| In many industries there are processes that create products. These products must be made to certain specification limits so the products are not defective. Process capability indices are use to ensure the processes making these products are within certain specification limits. To calculate the capability of a process three assumptions must be met: (1) the process follows a normal distribution, (2) the process is in statistical control, and (3) the process mean is centered between the specification limits. Of the three assumptions the first assumption is the hardest to meet since data rarely follows a normal distribution. This paper explores using normal mixtures to approximate the distribution of the data. Process capability indices are then estimated for each normal model in the mixture and the weighted total and median are compared to see which does the best job of estimating the process capability. |
