Abstract :
A screening methodology based on measurements of several parameters is proposed. The methodology provides an improved semiquantitative basis for the selection and evaluation of screening criteria The method is devised 1) to yield a minimum number of parameters required for effective screening with a linear function, 2) to determine the gain in reliability obtained by screening on the basis of two parameters rather than one, three rather than two, etc., 3) to determine the parameters which may be interchanged for measurement or cost reasons without changing the effectiveness of the screening procedure, 4) to determine the probabilities of screening out a superior component and failing to screen out an inferior component, 5) to determine the costs associated with making the screen:rig procedure more stringent, 6) to permit modification of the screening criteria for small changes in component-part design, or lot characteristics; 7) to determine the effect of alternative failure criteria on the screening criteria, and 8) to indicate when practical screening cannot be achieved using a linear function of the parameters selected for measurement. The methodology is based on a combination of standard statistical techniques, and is novel only in maintaining a tractable analysis of the over-all problem of screening individual component parts by variables inspection.
