‘On‐line’G‐control chart for attribute data

Publisher Website
Google Scholar
Abstract
On the basis of thetheory of random processesthe concept of aG‐chart is elaborated. In this case the observed variableGis a number of conforming units between two consecutive appearances of non‐conforming ones. If the process is a Poisson one thenG‐variables are geometrically distributed (theG‐chart is called after the distribution type). Application of a new SPC concept for attribute data makes it possible to improve SPC employment for:high quality processeswith ṗ < 10−4(100 ppm);low volume manufacturing, short runsand‘stepped’ processes.In the paperindividual G, G‐barandstabilized G/Ḡcharts are presented. The sensitivities of aG‐chart and a classicalp‐chart for the detection of process changes are compared by constructing operating characteristic curves and ARL curves. Depending on the required degree of the detection process changes, an optimal size of subgroup is found. For this size of subgroup the average number of non‐conforming units and the average number of observed units between the process change and this change detection are minimal. Compared with the classicalp‐chart of the same sensitivity, theG‐chart requires on the average fewer observed units for process change detection and also on the average fewer non‐conforming units are produced.

This publication has 2 references indexed in Scilit: