Control Charts for high-quality Binomial Processes

RIGHETTI A.

INGENIERíA E INVESTIGACIóN

UNIV NAC COLOMBIA

Lugar: Bogotá; Año: 2022

0120-5609

he use of the zero-inflated binomial model can be a very good alternative when analyzing data of the processes in which many conforming samples are observed, that is, many samples without non-conformities. This is especially true for the case where the data show a higher frequency of zeros than would be expected if the sample had been generated by a binomial distribution. Traditional-ly, these processes were monitored using the binomial distribution but, under these circumstances, the binomial distribution tends to underestimate the variability of the process. In this context, the control charts have very strict limits yielding to an excessive number of false alarm signals, high inspection costs, and frequent process stops. When the excessive number of zeros is not taken into account, an incorrect model is created and consequently, the resulting control chart does not comply with the function for which it was built. This paper proposes the use of the generalized linear model to establ