Prerequisites for Successful Process Capability Cp, Cpk Analysis

  1. Measurement System: The measurement system must be capable of acceptable accuracy and precision.
  2. Order of Production: Always, ensure that parts are measured in the order of production, and associated with location if produced on multi-station tools (see #6 below).
  3. Data Independence: The data must be independent (i.e., not auto-correlated) of one another. In other words the next reading should not be a function of the current reading. Auto-correlated data is common when measuring variables such as temperature, where the next temperature reading is a function of the current temperature.
  4. Process Stability / In-Control: The process must be stable or in-control for process capability analysis results to be valid.
  5. Sub-Grouping: There are two ways to create sub-groups for process capability analysis. In either case, all factors that affect the output of the process (Man, Machine, Method, Measurement, Material, and Environment) must remain constant within the sub-group.
    1. You could measure 30 to 50 consecutive parts, and record the data using an Individuals & Moving Range chart (I-MR), or an Individuals & Moving Standard Deviation (I-MS) chart. In this case you have by default chosen a sub-group size of 2.
    2. Alternatively, you may choose to sample 5 parts every hour. In this case the sub-group size is 5, and the data is recorded on a Mean & Range chart (x-bar–R), or a Mean & Standard Deviation chart (x-bar-S).
  6. Manufacturing Process, Equipment, and Parameters: Understanding the manufacturing process will allow you to keep populations separate, and will prevent you from reaching incorrect conclusions. Multi-station tools (multi-spindle, multi-cavity molds, etc.), where each station operates under different conditions (temperature, speed, feed, etc.), will produce parts with different mean and variability. These parts are considered to come from different populations.
  7. Sample Size: Your choice of sample size and the variance of the capability metric will determine the confidence interval for the process capability index. The larger the sample size, the narrower your confidence interval for a given confidence level.
  8. Modification of Spec-Limits: You may have chosen to guard-band your process to accommodate measurement error. In this case, the guard-band limits are your specification limits for process capability calculations.
  9. Measurement Tools: You should be careful not to change inspection/test methods during a process capability analysis. Of course the gages should be calibrated, in-control, and capable (perform a Gage R&R). If you are comparing data provided by a supplier, both you and your supplier must use the same inspection/test method, with gages of the approximately the same accuracy and precision.
  10. Unmodified Data: Be careful not to truncate the data distribution by leaving out “out-of-spec” parts.
  11. Rounding-off Data: Rounding off data can lead to inadequate measurement units i.e., lack of necessary discrimination. The result of which is a histogram with too few distinct categories.
  12. Number of Distinct Categories: The number of distinct categories checks the ability of a measurement device to distinguish one part from another. After you have collected the data, plot a histogram to check the number of distinct categories. If you have less than 5 distinct categories in your data set, you may not be able to detect shifts in the manufacturing process.
  13. Interpretation: And finally, remember that a batch-by-batch application of the process capability index may not reflect the long-term variability of the process.