presence of cracks and packed for shipment. This final test just prior 
to delivery of the windows to EDU was intended to remove any remaining 
doubts about the quality and safety of the supplied windows. 
EXPERIMENTAL TEST PROGRAM 
Testing Arrangement 
The experimental test program for evaluation of the chosen window 
design for EDU consisted of testing to destruction under hydrostatic 
pressure a series of EDU windows. While the type of loading differed 
from test to test depending on whether the tests were of short-term, 
long-term, or cyclic nature, the method of loading and the test arrange- 
ments were the same in every case (Figure 4). 
The 9-inch diameter NCEL pressure vessels were used in every case 
for the containment of windows. The pressure was raised with positive 
displacement air operated pumps at 650 psi/minute rate. For long-term 
tests the desired pressure level was maintained inside the vessel by 
closing valves leading to the vessel. Only periodically were they 
opened to adjust the pressure if it deviated more than 50 psi from the 
desired pressure setting. During cyclic tests the sustained pressure 
was maintained for 7 hours followed by depressurization proceeding at a 
rate equal to the pressurization rate. The depressurization was followed 
always by a 17-hour long relaxation period. The overall 24-hour length 
of the cycle was patterned on a typical working day. 
To eliminate as many extraneous variables as possible from the tests, 
the windows rested on a 0.025-inch thick nylon fiber reinforced gasket 
(DuPont's Fairprene 5722A) and no retaining rings were used for clamping 
the windows inside the test flanges. The sealing was accomplished by 
placing a bead of room temperature curing silicone rubber around the 
circumference of the window. 
Test Specimens 
Test specimens were windows selected at random from the lot supplied 
by the manufacturer for installation in the EDU test chamber complex. 
All of the tests except for 6 short-term tests were conducted for economy 
with the small (4.450 x 1.040 inches, t/D; = 0.346) windows. The 6 short 
tests were conducted with the large windows (6.950 x 1.650 inches, t/D; = 
0.330) to determine whether there was a substantial difference between 
the strengths of the large and the small windows. Also for economy only 
one window was tested for each of the many chosen long-term and cyclic 
loading conditions making any subsequent statistical reliability analysis 
of data impossible. 
Clamping sometimes tends to strengthen the windows. Testing unclamped 
windows always produces conservative data. 
