INSPECTION IN MANUFACTURING PLANT 45 



accumulated since the last computation should be used. When it is 

 known that the quaUty of the product has changed significantly during 

 the period, use only the data collected since the change. 



Record the results of all process average computations and the periods 

 covered by them in the space provided on the sampling layout. 



1.42 Periodic Totaling of Lot by Lot Record and Posting on Summary of 

 Residis Form, Fig. 4 



The following columns on the Lot by Lot Record should be totaled 

 at intervals, and may, for contrast, be marked in red : 



List of Defects 

 No. of Lot hi Sample 2nd Sample in 1st Sample 



Lots Size SS Def. Total Total by Columns 



XX XX X XXXXetc. 



Totals after approximately 20 entries are generally considered satis- 

 factory. The required totals are then posted to the Summary Form, 

 Fig. 4. 



L43 Per Cent Defective 



The purpose of this figure is to show the average quality of the product 

 as received by the inspector during the period covered and is obtained by 

 dividing the total number of defectives found in first samples by the total 

 number of articles inspected in first samples, and multiplying by 100. 

 Only the results of the first samples are used, in order to accord equitable 

 treatment to all lots. 



1.44 Control Chart 



The per cent defective is plotted on the graph at the right of the sum- 

 mary form (Fig. 4). Control Limit Lines are drawn around the process 

 average to indicate the variation that may be expected due to random 

 sampling. These control limits are determined by the following for- 

 mulae^: 



Upper Control Limitj^ _ j_ 9 ^ /P (^ ~ P) 

 for fraction defective! 



p+V^- 



Lower Control Limit'*! _ - _ ^ /p(l ~ P) 

 for fraction defective j P " ^ n 



^ The considerations involved in the establishment of control chart limits are discussed 

 in ASA Standards Zl. 1-1941 and Zl. 2-1941, "Guide for Quality Control and Control 

 Chart Method of Analyzing Data". In this case, 2-sigma limits have been chosen. For 

 this particular application within the manufacturing plant over a period of 3'ears, this 

 choice has appeared to strike an economic balance with respect to the net consequences of 

 two kinds of "errors" that may occur in practice; namely, looking for trouble that does not 

 exist and not looking for trouble that does exist. 



^ If this result is negative, the lower control limit is to be taken as zero. 



