Using Double Sampling Inspection in a Manufacturing Plant 



By D. B. KEELING and L. E. CISNE 



THE necessity for quality control in a manufacturing plant arises from 

 the fact that all units of product cannot be made identical. To limit 

 variations and attain controlled uniformity some sort of inspection must be 

 established. It has been the experience of the Western Electric Company 

 that quality control may be attained most economically by the use of a 

 sampling inspection wherein only a portion of the entire output is examined 

 for desired quality characteristics. 



Advantages which have been gained through the use of sampHng inspec- 

 tion, and with no adverse effect on previously existing quality levels, are: 

 a reduction in the cost of inspection by economies in inspection time; a 

 reduction in the amount of scrap produced by making available for super- 

 visory action useful records of the results of inspection; and as an end result, 

 the attainment of uniform quality of a satisfactory level. 



It is the purpose of this paper to provide a detailed method of procedure 

 that has proved successful in establishing and maintaining one type of 

 sampling — the "Average Outgoing Quality Limit" Double Sampling Plan. 

 Statistically determined tables of lot sizes and corresponding sample sizes 

 which guarantee a certain degree of protection have been used by the West- 

 ern Electric Company for approximately fifteen years. They were fur- 

 nished by the Bell Telephone Laboratories and have recently been made 

 generally available in an article published in the January 1941 issue of the 

 Bell System Technical Journal.^ A typical sampling table, reproduced 

 from the preceding article, is shown in Fig. 1. 



Briefly stated, the AOQL Double Sampling Plan involves the examination 

 on a "go — no go" basis of a specified number of articles taken at random 

 from a large group. The acceptance or rejection of this group is usually 

 made on the basis of results obtained from the first sample alone. How- 



1 The AOQL Double Sampling Tables specifically referred to in the present article are 

 Tables DA-0.1 to DA-10.0 shown on pages 49-61, inclusive, of a preceding article: H. F. 

 Dodge and H. G. Romig, "Single Sampling and Double Sampling Inspection Tables", 

 Bell Sys. Tech. Jour., pp. 1-61, Jan. 1941. These tables give sample sizes and allowable 

 numbers of defectives for a variety of AOQL values, lot sizes, and process average values. 

 The sampHng table reproduced in Fig. 1 is based on an AOQL value of L5 per cent defec- 

 tive. In tables prepared for shop use, it has been found preferable to use a notation 

 slightly different from that shown in Fig. 1, specifically to use AN instead of c to represent 

 "allowable number of defectives" and to use SS instead of n to represent "sample size". 

 The shop notation is used in the present article. It should be noted that, in the original 

 article, C was generally referred to as an allowable number of "defects", since primary 

 attention was given to inspection of a single characteristic. See footnote 2 for explana- 

 tion of the terms "defective" and "defect". 



37 



