A couple of points:

I stand corrected. While the thermal expansion coefficient for glycerin is an order of magnitude less than the thermal expansion coefficient for air, the volumes are significantly different. Thus, you are correct that it is the compression of the air bubble that matters as opposed to the expansion of the air bubble.

In prior publications, the root cause of the unwanted activation were the result of physical damage to the sprinklers. It was found that installers were unpacking the sprinklers and putting them into buckets (e.g. "five gallon buckets") prior to installation. The movement of these buckets caused physical damage to the sprinkler arm and perhaps the glass bulb also. Movement of the arm would produce an added stress to the glass bulb that was not present before the sprinklers were removed from the shipping box" or anticipated in the design and testing of the sprinklers. Scratches could be part of the physical damage.

While there may be no specific "study" on this issue, the application of fundamental knowledge can provide significant insight as to a specific problem.A quick check of "hardness" shows that brass (3 to 4) is softer than "glass" (4.5 to 6.5). "Steel" is harder than brass, so the scratches might occur with stainless steel arms than for brass arms. I doubt that many reasonable, practical, or recommended cleaning methods would produce contact of a materials that is "harder" than glass. A wire brush could induce scratches on the glass, but I am unaware of the use of such as a recommended practice.

Douglas J. Carpenter, MScFPE, CFEI, PE, FSFPE
Vice President & Principal Engineer
Combustion Science & Engineering, Inc.
8940 Old Annapolis Road, Suite L
Columbia, MD 21045
(410) 884-3266
(410) 884-3267 (fax)
www.csefire.com