Reaction to Glassonweb article from Miriam Schuster, Jens Schneider & Tuong An Nguyen “ Investigations on the execution and evaluation of the Pummel test for polyvinyl butyral based interlayers” (08/07/2022).
Dear Miriam, Jens, Tuong An,
Allow me to add some (mainly historical) remarks/observations on your detailed article about Pummel evaluation on laminated glass.
In the 70’s, customers (mainly automotive laminators) had been asking the PVB suppliers of these days for an easy and cheap adhesion evaluation test.
DuPont used at that time Compressive Shear Strength (PVB release as well as customer laminated glass control) and Monsanto used internally peel test for product release and CSS(?) for customers product evaluation. (both methods have been described in your article). Both tests are considered cumbersome, expensive and needing special sample preparation, not ideal for a quality control test (little preparation/representative for regular production, not needing sophisticated equipment).
I was not around (in the PVB/laminated glass world) at that time (1977) but learned the history in 1985 when I started as a young scientist amongst “old foxes”.
In 1977, R.Beckman from Monsanto came up with Pummel testing (U.S.Patent 4 144 376).
A picture of the PVB world at that time:
PVB laminated glass was used almost exclusively for the auto industry (glass thickness varied between 2.0 and 3.5 mm, PVB thickness had 0.76 mm thickness goal).
The request/goal had been to develop an understanding of ideal adhesion/impact performance (when is there a risk for long term delamination (based on adhesion) and when would one run the risk of a ball drop/impact failure (ECE R43 the 5 pound ball falling through the laminated glass (and headform test)).
At that time (1985) the Pummel ‘adhesion’ variability between Tin and air side of glass was about 1 Pummel unit. Today with thin automotive glass the adhesion difference is barely noticeable (also for thicker glass ?) due to the limited residence time on the tin bath during the float process.
I was challenged (as a new comer) to reduce the subjectivity and variability of Pummel test results…
We spent months/years and hundreds of samples trying to obtain good Pummel samples/standards and better understanding of the tests and variability parameters.
What did we learn (sorry but the test reports must be in DuPont or Kuraray archives, so this is based on my memory):
· To obtain a good representative Pummel sample/evaluation one needs a sample from 10 cm X 30 cm of which an area of 10 X 5 cm (at least) representing app. 10 hammer strokes X 4 rows.
· The variability is mainly caused by the evaluator of the Pummel samples (less by the executer of the Pummel test), we solved this by having 3 people reading the samples and in case variability was larger then 1 Pummel unit, a re-rating was performed (in case of disagreement the test was repeated).
This way we obtained a person to person variability of app. 0.5 Pummel unit.
· We have been working with spinning samples and reflected light in an attempt to automate and reduce the reading variability, but we had to stop because of lack of (financial) support.
· Other contributing variables:
· Glass thickness (becoming more and more important with increasing architectural laminated glass (the glass fragments tend to increase in size (with glass thickness) and potentially interfering with the read out of Pummel evaluation.
· PVB thickness: In architected laminated glass various PVB thicknesses are used (versus 0.76 mm goal for automotive applications).
0.38 mm PVB laminated glass tends to tear/break easily (when pummel testing at — 18 °C).
Over the years we developed correlations between Pummel (as executed by DuPont or Everlam) and CSS, for 3GO plasticized PVB (at the current ratio) we obtained the following limits:
Statistical risk for long term delamination (too low adhesion): lower then 1500 psi in CSS test or a Pummel of 2
Statistical risk for ball drop failure (too high adhesion): higher then 2500 psi in CSS or a Pummel of 5
( statistic limit of 4.5 sigma was used or 1 (failure) in 100 000 units).
Conclusion;
Pummel evaluation of laminated glass remains subjective and variable but on the other hand the tolerance between good and “bad” quality is fairly large ( 1500–2500 psi or Pummel 2–5).
Pummel can be used as process control (adhesion/impact performance) tool as long as one understands/controls subjectivity of the test.
