A process capability calculator tells you whether a process can meet its specification. Enter the upper and lower specification limits and the mean and standard deviation of the process, and you get Cp, Cpk, and with a target Cpm and Cpkm, an estimate of the defects and a comparison with the usual minimums. A second mode gives the control chart limits, the upper and lower control limits for the process. It runs in your browser.
How to use the process capability calculator
- Choose process capability or control chart limits.
- Enter the specification limits, the mean and the standard deviation.
- Add a target, or a subgroup size for control limits, if you want them.
- Press Calculate.
Cp and Cpk
The Wikipedia article on the process capability index gives Cp = (USL − LSL) ÷ (6σ), which compares the width of the specification with the spread of the process, and Cpk as the smaller of (USL − μ) ÷ (3σ) and (μ − LSL) ÷ (3σ), which also takes account of how well the process is centred. It adds that Cpk is negative if the process mean lies outside the specification limits.[1] Cp shows what the process could do if it were centred, and Cpk what it does. If only one limit applies, the calculator gives the one-sided value.
A worked example from the page
The page's example has a target of 100.00 micrometres, an upper limit of 106.00, a lower limit of 94.00, a process mean of 98.94 and a standard deviation of 1.03. Cp is 1.94, Cpk is 1.60, Cpm is 1.35 and Cpkm is 1.11. The page remarks that the process runs about one standard deviation below its target.[1] Enter these numbers to see the same results. Cpm is Cp divided by √(1 + ((μ − T) ÷ σ)²), where T is the target, and Cpkm is Cpk divided by the same factor. They fall when the mean drifts from the target even if it stays inside the limits.
How big should it be
The page lists the recommended minimums for Cpk, and says that they are a matter of personal opinion, citing one academic expert. For an existing process the value is 1.33 for a two-sided specification and 1.25 for a one-sided one. For a new process it is 1.50 and 1.45. For a safety-related or critical existing process it is 1.50 and 1.45, for a new one 1.67 and 1.60, and for a Six Sigma process 2.00. The page adds that a value above 2.5 may mean unnecessary precision, which can be expensive.[1] The calculator reports the highest minimum that your Cpk meets. In the example, Cpk is 1.6, which meets 1.50 but not 1.67.
From the index to defects
The page has a table that links Cp to the sigma level and to the defect rate for a short-term, centred process: 1.00 is 3 sigma and 2,700 parts per million, 1.33 is 4 sigma and 63 parts per million, 1.67 is 5 sigma and 1 part per million, and 2.00 is 6 sigma and 0.002 parts per million. With the usual 1.5 sigma shift for the long term, the rates are 6,209 for 1.33, 232.6 for 1.67 and 3.40 for 2.00.[1] The calculator estimates the defects directly from the normal curve for your mean, standard deviation and limits, which is our own step. For the example it is about 0.8 parts per million.
Control chart limits
The article on control charts says that the centre line is at the mean or median of the plotted statistic, and that the control limits are typically drawn at three standard deviations from it. With the process in control, 99.73 percent of the points are expected between the limits, and a Shewhart chart gives a false alarm about once in 370.4 points.[2] The limits have no intrinsic link to the specification targets or the engineering tolerances, so they must not be treated as specification limits.[2] The page also mentions optional warning limits at two standard deviations.[2]
Limits for averages
For a chart of the averages of subgroups of n values, the limits are the mean plus and minus three standard errors, where the standard error is the standard deviation divided by √n. This is our own working, since the pages we read do not give the formula. For a mean of 50, a standard deviation of 2 and subgroups of 4, the limits are 47 and 53, while for single values they would be 44 and 56. The calculator gives both, with the 2-sigma warning limits.
Using the results
If Cp is high and Cpk is much lower, the process has enough precision but is off centre, and the fix is to adjust the mean. If both are low, the process varies too much, and you need to reduce the variation, for example by better materials, tools or procedures. Track the indices over time, and recalculate them after any change. Remember that the indices describe a stable process, so first check, for example with a control chart, that the process is in control.
A one-sided example
A strength test only has a lower limit, which is 50 units. The process has a mean of 58 and a standard deviation of 2. Then Cpk = (58 − 50) ÷ (3 × 2) = 1.33, which just meets the minimum of 1.25 for an existing process with a one-sided specification. Leave the upper limit empty to get this result in the calculator.
Common mistakes
- Using the specification limits as control limits. They are different things.
- Quoting Cp alone. It ignores an off-centre process.
- Using the range, not the standard deviation. Enter the standard deviation.
- Applying the indices to an unstable process. Get it in control first.
What next
Use this often? Add Process Capa to your Home screen.
Frequently asked questions
What is a process capability calculator?
A tool that finds Cp, Cpk and related indices that show whether a process can meet its specification limits.
What is the formula for Cpk?
Cpk = min[(USL − mean) ÷ (3σ), (mean − LSL) ÷ (3σ)].
What is a good Cpk?
Commonly 1.33 or more for an existing process, 1.50 for a new one and 1.67 for a new safety-critical one.
What are control limits?
Lines drawn at three standard deviations from the centre line of a control chart. About 99.73 percent of points fall between them for a stable process.
Are control limits the same as specification limits?
No. Control limits describe the process, and specification limits come from the customer or the design.
Was this tool helpful?
What went wrong?
We only receive the reason you pick, never what you typed into the tool.
Thanks for your feedback.