by Bill Inman, Member, Reaction Research Society

A few of us met at the Mojave Test Area on Saturday, January 15, 2022, to conduct an elevated temperature burst test of a 5-gallon (20-pound) propane container partially filled with water, Dave Nordling was the pyrotechnic operator in charge for this event, The objective was to determine practical limits for use as steam rocket vessel. This was an extremely dangerous task and having only those necessary to conduct the test was prudent.

The 20-lb propane bottle that was to be failure tested was an old surplus asset retired from serivce, The capacity of this particular propane fuel cylinder was measured at 46.4 lbs. of liquid water (1285 cubic inches). Propane containers by regulations are only filled to 80% of their liquid volumetric capacity. The water fill level for this test, since higher temperatures than reached by normally reached in steam rocket operations were anticipated was only 4 gallons or 71% of the 5.57 gallon total to provide more internal room for thermal expansion.

Filling of the vessel was done through a reducer bushing in the factory opening via a siphon tube from water bottles. Heating of the sealed vessel charged with water was done by a propane fired turkey fryer burner, The burner was positioned directly underneath the center of the bottle which was propped up by a metal frame. The positioning of the burner was both by eye and by feeling for the weld seam running around the middle of the tank. The propane fuel hose and pull cable to remotely pull away the burner if necessary were both on the right side as viewed from the blockhouse. Automotive brake line was used to connect a pressure gauge for visual readout at a distance and manual ball valve on a tee to allow remote venting of the setup if the need arose, Mechanical pull cables were carefully routed back to the blockhouse. All mechanical control devices were tested and safe operation verified before starting the heating process.

The test article on the north side of the vertical test stand just behind the large I-beam. The pressure gauge and the video camera recording the gauge readings were on the opposite side of the I-beam and all controls were remotely handled from the blockhouse.  Due to the expected destruction of most of the test stand and related components, everything was kept minimal, with no planned provision for securing it beyond the clamps holding the sheet metal shroud in place over the test article.

All operations went smoothly and everyone was safely secured in the blockhouse. The heating rate from the turkey fryer bunrer was somewhere over 100,000 BTU/hr based on literature which was sufficient for a steady increase in pressure which took just a little longer than 45 minutes. The propane container used to feed the burner had sufficient fuel to last 2 to 3 hours in effect limiting the test if no action was taken once the burner was lit. As long as all parties remained safely behind cover and at a safe distance, we only had to wait. If somehow, the vessel failed to burst and the manual valve would not open remotely by the mechanical pull cable from the blockhouse. the test article would be left untouched and we would allow 24 hours for the vessel to return to ambient temperature.

Listed design burst pressure was about 960 psig based on a four-fold safety factor of the nominal rating of 240 psig for propane service. As these containers are meant for public use and rougher handling at campsites, they are likely way over-designed. Conversely, these containers often get dented, abused and corroded over time. The exact failure point on any given vessel is not easily determined from so many uncontrolled factors. According to the graphs in the report of a testing program commissioned by the propane industry (1), the average as-tested burst pressure of used cylinders of this type appeared to be in the 1250 to 1600 psig range. But these vessels aren’t normally actively heated which makes analysis less certain, thus the reason for this testing program, and this being the first test. The cylinder chosen for this test had a fair amount of rust evident around the bottom and a few inches up from it. For the next test, a cylinder in better condition will be used to see how they compare.

The vessel failed at 1135 psig in a sudden violent burst a little above the expected burst pressure but within the 1500 psig range of the gauge. The pressure wave was enough to shatter the row of cinder blocks put beside the test article to block the wind. The test article ruptured at the weld seam. The metal support structure was blown apart, converted into twisted pieces of steel, and the sheet metal shroud was shredded in numerous pieces scattered all across the area and crumpled up like aluminum foil. Some parts, including the largest piece of the tank weighing 10.5 pounds, were found almost 100 yards away. The I-beam deflected the debris away from the occupied blockhouse, but the shockwave, which was felt by us inside the blockhouse, managed to break one of the windows in the Dosa Building. The blockhouse with the blast windows continues to be a useful asset for the society.

Preliminary answers to questions going in (pending confirmation of these results in a follow-on test):

What is the “real-world” burst pressure of a retired propane cylinder on it’s first use as it would be for a steam rocket motor, and would it be significantly less than that of a cylinder used only in normal propane service?

ANSWER: 1135 psig, and apparently some less (approximately 20.5%), although this one was not in pristine condition, either.

Is the prediction that it will fail along a seam (weld) true?

ANSWER: it appears to be, as there was a long tear along the seam, although there were many other tears in quite a few other locations as well. Six pieces were recovered but there is still over a pound missing compared to the starting empty weight, meaning there are at least seven pieces.

Will the area the burner flame impinges upon be weakened more than the rest of the tank?

ANSWER: It appears so, as the area where the longitudinal tear and the tear along the seam intersect shows evidence of the paint being more completely burned away than elsewhere. But again, there were many additional tears as well, so not sure exactly how to factor that into the analysis – and would an arrangement to keep the burner moving back and forth while heating reduce any such tendency?

Reference 1: National Propane Gas Association, Final Report on Testing and Assessment of CG-7 Pressure Relief Valve and Propane Cylinder Performance, Volume One: Results and Evaluation, January 31, 2003, by D. R. Stephens, M. T. Gifford, R. B Francini, and D. D. Mooney