Statistical process control of fire-resistant coating production based on thickness measurements

U radu je provedeno statističko praćenje procesa nanošenja protupožarnog premaza. Statistički promatrano radi se o procesu samo s jednom, donjom, granicom specifikacije. Upravo se takovi procesi, vrlo često, zbog svoje prirode ne ponašaju po normalnoj razdiobi. U cilju dovođenja procesa pod kontrolu, proces nanošenja protupožarnog premaza je praćen x-R kontrolnom kartom. U radu je provedena analiza sposobnosti procesa na podacima aproksimiranim Weibullovom 3 – parametarskom razdiobom, podatcima transformiranim Johnsonovom transformacijom te je za usporedbu procjena provedena na podatcima aproksimiranim normalnom razdiobom. U okviru eksperimentalnog dijela provedeno je mjerenje debljine protupožarnog premaza na sedam čeličnih nosača. Rezultati su analizirani probnom verzijom programskog paketa Minitab. U cilju procjene mjerne nesigurnosti rezultata mjerenja debljine premaza primijenjena je metoda Monte Carlo simulacija.This article focuses onto the process of fire-resistant coating production. That process belongs to the class of processes, considered statistically, which have only one, lower specification limit. Such processes are interesting in that they usually do not have the accompanied normal distribution. In order to bring the process under control, we monitored the fire-resistant coating production using the x-R control chart. The process capability analysis was applied on one set of data analyzed using three different methods. In the first method the data were approximated using the 3-parameter Weibull distribution. In the second method, data were transformed using the Johnson transformation while in the third method, applied for the purpose of comparison of sets, data were approximated with normal distribution. In the accompanied experiment, the thickness of the fire-resistance coating was measured on seven steel girders. The results have been analyzed using the trial version of the Minitab software package. Using the Monte Carlo method, numerical simulations were performed in order to estimate the uncertainty of measurements of coating thickness