OK, so the system is broken and has to be repaired to restore production. Since the spares needed are imported, it’s going to be expensive and, if the parts needed are not in the stores as spares, it will take time. In addition, we have the cost of the actual repair itself over and above the cost of the spares, and the cost of production loss, which can easily exceed the sum total of the rest.
So why add an additional cost of analysing the failure? Failure analysis is not a cheap undertaking. If it is suggested that such an exercise can be carried out cheaply, the resulting report probably won’t be worth the fee, no matter how small.
An on-site examination is always a good place to start when it’s time to gather evidence, although the evidence may be ‘well distributed’ over the surrounding scenery in some of the more spectacular failures.
Failure analysis frequently requires the use of complex equipment and techniques, such as scanning electron microscopy. Electron microscopes are neither cheap to install or operate. The costs of hiring one can be over R1 000 per hour, while using an up-to-date state-of-the-art instrument may cost a few million Euros.
Similarly, determination of the material of construction of the failed part – involving chemical analysis, mechanical testing and examination of the microstructure – is not cheap either. Like the SEM, these techniques use equipment that is expensive to install and operate, and the procedures can be very time-consuming for highly skilled and experienced staff. Such people are hard to find and costly to keep.
Examination of the fracture surface of the broken part, a technique known as ‘fractography’, requires specialist skills. I have known supposed failure analysts who simply compare the appearance of the fracture with published images in the appropriate reference text-book, a method that will frequently lead directly up the garden path. The analyst needs to have a thorough understanding of fracture mechanisms in order to determine which mechanism or mechanisms were responsible and what factors were involved in the initiation of failure.
Sooner or later, those practicing the ‘comparison’ method are sure to meet someone who works from first principles and understands fracture mechanisms, usually on the other side of the courtroom, and be made to look incompetent and foolish. Not to mention probably costing their client a great deal of money.
Detailed knowledge of the normal operation of the system and its history are necessary, which can then be compared with the fracture mechanism, leading towards what abnormal operation may have occurred.
Let us consider a simple case, a leaking compression fitting on a small pipeline carrying a mixture of ethanol vapour and ethylene gas. The leak drew attention to itself by catching fire, fortunately without significant collateral damage owing to an attentive operator with a fire extinguisher handy.
The investigation was relatively simple in this case. Dismantling the joint showed that the cone ring had been installed upside down, and a futile attempt made to seal the joint by over-tightening, destroying the sealing ability of the system. A leak, even in a moderate pressure line carrying any gas will generate a charge of static electricity, and when the gas is flammable, the leak becomes self-igniting.
More complex was a series of failures in a beverage can seamer, where the machine operates at a speed that makes direct observation of the system in operation very difficult. The other problem was that, on failure, the speed of the machine effectively destroyed most of the components in the seamer head, leaving little evidence for the investigator to work with. It required the use of high-speed cinematography to find out what was actually happening when the machine was in operation.
We didn’t manage to capture an actual failure, but the knowledge obtained led to a better understanding of the machine operation and resolution of the problem, which dropped subsequent failure rates significantly. This is the real value of failure analysis.
Failures in motor vehicles add another set of variables, that of the driver and the witnesses. I have lost count of the number of times I have read the driver’s statement that the vehicle “was only travelling at 55-60 kph”, but the vehicle, having hit a kerb, had had one wheel torn off. That just doesn’t happen in a modern, high performance vehicle, nor even in the ordinary family hatch-back. It’s not always the case, drivers don’t necessarily lie. In one instance a bus driver, charged with twelve cases of culpable homicide when the bus he was driving rolled over, was exonerated when severe wear and corrosion was found in the steering linkage and demonstrated to the magistrate. The company responsible for maintenance, which was neither the owner nor the operator, had some awkward questions to answer, having signed the vehicle off as roadworthy only six weeks prior to the incident.
In a similar case, following an incident involving several fatalities, examination of a number of supposedly ‘identical’ vehicles revealed that the layout of the instruments on the dash-board varied, leading to confusion for the driver. When he detected what he thought was brake failure and attempted to change to a lower gear, the vehicle’s computer systems, guarding against probable over-speeding of the engine, refused and left him in neutral. Disaster became inevitable. Sadly, the brakes were fully operational and catastrophe could have been averted.
The authorities in this instance refused to release the data from the vehicle’s data recording system. We had to reconstruct the data from the on-board computers controlling the engine, transmission and braking systems, all of which had non-volatile memory units. The driver’s sentence of five years was suspended on appeal for, as the judge said. “The man will live in his own prison for the rest of his life”. He never drove a bus again, though his employers were prepared to let him. He was given a desk job.
In an aircraft, analysing the cause of failure becomes a more pronounced problem. It’s one thing to sit in the safety of the laboratory and analyse failed components for days, weeks or even months. One problem took me four years before I found evidence from a similar incident, fortunately this time without fatalities, though with a very large bill to be able to prove the true cause of failure. Too late by then.
It is quite another to be the pilot faced with a failure, who has only minutes or seconds to diagnose the problem, find the correct solution and then apply it, all in a situation that is stressful and life-threatening. And yes, I’ve been there as the pilot in command of an aircraft. All right, it was a two-seat Cessna, not a Boeing with a hundred or more paying passengers, but I know what it feels like.
In one case (not one of mine) the flight crew detected a severe vibration in one engine. Because of a confusing instrumentation system, they shut down the wrong engine. The aircraft crashed with loss of life. It did leave a newly delivered aircraft having a similar problem being grounded after post-delivery inspection by the local airworthiness authority. “Ok,” said the manufacturer, “we’ll take it back to the factory and fix it.” The inspector quite rightly refused, saying: “You’re not flying that thing in my airspace, you’ll fix it right here!”
So yes, failure analysis is costly. But it’s worth it. It stops problems coming back and it’s not as expensive as having to fix the same problem a second time. General George S Patton summed it up rather nicely. “I’m not paying for the same piece of real estate twice with my soldiers’ lives,” he said.
There is no point in sitting back and waiting for the insurance company to carry out an investigation. Usually they won’t and they may even decline the claim for lack of a properly carried out investigation. Failure analysis is expensive, but the results are frequently well worth the cost, and certainly much cheaper than having to fix the same problem twice, an event which will, at the very least, get one crossed off the insurer’s Christmas card list.