Statistical Process Control Accuracy Estimation of a Stamping Process in Automotive Industry

摘要

The goal of this paper is to unravel a quality problem, mainly the quality improvement of a process by employing statistical methods. The normality test utilized in the case is cumulative frequency distribution with regression analysis. The statistical process control (SPC) technique control charts reveal that the process is centered and meets the acceptance criteria and the regression analysis reveals that the recorded data follow a normal distribution. In this paper the results of the K-S, Anderson-Darling and Shapiro-Wilk tests are analyzed and discussed since these tests are, according to the literature more powerful statistical tools for detecting most departures from normality. In order to estimate more accurately if the tested data came from a normally distributed population, three goodness of fit tests were performed on the same collected values: Kolmogorov-Smirnov, Anderson-Darling and Shapiro-Wilk. Then, the results were analyzed, discussed and compared. Since the tests are more sensible to detect most departures from normality, they allow a more accurate assessment of the collected data - in turn increases the confidence in the control chart.