Estimating the Risk associated with Concorde fuel tank puncturing

As seen in other articles using the Concorde case as an example, the energy source that characterises the Concorde crash is that of the chemical bonding energy in the fuel.  The subsequent chained Occurrence of the aircraft crashing is not unique to this case, nor is the prior chained Occurrence of the tyre burst.

It is possible to conduct a post-hoc Risk estimation of the case as there is the data from the operational life of the Concorde to support this.

An estimation of the Risk starts with estimating the LLC and the LWC.

What constitutes a LLC needs to be defined and I’ll say that it is a case in which fuel is released from the tank but does not ignite and the problem is discovered resulting in the aircraft flying to burn off fuel so that it may land and then the associated delay and repair costs.  Fuel loss could be discovered en route as increased fuel usage is noted and the aircraft is landed at an en route alternate aerodrome.  Or fuel loss could be seen from the ground and the pilots notified so that the aircraft flies to a location at which it can dump or burn off fuel and it returns to its departure aerodrome.  LLC would be made up of the cost of fuel lost for no productive use, the cost of the delay for passengers and crew (hotel etc.) and the cost of repair. Of course, someone working for an airline would have a better understanding of the actual components of delay costs due to aircraft unserviceability.  My estimate of this is a total of some $300,000.

The LLC frequency can be estimated from information provided by news services at least, eg.‘Concorde fuel tank ruptured six times’, BBC News Online. 5 January 2001. Seen 25 April 2015.  Hence the historical Frequency of fuel tank leak from rupture is 6 rupture leaks /27 years = 0.22 [rupture leaks per year]

What constitutes a LWC is given us by the actual case which arose from a significant rate of fuel leak and ignition leading to the subsequent chained Occurrence of the crash.  113 people died, the hotel and the aircraft were destroyed and much engineering investigative effort not to mention the charges of lawyers was incurred, apart from brand damage.  My estimate is that the total of these would be something of the order of $450,000,000. Note this source – ‘Tragic Concorde Crash Could Cost Insurers $350 Million’, 26 July 2000. Retrieved on 25 April 2015

We have a single data point for this case as one of them occurred in the 27 years in service, or a Frequency of 1LWC/27 years = 0.037 [LWC per year].  This is a subjective estimate, of course, not an objective one.  As pointed out in Chapter 6 (p88) objectively determined risk is almost an anomaly so we should recognise the subjective nature of this estimate rather than being deterred by it.

The resulting Risk Diagram is:


The resulting Risk estimated in the manner described in Chapter 4 is of the order of $18M pa. (about $5000 per flight given an estimated 3704 flights per annum).

I gave this task as an assignment to a group of students, without specifying anything other than that a Risk Diagram for the fuel leak Event was to be constructed using whatever they felt necessary by way of assumptions as to components of costs and the frequency of the LLC and LWC.  Different students found different sources with information about tyres breaking and fuel tank rupture.

This is a list of results of those who successfully completed the task, i.e. whose analysis was logically correct.  The bold outlier resulted from significantly greater estimates of the LWC than those of the majority.

$27, 291,000

The mean value is $25.6M if the outlier is excluded and my own estimate included.  This gives some sense of the uncertainty associated with the estimates.  However, if this task was to be completed by a group who did more research into actual costs than could be done for the purposes of an assignment task primarily concerned with logic and understanding and given the validity of the date reported by the press, one can readily see a Risk being derived that all would accept.  The fact that the one data point for the LWC Frequency is an inherently subjective data point can be recognised by the group and its essential uncertainty about the conclusion limited to that figure.

It is interesting to note how different from a 45 degree line the derived Risk line is.

A 45 degree slope projected ‘backwards and upwards’ from the LWC predicts 52 fuel leaks per annum, which is clearly not the case.

A projected 45 degree line downwards from the LLC case predicts the LWC Frequency to be once in 4500 or so years, also hard to believe.  To understand the LWC, one has to imagine a scenario in which a larger tyre fragment than had been experienced punctures the fuel tank and tyre fragments also damage undercarriage bay wiring, that then ignites the fuel or the fuel is ignited by hot engine exhaust gases and the subsequent likely effect on pilot decisions and structural integrity in a return to an emergency landing at high weight.  A formal Event Analysis (the Outcome process, see Chapter 9) could be conducted to look at these possibilities and probabilities.  As it is not too hard to envisage this, especially if undercarriage bay wiring is not protected, it is hard to accept an interval between LWC cases of over 4000 years.

Even though the chance of what actually happened to the catastrophic flight was remote, its ‘remoteness’ could be estimated and the significance of the process truly understood.  Given that knowledge it would perhaps have been considered a prudent precaution to ensure protection of undercarriage bay wiring.  Also, given the relatively small number of Concorde flights, prudent to ensure runway inspections prior to Concorde departures.

Ah, the value of 20:20 hindsight!  Our task as true risk engineers and managers is to bring that hindsight into foresight.

Leave a Comment