California Institute of Technology

Explosion Dynamics Laboratory

Explosion Research at Caltech and TWA Flight 800

Background | Facts | Documents | 1/4-Scale Experiments | Misconceptions |

Aviation Kerosene (Jet A) Combustion

"I thought that truth must eventually prevail, but I have found silent truth cannot withstand error aided by continual propaganda." - Orville Wright

The Misconceptions:

Unfortunately, misinformation about aviation kerosene combustion and misrepresentations about the official investigation of TWA 800 and the investigators have been published in print and on the worldwide web. The most common of these erroneous claims are discussed below:
You can put a match out in Jet A at room temperature so it can't possibly explode inside an aircraft fuel tank.
False. You can do this with many fuels that have sufficiently high flash or fire points. At room temperature, a combustible liquid fuel has enough heat capacity to absorb all the energy and extinguish a small match flame without raising the fuel surface temperature above the fire or flash point. A flame cannot be sustained over a liquid fuel until the surface temperature exceeds the fire point. A flash will not result unless the fuel temperature and vapor space are above the flash point. [Kuchta and Clodfelter] This does not apply to the case of TWA 800 because a) the explosion occurred in the fuel vapor-air mixture and did not require a fire on the fuel surface, b) the fuel and air inside the tank were hotter than the flash point, and c) the effective flash point was lowered due to the lower outside air pressure at the altitude of the explosion. The Jet A involved in the Center Wing Tank (CWT) explosion had a flash point temperature of about 115 F and the decreased air pressure at the explosion altitude of 13.8 kft lowered the effective flash point to about 100 F. [Exhibit20S] The temperatures inside the tank were between 100 and 130 F, and at some points, as high at 140 F [Exhibit 23F].
The CWT of TWA Flight 800 was completely empty of fuel.
False. It is impossible to completely empty the center wing using the fuel system due to the location of the scavenge pump, the nature of the tank bottom surface, the division of the tank into bays, and the arrangement of the fuel probes that supply the fuel quantity instrumentation system. At least 50 to 100 gallons of fuel will remain even when the fuel quantity system indicates no fuel is remaining. [NTSB Final report]
The CWT of TWA Flight 800 did not contain enough air to support combustion.
False. This is apparently derived from the notion that fuel tank ullages in automobiles are generally too rich to support combustion. Jet A is a much less volatile fuel than gasoline and the situation is very different in an aircraft fuel tank than an automobile. The fuel tanks in an airplane are vented directly to the outside atmosphere and each time the plane descends from cruising altitude, a fresh supply of air is brought into the tank ullage by the vent system. The amount of fuel added to this air by evaporation is determined by the vapor pressure of the fuel, which has a maximum level of about 1 to 1.5% of atmosphere pressure for a tank containing 50-100 gallons of Jet A heated up to 60C - a situation comparable to TWA 800. [Exhibit 20D] The rich flammability limit for Jet A is probably above 6% by volume, far higher than any possible conceivable fuel concentration that would have existed in the TWA 800 situation. In contrast, the fuel vapor concentration in the ullage of a gasoline tank can be as high as 40% on a warm day.
The fuel in the CWT of TWA Flight 800 was cold.
False. Fuel and fuel tank temperatures were measured in flight tests with a Boeing 747. The temperatures at the time of the explosion ranged up to 60 C in the fuel layer at the bottom of the tank. [Exhibit 23F]
The vents in the CWT would prevent the pressure from building up during an explosion inside the tank.
False. The effects of venting were studied in the 1/4-scale testing program and shown to be insignificant. The venting system area is too small to relieve the pressure from an explosion, even one that takes more than several seconds to pressurize the tank. [Exhibit 20E, 20O]
The NTSB investigators were unable to make Jet A explode in any tests.
False. Several hundred combustion experiments with Jet A were conducted to determine the peak overpressure, flame speed, and ignition energy. The conditions of these experiments bracketed the range of conditions observed in the flight tests. [Exhibits 20D, L, P, T]
The NTSB investigators had to use an ultra-sensitive explosive mixture of propane and hydrogen in 1/4-scale testing because they were unable to get Jet A to explode.
False. As discussed above, explosions were obtained in numerous experiments with Jet A in the laboratory and in the 1/4-scale tests. Several hundred experiments with Jet A were carried out in laboratory tests [Exhibits 20T, L]. The first series of experiments in the 1/4-scale facility used a hydrogen-propane mixture to simulate Jet A. The purpose of using this mixture was to simulate flame propagation in a tank that was at a different temperature and pressure than the explosion altitude. The way in which this was accomplished and the choice of the simulant mixture are documented in Exhibits 20E and O. Subsequent experiments in this facility used Jet A at the temperatures and pressures corresponding to the conditions in TWA 800 at the explosion altitude. A movie from a Jet A experiment was shown at the August 22, 2000 board meeting illustrating the explosive combustion of Jet A vapors in the CWT model. This experiment demonstrates that sufficient pressure was created to reach the failure pressure on SWB3 despite the presence of venting and the failure of the model SWB3 and FS panels. [Exhibits 20O, P]
The NTSB investigators had to use excessively large spark energy in their 1/4-scale testing.
False. A spark was never used to ignite the 1/4-scale tests and ignition energy studies were not a part of the 1/4-scale testing program. Extensive results are available on ignition energy in separate laboratory tests [Exhibits 20T, L]. The ignition system used in the 1/4-scale was actually a hot filament ignition system and no sparks were created. [Exhibit 20O]
The NTSB investigators had to use a torch in order to start combustion in Jet A laboratory experiments.
False. Jet A vapor-air mixtures were ignited by many different ignition sources in the Caltech experiments. Experiments were conducted with spark ignition sources [Exhibits 20T and L] with energies ranging between 1 mJ and 100 J and the 1/4-scale experiments simply used a hot filament [Exhibit 20O]. The preliminary experiments discussed in Exhibit 20D used a flame jet from a 1/4-inch diameter nozzle, comparable to some of the smaller openings in the spanwise beams and spars within the tank. Other experiments with pools of liquid creating vapor-air mixtures in the entire ullage with spark ignition in both the laboratory testing [Exhibit 20L and T] and hot filaments in the 1/4-scale experiments [Exhibit 20O, P] demonstrate that the explosion of Jet A vapor mixtures in an ullage over a thin layer of fuel is associated with a propagating flame (deflagration) in the vapor-air mixture rather than a pool fire on the liquid layer. The combustion time is sufficiently short so that the combustion is completed before appreciable venting occurs. [see 1/4-scale experiment number 67 - Exhibit 20P]
Jet A has to be violently shaken and lofted as a mist in order to explode.
False. Hundreds of experiments documented in Nestor, Ott, and the Caltech studies [Exhibit 20D, L, T, P] demonstrate that an explosion can be produced in a quiescent mixture of Jet A vapor in air.
The studies that Caltech carried out on Jet A explosions refute the NTSB crash investigation conclusions.
False. Caltech's studies support the NTSB findings of probable cause. [NTSB Final Report]
The combustion of Jet A inside the CWT was not a "real" explosion.
False. An explosion is a vague term used to describe an event associated with rapid energy release (see the glossary). The type of explosion considered in our studies is technically a deflagration or propagating flame occurring in the premixed volume of Jet A vapor and air within a multi-compartment vessel (the CWT). The deflagration pressurized the CWT until the failure pressure of Spanwise Beam 3 (SWB3) was reached. When SWB3 fails a blast wave is created. The combination of this blast and the rotation of SWB3 into the Front Spar (FS) cause the failure of the FS and ultimately, the fractures in the fuselage that destroyed the hull. [Exhibit 18A]
The peak explosion pressure was too small to rupture the CWT of a 747.
False. Boeing estimated the failure pressure of the components of the CWT. [Exhibit 20E and Report on Observable Early Event Damages in Exhibit 18] The peak explosion pressures were measured in tests, including 1/4-scale experiments with Jet A, venting, and simulation of structural failure of SWB3 and FS. [Exhibit 20E, O, P] The peak explosion pressures significantly exceed the failure pressure of SWB3 and the FS.
Lightning strikes frequently occur on aircraft without causing explosions, proving that Jet A must never be explosive.
False. Lightning strikes do occur on most commercial aircraft on the average of about one per year. Airplanes are designed to accommodate these through special design and engineering safety features (explosion suppression systems located at the wing tips on the vent lines, bonding of components) that prevent the stray electrical energy from reaching the interior of the fuel tanks. These features are not designed to prevent or suppress explosions that start within the fuel tanks. [See the Fischer and Plummer reference in Exhibit 20T
Why doesn't the FAA do something about fuel tank flammability if this is a real problem?
They are. The FAA has issued a Special Federal Aviation Regulation (SFAR No. 88) to minimize the potential for failures that could cause ignition sources in fuel tanks on new and existing airplanes. They have issued (as of 2002) over 40 Airworthiness Directives (AD) related to issues derived from TWA 800 and related fuel tank hazards. The FAA initiated a research program on fuel tank protection and are carrying out research and proposing regulations for fuel tank inerting, which is being carried out with industry participation. See also Advisory Circular AC 25.981-2 Fuel Tank Flammability Minimization. The FAA announced on Feb 17, 2004 that it is considering issuing a Notice of Proposed Rulemaking (NPR) requiring a fuel tank inerting system to be installed on existing aircraft with center wing tank flammability hazards. The FAA issued the proposed rule in November 2005, the final rule in July 2008, and the NTSB closed the issue in October 2008.
Why didn't the industry identify the problem earlier?
The industry was not required to consider this as part of the certification process. SFAR No 88 and Advisory Circular AC 25.981-1B FUEL TANK IGNITION SOURCE PREVENTION GUIDELINES now require that "The design approval holder would be expected to develop a failure modes and effects analysis (FMEA) for all components in the fuel tank system. Analysis of the FMEA would then be used to determine whether single failures, alone or in combination with foreseeable latent failures, could cause an ignition source to exist in a fuel tank. A subsequent quantitative fault tree analysis should then be developed to determine whether combinations of failures expected to occur in the life of the affected fleet could cause an ignition source to exist in a fuel system."
If this is such a problem, why aren't airplanes falling out of the sky all the time due to fuel tank explosions?
You wouldn't expect them to. The frequency of commercial aviation fuel tank explosions worldwide is estimated by the industry to be about one every 5 years without any preventative measures (see the FTHWG ARAC report for a complete discussion). Given that major airplane accidents are relatively rare events, the observed and predicted frequencies of major accidents due to fuel tank explosions are consistent. (PAL 1991, TWA 1996, Thai Air 2001) In the last two decades, there has been an average of about 1.5 major accidents per year in the USA (due to all causes) with relatively large fluctuations from year to year. Of these, one (TWA 800) has been a fuel tank explosion.

Accidents have been reduced to this very low rate in commercial aviation through careful attention to design, engineering safety features, and redundant systems. This applies to fuel tank explosions just as it applies to any other mode of failure. The fact that a failure mode or safety hazard is identified by an accident investigation does not mean that it has a high probability of occurring. On the other hand, the fact that a system has engineering safety features does not mean that accidents can never occur. Accidents in complex engineering systems like aircraft are frequently due to a combination of failures that were never considered by the original designers.

In the interest of safety, the NTSB does investigate every accident in commercial (and general) aviation and attempts to find a probable cause. The purpose of this is not to fix blame but to improve safety. Identification of a potential safety hazard requires that the investigative and regulatory agencies respond with proposals to reduce or eliminate the problem. The burden of evaluating the solutions and carrying out a cost-benefit study (required by federal statute) falls to the FAA - who relies almost exclusively on the industry to make the evaluation of technical feasibility and economic viability of any proposed solutions.

Cars contain fuel pumps and wiring inside the fuel tanks - why don't they blow up more often?
Gasoline tank vapor spaces are almost never flammable while Jet A tanks in airplanes will always pass through a flammable regime during normal flight operations.

The ullage of gasoline fuel tanks in automobiles is almost always too rich to be flammable except at very low temperatures. This is due to the much lower flash point (about -40 C) of gasoline in comparison to Jet A. The vapor space in a partially-filled gasoline tank does not become flammable until the temperature has dropped below about 10F and a serious hazard will exist below 0 F down to about -40 F for a typical gasoline (Reid Vapor Pressure (RVP) of 9.5 psi, flammability limits between 1.4 and 7 % by volume). [See W.F. Marshall and G. A. Schoonveld, SAE Transactions, Vol. 99, No. 4, 594-617, 1990]

For this reason, the probability that a fuel tank containing liquid gasoline has a flammable vapor space is extremely small in most climates except in the artic regions. The exception to this is when the tank is removed for servicing and the fuel is drained from the tank. Since gasoline has such a high vapor pressure, the tank can have a flammable vapor space even if there is no liquid fuel visible inside that tank. For example, the complete vaporization of about 1-2 tablespoons of gasoline will result in a flammable mixture inside a 15-gallon capacity automobile tank! This is the reason why welding on or near "empty" gas tanks is extremely hazardous and thorough purging of the tank with steam, carbon dioxide, nitrogen, or other inert gas is required before repair work is started.

The in-tank fuel pumps in modern fuel-injected vehicles are designed to run submerged in fuel, as are the fuel pumps in aircraft, and the fuel circulates through the pump, including the sparking brush-commutator system. Unless the gasoline tank is almost completely emptied of fuel, this will not pose a hazard since there will be only fuel and not a fuel-air mixture within the pump. So in order for an in-tank fuel pump to pose an ignition hazard, the tank must be "run dry" under very cold conditions so that flammable vapor surrounds the commutator.

Fire and explosions occasionally (about 3 accidents out of 1000) do happen when gasoline-powered vehicles crash. Post-crash fires are a serious safety issue for automobiles and are the subject of ongoing study, legislation, and litigation. Fires occur after the gas tank is ruptured and the accident results in an ignition source, often arcing electrical wiring or exposed hot lamp filaments. It is actually much harder to ignite gasoline by spilling it on a moderately hot surface that it is to ignite Jet A so that tailpipes and exhaust manifolds are not good ignition sources for gasoline. On the other hand, gasoline makes a large vapor cloud very quickly and if a high-temperature ignition source is introduced, a very rapid or "flash fire" will be the result. This can serve as an ignition source for the puddle or pool of fuel under the leaking tank, causing a pool fire that may destroy the vehicle.

If a fuel tank temperature is above the flash point, then an explosion will always occur
False A potential explosion hazard exists when a liquid fuel is above its flash point but an explosion occurs only if an appropriate ignition source is present in the flammable portion of the fuel vapor-air mixture within the tank. Some possible ignition sources include a hot surface, an electrical arc or spark, static electricity discharge, open flame, or burning metal fragments (e.g. from grinding or incendiary devices). Unless the surface temperatures of a fuel tank containing kerosene exceed the minimum AIT of 190 C, ignition will not occur due to the elevated temperature alone. The term "flash" in flash point temperature refers to the measurement technique, in which an open flame is introduced into the vapor space above a warm liquid fuel. The flash point is the minimum liquid temperature for which a "flash" is observed by the operator when the flame is inserted in the vapor space.

In general, in order for a explosion to take place, three elements are required: fuel, oxidizer, and an ignition source. This is usually refered to as the "fire triangle" by fire and explosion investigators. In addition, two conditions are needed before an explosion actually occurs: 1) The fuel and oxidizer must be molecularly mixed in the correction (flammable) proportions, 2) An ignition source of the appropriate size and duration must be present in the flammable portion of the mixture.

Copyright © 1993-2021 by California Institute of Technology, Joseph E. Shepherd