 0.99484

R 0.91904 0.23794 0.01998 0.00530 0.00309 0.00270 0.00412 0.00795 0.02728 0.48302 0.96550 S 0.98700 0.32084 0.02356 0.00565 0.00304 0.00270 0.00524 0.01273 0.05487 0.73307 0.99493 0.98700 0.32084 0.02356 0.00565 0.00304 0.00270 0.00524 0.01273 0.05487 0.73307 0.99493

3.1:當0.0027,三種管制圖在不同nm下的調整 修正管制界限因子;其中m 代表參數已知的情況,

3.2:當0.0027,調整 的修正管制界限在各種情況下的整體 ARL,其中m 代表參數已知的情況。

3.2(續):當 0.0027,調整 的修正管制界限在各種情況下的整體 ARL,其中m 代表參數已知的情況。

3.3:當0.0027,三種管制圖在不同n m的整體ARL 不偏之修正管制界限因子;其中m 代表參數已知

的情況。

n m 管制圖 , , n m 管制圖 , ,

5 25 R 0.002109 0.000292 0.444589 5.899378 15 25 R 0.001603 0.000741 1.617983 6.32793 S 0.002142 0.000263 0.182934 2.313219 S 0.001752 0.000598 0.489816 1.638876

0.002166 0.000257 0.183457 2.316058 0.001768 0.000589 0.490205 1.639784

50 R 0.002152 0.000375 0.446488 5.817729 50 R 0.001666 0.000827 1.623561 6.29335 S 0.002177 0.000345 0.183687 2.281107 S 0.001807 0.000688 0.491173 1.630063

0.002190 0.000342 0.183961 2.282203 0.001815 0.000684 0.491374 1.630485

75 R 0.002152 0.000424 0.446488 5.777903 75 R 0.001688 0.000868 1.625469 6.279252 S 0.002191 0.000381 0.183990 2.269117 S 0.001829 0.000726 0.491706 1.626722

0.002200 0.000379 0.184176 2.269748 0.001835 0.000723 0.491842 1.626993

100 R 0.002173 0.000431 0.447626 5.771497 100 R 0.001709 0.000880 1.627314 6.274803 S 0.002199 0.000401 0.184155 2.262793 S 0.001841 0.000746 0.491992 1.624948

0.002205 0.000400 0.184296 2.263222 0.001845 0.000744 0.492094 1.625148

R 0.002194 0.000506 0.448974 5.717122 R 0.001751 0.000949 1.631079 6.251011 S 0.002225 0.000475 0.184723 2.242319 S 0.001881 0.000819 0.492952 1.619101 0.002225 0.000475 0.184723 2.242319 0.001881 0.000819 0.492952 1.619101

10 25 R 0.001793 0.000574 1.167455 6.152463 20 25 R 0.001498 0.000841 1.932927 6.464918 S 0.001880 0.000484 0.387142 1.818083 S 0.001675 0.000668 0.553363 1.540296

0.001899 0.000475 0.387627 1.819500 0.001688 0.000660 0.553684 1.540957

50 R 0.001835 0.000669 1.171196 6.103699 50 R 0.001561 0.000931 1.939085 6.43343 S 0.001928 0.000574 0.388389 1.804717 S 0.001733 0.000759 0.554735 1.533547

0.001938 0.000569 0.388641 1.805354 0.001740 0.000754 0.554900 1.533859

75 R 0.001856 0.000705 1.172782 6.086797 75 R 0.001582 0.000971 1.940988 6.420225 S 0.001948 0.000612 0.388882 1.799671 S 0.001756 0.000796 0.555272 1.530982

0.001955 0.000608 0.389052 1.800074 0.001761 0.000793 0.555383 1.531184

100 R 0.001877 0.000717 1.174469 6.081235 100 R 0.001603 0.000979 1.943355 6.418352 S 0.001958 0.000632 0.389147 1.796998 S 0.001769 0.000817 0.555560 1.529619

0.001963 0.000630 0.389275 1.797290 0.001773 0.000814 0.555644 1.529768

R 0.001909 0.000791 1.177226 6.049608 R 0.001651 0.001049 1.947128 6.396688 S 0.001993 0.000707 0.390022 1.788136 S 0.001812 0.000888 0.556523 1.525110 0.001993 0.000707 0.390022 1.788136 0.001812 0.000888 0.556523 1.525110

3.4:當0.0027,整體ARL 不偏的修正管制界限在各種情況下的整體 ARL,其中m 代表參數已知。

3.4(續):當 0.0027,整體ARL 不偏的修正管制界限在各種情況下的整體 ARL,其中m 代表參數已知。

表4.1:前 25 筆樣本中,硬烤製程的流量寬度(flow width)資料(單位:微米)。

階段一 樣本i

晶圓

Ri Si S i2 1 2 3 4 5

1 1.3235 1.4128 1.6744 1.4573 1.6914 0.3679 0.1635 0.0267 2 1.4314 1.3592 1.6075 1.4666 1.6109 0.2517 0.1111 0.0123 3 1.4284 1.4871 1.4932 1.4324 1.5674 0.1390 0.0565 0.0032 4 1.5028 1.6352 1.3841 1.2831 1.5507 0.3521 0.1389 0.0193 5 1.5604 1.2735 1.5265 1.4363 1.6441 0.3706 0.1412 0.0199 6 1.5955 1.5451 1.3574 1.3281 1.4198 0.2674 0.1168 0.0136 7 1.6274 1.5064 1.8366 1.4177 1.5144 0.4189 0.1614 0.0260 8 1.4190 1.4303 1.6637 1.6067 1.5519 0.2447 0.1077 0.0116 9 1.3884 1.7277 1.5355 1.5176 1.3688 0.3589 0.1439 0.0207 10 1.4039 1.6697 1.5089 1.4627 1.5220 0.2658 0.0988 0.0098 11 1.4158 1.7667 1.4278 1.5928 1.4181 0.3509 0.1548 0.0240 12 1.5821 1.3355 1.5777 1.3908 1.7559 0.4204 0.1682 0.0283 13 1.2856 1.4106 1.4447 1.6398 1.1928 0.4470 0.1699 0.0289 14 1.4951 1.4036 1.5893 1.6458 1.4969 0.2422 0.0937 0.0088 15 1.3589 1.2863 1.5996 1.2497 1.5471 0.3499 0.1568 0.0246 16 1.5747 1.5301 1.5171 1.1839 1.8662 0.6823 0.2423 0.0587 17 1.3680 1.7269 1.3957 1.5014 1.4449 0.3589 0.1432 0.0205 18 1.4163 1.3864 1.3057 1.6210 1.5573 0.3153 0.1289 0.0166 19 1.5796 1.4185 1.6541 1.5116 1.7247 0.3062 0.1195 0.0143 20 1.7106 1.4412 1.2361 1.3820 1.7601 0.5240 0.2230 0.0497 21 1.4371 1.5051 1.3485 1.5670 1.4880 0.2185 0.0819 0.0067 22 1.4738 1.5936 1.6583 1.4973 1.4720 0.1863 0.0832 0.0069 23 1.5917 1.4333 1.5551 1.5295 1.6866 0.2533 0.0922 0.0085 24 1.6399 1.5243 1.5705 1.5563 1.5530 0.1156 0.0431 0.0019 25 1.5797 1.3663 1.6240 1.3732 1.6887 0.3224 0.1482 0.0220

平均值 0.3252 0.1316 0.0193

表4.2:後 20 筆樣本中,硬烤製程的流量寬度資料(單位:微米)。

階段二 樣本t

晶圓

Rt St S t2 1 2 3 4 5

1 1.4483 1.5458 1.4538 1.4303 1.6206 0.1903 0.0811 0.0066 2 1.5435 1.6899 1.5830 1.3358 1.4187 0.3541 0.1391 0.0194 3 1.5175 1.3446 1.4723 1.6657 1.6661 0.3215 0.1367 0.0187 4 1.5454 1.0931 1.4072 1.5039 1.5264 0.4523 0.1877 0.0352 5 1.4418 1.5059 1.5124 1.4620 1.6263 0.1845 0.0716 0.0051 6 1.4301 1.2725 1.5945 1.5397 1.5252 0.3220 0.1265 0.0160 7 1.4981 1.4506 1.6174 1.5837 1.4962 0.1668 0.0689 0.0047 8 1.3009 1.5060 1.6231 1.5831 1.6454 0.3445 0.1395 0.0195 9 1.4132 1.4603 1.5808 1.7111 1.7313 0.3181 0.1434 0.0206 10 1.3817 1.3135 1.4953 1.4894 1.4596 0.1818 0.0783 0.0061 11 1.5765 1.7014 1.4026 1.2773 1.4541 0.4241 0.1628 0.0265 12 1.4936 1.4373 1.5139 1.4808 1.5293 0.0920 0.0353 0.0012 13 1.5729 1.6738 1.5048 1.5651 1.7473 0.2425 0.0966 0.0093 14 1.8089 1.5513 1.8250 1.4389 1.6558 0.3861 0.1659 0.0275 15 1.6236 1.5393 1.6738 1.8698 1.5036 0.3662 0.1440 0.0207 16 1.4120 1.7931 1.7345 1.6391 1.7791 0.3881 0.1571 0.0247 17 1.7372 1.5663 1.4910 1.7809 1.5504 0.2899 0.1264 0.0160 18 1.5971 1.7394 1.6832 1.6677 1.7974 0.2003 0.0757 0.0057 19 1.4295 1.6536 1.9134 1.7272 1.4370 0.4839 0.2048 0.0419 20 1.6217 1.8220 1.7915 1.6744 1.9404 0.3187 0.1258 0.0158  

  圖2.1:當 0.0027,參數已知時,傳統S 管制圖對應不同 n 的函數2 ARLw1( ) 之圖形。

0.0 0.5 1.0 1.5 2.0

0100200300400

rho

ARL n=5

n=10 n=15 n=20

ρ

ARL=370.37 

  圖2.2:當 0.0027,m25時,傳統S 管制圖對應不同 n 的函數2 ARL( ) 之圖形。

 

0.0 0.5 1.0 1.5 2.0

0100200300400

rho

ARL n=5

n=10 n=15 n=20

ρ

ARL=370.37 

  圖2.3:當 0.0027,n5時,傳統S 管制圖對應不同 m 的函數2 ARL( ) 之圖形。

0.0 0.5 1.0 1.5 2.0

0100200300400

rho

ARL m=25

m=50 m=75 m=100

ρ

ARL=370.37 

  圖3.1:當 0.0027,n5, m25之下,傳統管制界限與兩種修正管制界限的

整體ARL 函數ARL( ) 之圖形。

 

0.0 0.5 1.0 1.5 2.0

0100200300400

rho

ARL

traditional method adjusted alpha

overall ARL-unbiased

ρ

ARL=370.37 

  圖3.2:式(3.17)的函數圖形。

  圖4.1:R管制圖的修正管制界限範例。

 

  圖4.2:S管制圖的修正管制界限範例。

   

  圖4.3:S 管制圖的修正管制界限範例。 2

相關文件