In a space equipment testing site, internal combustion occurred in a liquid hydrogen tank (H2) under a high pressure of 12 m³ (700 kg of liquid H2). The test was stopped as the malfunction was detected by the indicators. No consequences were reported. The tank commissioned in 1988 was connected to H2 gas and liquid oxygen (O2) storage tanks. The tank structure included a double casing separated by vacuum. The multilayer inner wall was made of austenitic stainless steel and 157 mm thick. Its maximum operating pressure was 400 bar, with a service pressure of 250 bar.

In July 1991, the test tank was emptied and purified (by injecting helium) as the H2 gas network was suspected of being polluted in October 1990 following an intake of air from a compressor.

The day of the accident, a test was performed: the liquid H2 tank was pressurised by a control device (+ 250 bar in 30 sec). While the pressure in the tank reached 160 bar, a noise indicating a pressure drop was observed and the detectors recorded a pressure fluctuation of 10 bar. The sudden and significant fluctuations in the measured parameters (temperature of wall, pressure, flow) led the test to be stopped. The combustion of an O2 pellet in an abandoned pipe of the tank’s drain pipe was responsible for the accident. The O2 pellet was formed by the regular addition of air due to the impurities present in the H2 and the accidental pollution of the network in October 90, the presence of air in the abandoned pipeline during purification in July 91 (helium is lighter than air), re-solidification of the air while refilling the tank with liquid H2. The abandoned pipe was thus obstructed by the air pellet causing a mixture of air and H2 to be trapped in its lower portion.

Several combustion spots were detected: in condensed phase between solid O2 and H2 gas, in less dense phase between H2 and O2 gas. The combustion was undoubtedly triggered by the friction of the pellets against the walls and the adiabatic compression of the H2/O2 gas mixture. The tank was re-commissioned 4 months later after investigation and implementation of a bleed valve to regularly drain the abandoned pipe and avoid the accumulation of O2 particles.