12 SHARAD S.PRABHU AND GEORGE C.RUNGER TABLE 2.Optimal Multivariate EWMA Controi Scheme P=4 p=10 p=20 ARLo= 500 500 1000 500 100 05 02 10 9 14.60 器 1.5 A 器 2.0 器 30 Note:ARLo and ARLmin are zero-state average run lengths. the MEWMA and an efficient algorithm.that avoids variables (r=0.30)is slightly smaller than the 0.40). Design of a MEWMA variables. ovidess the the corres and fo of the number of variables the on-target ARL,and =2.These are zer o-state run length edonthe7rotae The zero-state ARL is the ARLc used Consequently,the optimum ues for5 ed to calculations are sensitive to the umber of states used o e on-target case From Table 2,an r less than 0.05 is preferred to magnitude 6=0.5.Although the presents the zero-state ARL of MEWMA an 0.03 migh mately=,r=0.1 is close to the best choice over for each row in Table 1,there 0=1.5t0 onig the as large as6=3.an r of 0d is Although it is useful to obtain insights from a The steady-stat Journal of Quality Technology Vol.29,No.1.January 1997 Reproduced with permission of the copyright owner.Further reproduction prohibited without permission.Reproduced with permission of the copyright owner. Further reproduction prohibited without permission