Process capability indices for skewed process data

A heuristic approach of constructing new process capability indices, based on weighted variance control charting method, without any assumption on pop ulations is presented. This method adjusts the values of the capability ind ices in accordance with the degree of skewness and kurtosis estimated from the sample data considering the variation above and below the target value separately. For the symmetric populations, however, these capability indice s would be equivalent to the normality-based capability indices. A simulati on study was conducted to evaluate the robustness of this new approach unde r a variety of non-normal skewed process data when the sample sizes are 75, 100, 150, and 200, and the Johnson-Kotz-Pearn method is compared under the same situation as well. The results show that this heuristic approach perf orms better in evaluating process fallout when the underlying distribution belongs to lognormal and skewed unbounded distributions. Significance: This paper proposed new process capability indices according to the degree of skewness and kurtosis in dealing with non-normal skewed pr ocess data. The results show that this new approach performs well compared with the Johnson-Kotz-Pearn method in lognormal and skewed unbounded distri butions.