Though the content is dramatically reformated, I don't really see anything of much substance changed.
The control chart methodologies are the same. Instead of repeating the control chart process for each type of data, they give an overall methodology and then give the formulas for each type of data. It shortens this section by quite a bit to make room for all the other new stuff.
The capability metrics are essentially unchanged, but has a greatly expanded description of the metrics and their use. One thing new is that this version allows use of sbar estimation of the variance for Cp & Cpk (the older version only mentioned use of Rbar). There is still no mention of use of pooled variances. The coverage of this topic is much easier to read/understand.
There is a section called "Suggested Use of Process Measures" which says "It is strongly recommended that all four indices (Cp, Cpk and Pp, Ppk) be calculated on the same data set. The comparison of the indices amoung themselves can provide insight to potential process issues and aid in measuring and prioritizing improvement over time." Each metric has its own story to tell  Ppk is the actual performance, Pp is the "capability" if the process was on center, Cpk is the capability if instability was removed, and Cp is the capaibilty if instability was removed AND the process was on center.
The Second Edition finally cleary states "If the process is in statistical control the process capability will be very close to the process performance. A large difference between the capability and performance estimated sigma indicates the presense of special cause(s)." It is about time.
The nonnormal section starts by pointing out the relationship between Cpk, Ppk and the proportion nonconforming (binomial). It then provides two nonnormal methodologies, none of which are new: (1) Transformations & (2) NonNormal Forms (use of statistical distributions other than "Normal").
For transformations they specifically mention BoxCox and Johnson transformations.
For use of nonnormal forms (other distributions) they say to identify the distribution of best fit and then calculate:
Pp = Spec Range / (Q0.99865  Q0.00135)
where Q0.99 would represent the 99th percentile (Q=quantile; 99% of the population is less Q0.99).
It says that "Cp is calculated as above replacing s with Rbar/d{2}". but I personally don't see where the s is in the Pp formula. I wouldn't have thought Cp was defined when using the percentile (quantile) method.
They do say that there is no analoque of a nonnormal Cpk available.
Ppk = min[(USLxbar)/(Q0.00865xbar) , (xbarLSL)/(xbarQ0.00135)]
