Meaning
Statistical indices measure the performance of a process when the output data does not follow a bell-shaped distribution. Calculating non-normal capability requires the use of transformations or specialized probability models to determine if a system can meet customer specifications. Standard formulas like Cpk yield misleading results when the data is skewed or multimodal.
Distribution Skew
Processes involving chemical reactions or human wait times often produce results that cluster at one end of the scale. The non-normal capability analysis accounts for this lopsidedness by using the median rather than the mean as a center point. This adjustment provides a more realistic view of the defect rate.
Analysis Method
Software tools apply the Box-Cox or Johnson transformation to force the data into a shape that can be measured. When assessing non-normal capability, engineers must identify the underlying distribution, such as Weibull or Lognormal, before drawing conclusions. Choosing the wrong model leads to an overestimation of process reliability.
Process Yield
High capability scores indicate that the system produces very few defects even under varying conditions. The non-normal capability report shows the parts per million that fall outside the tolerance limits. Improving this score requires fundamental changes to the process logic or material quality.
