The DH. Comet crashes: How unknown-unknowns brought down Britain’s entry into the Jet Age

The 27th of July 1949 was a momentous day. After having made two 500-yard hops in the morning, the first de Havilland Comet took to the air for a 31 minute test flight. (You can see footage here). Britain had brought the world truly into the jet age.

No longer would airline passengers have to suffer the indignity of plodding along at a mere 500 km/h, flying at the same level as the turbulence. Instead, the jet set could look forward to zooming around at 740 km/h, and cruise high enough to go over the top of all but the largest thunderstorms. The USA might have been the first to break the sound barrier (at least in a way that allowed the pilot to talk about the experience afterwards); the Comet showed that Britain was still capable of world-class aeronautical engineering.

Development took another 3 years, but on 2 May 1952 the Comet carried its first paying passengers. All seemed to be going well for the Comet, but soon disaster struck.

On 2 May 1953 Comet G-ALYV crashed six minutes after take-off from Calcutta airport after flying into a thunderstorm. Although subsequent events might call the conclusion into question, the official cause of the crash is still the overstressing of the airframe by the thunderstorm.

Not so easily explained was the crash, on 10th of January 1954, of BOAC Flight 781 (Comet G-ALYP), en-route from Rome to London. 20 minutes after take-off the plane broke into pieces and fell into the Mediterranean near the island of Elba. BOAC voluntarily grounded their Comets and performed a number of modifications suggested by de Havilland’s engineers. By the end of March the Comets were returned to service again.

But on the 8th of April there was another in-flight break-up of a Comet, this time G-ALYY, operated by BOAC as South African Airlines flight 201. Again the Comet crashed shortly after take-off from Rome, the pieces splashing into the sea near Naples. This time the grounding of the Comet fleet was not just voluntary; the type’s Certificate of Airworthiness* was revoked, and the Royal Aircraft Establishment (RAE)   was tasked to investigate.

The RAE set up a large test programme, which included performing fatigue tests on Comet G-ALYU, which was donated by BOAC (who must have been extremely worried by the prospect of having to retire their most prestigious aircraft). The results of the investigation are well-known, even to people who are not all that interested in aviation: the Comets crashed due to fatigue, caused by having square windows. Generally it is implied that de Havilland’s engineers didn’t know about fatigue; the Comet’s story is used as a cautionary tale about the danger of unknown unknowns.

However, as so often, the story is more complicated than that. De Havilland’s engineers did in fact known about fatigue. In fact the phenomenon has been studied ever since the Versailles rail accident in 1842, which killed between  52 and 200 people. It was precisely because de Havilland’s engineers were worried about fatigue, that they performed fatigue tests on a test fuselage. This fuselage failed after 16,000 (simulated) flights, whereas G-ALYP and G-ALYY had only performed 1290 and 900 flights respectively, before they crashed. The air-frame tested by the RAE, G-ALYU, failed after a total of 3,057 cycles (in-service and simulated added together).

So what caused such a big difference between de Havilland’s development tests, and the real aircraft? It is our old friend plastic deformation.

At the tip of a fatigue crack, there is a high stress concentration. So, when a load is applied (for example the cabin is pressurised), this causes the yield stress of the material to be exceeded in a small area surrounding the crack tip. As a result, a small zone around the crack tip is plasticly deformed. Remember that plastic deformation is permanent, even if the load is removed. However, the surrounding material is still elastic. That means that when the load is removed again (for example the cabin is depressurised after landing), the elastic material ‘wants’ to return to its original shape. However some of this material will find a lump of plasticly deformed material blocking this return. This results in a compressive stress (think of it as a pressure) on the plastic zone, squeezing it together. As the tip of the crack will be in the centre of the plastic zone, the end-result is that the sides of the crack are forced together. This makes it harder to grow the crack, so any crack growth will be slowed down. The larger the load you apply, the larger the plastic zone, and the stronger this effect will be.

Plastic Def Crack Tip
Example of the effect of plastic deformation at the crack tip. When the elastic material ‘wants’ to return to its original shape when the load is removed, it is blocked by the plasticly deformed material. This causes the surrounding material to squeeze the plasticly deformed material, which will force the crack faces together.

Normally a large load will cause a large amount of crack growth. However, if you apply just one large load, and then lots of small ones, the plastic zone caused by the initial large load will slow down the crack growth during the subsequent cycles, and it will take the crack longer to grow.

This is what happened in the case of the de Havilland test. The test fuselage used for the fatigue test was first used to check whether the design was resistant to over-pressurisation of the fuselage. It was pumped up to twice the pressure difference expected to be encountered in service, and another 20 times up to pressures varying between the maximum service pressure difference and twice the maximum pressure difference. In other words, de Havilland’s engineers unwittingly applied a large load on the structure right before the fatigue test, causing large plastic deformations. Consequently, during the fatigue test the cracks that were present in the fuselage (caused by drilling the rivet holes needed for assembly) grew much more slowly than those in the aircraft used in service. The safe life calculated from the fatigue test was therefore much too long.

Although the window design was not ideal**, it is not knowing that the fatigue test was not reliable, that made the engineers sign off the Comet’s design as safe.

With the results of the RAE investigation de Havilland’s engineers set to work designing an improved version of the Comet that would not suffer from the same fatigue problems. This eventually became the Comet 4, which entered service in 1958. Although Comets continued to serve until 1997, the crashes of 1954 and subsequent groundings had allowed the US to catch up. The Comet had to face stiff competition from the Boeing 707 and Douglas DC-8, and never achieved the success that was hoped for it.

A Comet 4B operated by British European Airlines. Photo: Ralf Manteufel, licensed via Wikimedia Commons under the GNU Free Documentation License v 1.2

Whenever you are the first to do something, the spectre of unknown unknowns is always lurking. Although the Comet may not have been a commercial success, it did inaugurate the jetliner era, and it is the investigations into its crashes that gave us the knowledge that makes modern air travel so safe.

*A government document, basically saying that a certain type of aircraft is safe and granting permission to fly it.

** For the aircraft that crashed into the Mediterranean the fatal cracks didn’t start at a passenger window, but at a rivet near an ADF window; a cut-out in the fuselage covered with a fibre-glass panel. This window isn’t transparent to visible light, but it is to radio-waves, which are used by the Automatic Direction Finder, a navigational instrument.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s