Machine capability
Calculate Cm and Cmk, even when the characteristic is not normally distributed
my8data calculates machine capability with the normal, folded normal, Weibull or mixture distribution. For non-normal characteristics the quantile method from ISO 22514-2 applies, together with the calculated defect rate in ppm.
- Included in the free account
- No installation, servers in Germany
- Order on invoice, no credit card
| Cm and Cmk | Assessment |
|---|---|
| ≥ 1,67 | capable |
| 1,33 to 1,66 | conditionally capable |
| < 1,33 | not capable |
Common limits. You set which ones apply per analysis, for example per customer requirement.
Result
This is what the result looks like
Position of a bore, 50 parts, upper limit 0.04 mm, calculated with sample data. The Anderson-Darling test rejects normality; the folded normal distribution fits in the probability plot. Cmk 1.82 by the quantile method.
Distributions
The right model for every characteristic
Not every characteristic varies symmetrically. With the wrong model, the Cmk turns out too good or too bad.
Evaluation
What you get
For non-normal characteristics my8data calculates with the quantiles of the fitted distribution instead of 6s. Each side of the tolerance gets its own half of the distribution.
If there is only a maximum, the other field stays empty. my8data then calculates only the one-sided index; an empty field does not count as zero.
Distribution parameters and the quantiles X0.135 %, X50 % and X99.865 % are part of the result. Every index can be retraced by hand.
The fitted distribution tells you how many parts to expect above and below the limits. Histogram and probability plot show whether the model fits.
For the running process: process capability with Cp, Cpk, Pp and Ppk.
Checked against reference values
The folded normal distribution is checked against the formulas in the Minitab documentation and the parameter estimation by Leone, Nelson and Nottingham (1961); the indices against ISO 22514-2. The reference data sets with target and actual values are in the public validation report.
Pricing
Every method in every plan
All prices excl. VAT. 12-month term, cancellable with 30 days’ notice. Order by PDF form on invoice.
- All methods
- Unlimited calculations
- PDF and Excel export
- Save analyses
- Everything in FREE
- PDF report from your own Excel template
- Excel export with data and results
- Save and reopen analyses
- Personal support from the developer
- Everything in PRO, for every user
- Shared team storage
- Seats transferable when staff change
FAQ
Questions about machine capability
How many parts do I need for a machine capability study?
The usual number is 50 parts produced one after another under the same conditions. Fewer parts make the index less certain.
What do I do if the characteristic is not normally distributed?
Choose the matching distribution: folded normal for magnitudes such as position or unbalance, Weibull for skewed characteristics, mixture for multimodal data. The index then follows the quantile method from ISO 22514-2.
Does it work with only one tolerance limit?
Yes. Leave the missing limit empty. my8data then calculates only the index for the existing limit and does not treat the empty field as zero.
What is the difference between Cmk and Cpk?
Cmk describes the machine alone, from a short run under the same conditions. Cpk and Ppk describe the running process over a longer period with all influences.
What does it cost?
The calculation is included in the free account. PDF and Excel export and saving cost €490 per user and year excl. VAT with PRO.
Check it with one of your own data series
Create a free account, paste the readings of a machine acceptance and compare the Cmk with your current evaluation.