THE L3 SYSTEM QUALITY CONTROL IN MANUFACTURE 979 



same essential conditions with respect to materials, parts, workmanship 

 and processing. The criteria of this method apply to a series of batches 

 or lots of product arranged in the order of their production and are based 

 on the use of control charts for averages and standard deviations for 

 samples of 50 imits each. 



The control limits for averages are for 50 units of product and also 

 include an allowance for the long time variation in the level of the process 

 average. They are represented by a band N ± 0.23 A which is equal to 

 the customary^ 3-sigma control limits for averages of 50 units placed 

 outside a band corresponding to iV dr 0.1 A (N zb Ko-) and are desig- 

 nated in the general specification as "C" limits. In this method the ''E" 

 limit in the specification is the control limit for the standard deviation 

 (a) of a sample of 50 units and equals 0.41 A. 



For this method a batch, represented by the sample, is considered 

 conforming if the sample meets the following criterion. 



(a) The average X, falls within the C limits. 



(b) The standard deviation, <t, either: 



1. falls within the E Umit; or 



2. falls outside the E limit but at the same time all of the 6 pre- 

 ceding consecutive standard deviations fall within the E limit. 



2.33 Three-Cell Method 



This method is for application whenever the product fails to meet the 

 criteria of the first two methods or where the manufacturing conditions 

 dictate its use. When maximum and minimum limits are specified the 

 method consists of testing or measuring each unit of product and classi- 

 fying conforming units in one of three cells : Lower, Middle or Upper. 



The distribution is shown graphically in Fig. 6. Groups of 5 units of 

 product are selected from this classification so as to meet one of the two 

 distributions given in Table I and maintained as such for shipment. 



In order to provide a graphic record of whether or not the manufac- 

 turing process is in statistical control, this method includes procedures 

 for obtaining control chart data based on samples of product taken from 

 regular production before classification. 



2.34 Records Method: 



This method is for application when it is not practicable to impose the 

 requirements of the other methods but it is still desirable to systematic- 

 ally maintain a graphic history of the measurements made on a 

 quality characteristic. Action required to maintain the distribution of 



