Safety Recalls - what else could possibly go wrong...?

An old boss of mine once told me “… your design will definitely kill two people per year!”

That was 20 years ago, when I was a fresh faced engineering graduate in my first job at a global car company. I was designing part of an engine management system, and as ever I had gone through every type of possible failure and worked out how well it was catered for.

But one very obscure scenario involved the car stalling on a hypothetical level crossing near a strong radio transmitter, a bit tenuous but it was a situation that could happen. I had gone through the figures and worked out that it was a million-to-one chance that the engine would not restart, resulting in something bad involving a train and sudden localised distortion to the car (ok, a crash).

I thought this was a remote chance, but my then boss pointed out that the systems would be installed in over 2 million cars per year in Europe, hence his terminal conclusion.

I redesigned it. No one had to die.

Even so, I am sure there could have been even more obscure scenarios I had never even thought of, and I could probably have spent years going through more and more complex tests in which case the car might never have been made.

We had to draw the line somewhere.

So how common are uncommon faults?

It’s with a great deal of sympathy that I read about Toyota’s sticky pedal problem, millions of cars that work fine, yet a rare glitch has necessitated a wholesale recall.

You just can’t take chances, even if most of cars are absolutely fine.

Toyota are no worse than Ford, Mercedes or any other leading manufacturer, all volume products suffer from occasional problems, largely due to the scale of production and of course because we want them cheap, and that’s not going to change any time soon.

When an industry has to make very complicated machines, that are used by the general public, and have to endure a vast array of harsh environments, things are going to get tricky. And when this situation is compounded by having to make the car as cheaply as possible, then the probability of a fault occurring (eventually) is inevitable.

Multiply this probability by the millions of cars made every year and the law of averages is definitely not on the side of car makers.

If you think about it, the mere fact that when something does go wrong it usually makes the headlines, tells us something about the fantastic job these companies do most of the time.

If the average Joe knew anything of the vast amount of work that goes into creating cheap, economical, useful and reliable cars they would be gobsmacked, and those that fancy their chances at suing for spurious accidents would perhaps hang their head in shame.

But hardly anyone knows about all that fantastic engineering work, it doesn’t make for sexy TV programs, nor is it compelling or vacuous enough to make it into the tabloid press.

Society tends to expect that every machine should work perfectly no matter what, and are then surprised on the very rare occasion that it doesn’t.

Cars are amazing

Here’s a challenge for you; think of a machine that has to work in heavy rain, baking sun, snow, ice, deserts, tarmac, cobble stones, at temperatures between -40 to +50 C, last over a decade whilst being shaken, accelerated, decelerated by novice users in a crowded and complex environment.

There are no other machines, just motor vehicles, which have to contend with all this.

But it doesn’t stop there, the engine is retuned every combustion cycle, hundreds of times each second. The suspension analyses the road and adapts to suit, the auto gearbox monitors the drivers ‘style’ and changes the way it works to please them. The brakes check wheel speed thousands of times a second and deduce when a tyre is about to skid and relieve brake pressure just before it happens.

Components have to operate faultlessly for millions of cycles. If an engine or drive-line fault develops then the systems must identify it, adjust the mode of operation to minimise risk and alert the driver, just like having an expert mechanic on board.

In addition the car has to be comfy, economical, perform well, have a really good sound system and be near silent in operation.

Not even the Space Shuttle has to contend with this level of sophistication.

And here is the kicker; as well as coping with all that, it also has to perform special functions in a crash.

We have multiple air bags whose operation is tuned to the ‘type’ of crash detected. We have automatic engine cut-outs, hazard indicators, seatbelt pre-tensioning and some cars can even make an automatic phone call for help.

Name me one other machine that has to reliably detect when it is about to be destroyed and then deploy safety mechanisms during the actual process of destruction. You’ll struggle with that one.

Now this feat of engineering would be amazing even with an unlimited budget, but the fact is that cars are made as cheaply as possible, which just takes the achievement from amazing to utterly astonishing.

Please take a few moments to look at your own car, and marvel. And if one part goes wrong, have some sympathy for the scale of challenge faced by those who engineer them.

Very soon, drivers will be the ‘least’ intelligent component in a car

Cars have become so reliable these days and the drivers’ role less and less involving, that some people are often unable to cope with a simple problem. I would have thought that if the accelerator pedal became stuck then you’d either put your toe under it and pull it up or drop the transmission into neutral, park up and switch off. Simple. But many people have lost the ability to cope with even the simplest of problems, and that is just scary. [Watch the video from Edmunds InsideLine entitled “How Hard Is It to Stop a Prius”]

I say scary because we depend more and more on technology – cars, computers, the internet, mobile phones, and the list goes on. And for the most part the technology serves us amazingly well, but like all devices there remains the possibility of that one-in-a-million or one-in-a-trillion chance that a fault will occur.

I remember back in the 70’s there were regular power cuts, no problem; the lights went out so we lit a few candles and life goes on.

Now don’t get me wrong, I am a great fan of technology. As an engineer I work on car technology that won’t see the glowing lights of a showroom for maybe seven years, as a writer I would be lost without the word processor and its fantastic ability to correct my abysmal spelling. Oh yes indeedy I just can’t get enough of the techy stuff.

What I am scared of is the way people are losing the ability to do things for themselves. To even bother trying to solve problems seems too great a challenge, our minds are being numbed and making a less intelligent contribution to the process.

The Driverless Car

Here is an interesting observation: most drivers don’t like driving.

Unlike enthusiasts, who experience a deep fulfilment from the task of driving, in the mass market most car owners don’t actually enjoy driving at all, it’s just become a necessity of modern life. That’s why so many drivers don’t pay attention and would rather chat on the phone, listen to the radio or check the football scores on their smartphones.

Cars are a very unusual phenomenon in that respect. Where else would you find such a large, complex and potentially dangerous piece of machinery being operated by someone so potentially disengaged?

And because of the ‘non-professional’ nature of the vast majority of car owners, technology is being developed to meet their needs. That is; making the car make most of the decisions.

We are already seeing Volvos with ‘collision avoidance’ brakes which perform an emergency stop before you drive into the car in front. Many cars have adaptive cruise control using radar sensors to move with the flow of traffic, some cars have lane assistance which nudges the steering to keep the car between the two white lines. And fully autonomous cars are in development, you just get in, tell it where to go and it drives you there.

To many this is automotive heaven, just like having a chauffeur, and takes the irritating burden of ‘having to do some driving’ out of a journey completely.

Plus there are safety advantages which make a very compelling argument; the fact is that nearly all accidents are caused by the driver making a mistake, so taking the driver out of the equation would in theory save lives.

This argument alone is powerful enough to kill the ‘drivers car’ stone dead, no arguments. It is simply infeasible to argue that autonomous cars should not be compulsory just because we want to have a little bit of fun.

But to enthusiasts this is automotive hell, no control, no involvement, no enjoyment, nothing.

And it also takes a lot of skill and judgement away too. What if I want to drive on the left of my lane to get a good view past the truck I am about to overtake? Will the lane-control system let me? What if I need to gently nudge my driveway gate open because it has blown shut? Will the collision avoidance system let me?

Complexity

What’s the greatest challenge facing car engineers and designers? Meeting carbon emissions targets is a damn good one, as is crash safety. But by far the biggest problem facing car engineers is complexity, and it’s a problem that for the most part is being hidden from the general public.

With all the highly sophisticated systems on board a car, such as engine controls, ABS, crash avoidance, adaptive gearbox systems and even sat-nav, knowing exactly how each part will react to the behaviour of another part has become almost impossible.

No usability testing, no matter how comprehensive, will be able to catch or isolate every possible problem.

Modern cars don’t just have a set of complex independent systems, they are also linked together. This has provided some amazing cross-function capabilities, such as traction control where wheel slip is detected by the ABS system and the engine controls reducing power to suit, and it has given us seamless automatic gear shifts where the gearbox talks to the engine to ensure the speed and power are matched perfectly as each new gear is selected.

More importantly it has provided much greater safety protection, for example if the brakes fail then the electronic hand brake system can lend a hand and the engine and gearbox can work together to increase engine braking.

It can even compensate for driver incompetence; some people panic in an emergency and press both pedals to the floor, modern cars can detect this, apply the brakes and return the engine to idle. This simple step has saved many lives.

Now, the concepts of integrated safety systems are simple enough to understand, the arguments for and against them are also fairly simple. Even politicians can understand.

But the devil is in the detail, and when you get down to the actual software it gets mind-bogglingly complicated.

I will give you a relatively simple example.

In order to reliably detect if the accelerator pedal sensor has failed, the pedal has at least two independent circuits, the signals are compared to see if they agree, that way if a wire is broken then the system will detect it and the engine can be safely returned to idle.

But it has to do more; what if there is a mechanical failure such as a broken return spring? Well, the signal is also analysed for movement so that if it stays in exactly the same position for too long then there is a fair chance it’s stuck. But how long is ‘too long’?

This is where it gets tricky. The signal is also compared to other signals, such as the brake pedal as mentioned above. But even if both brake and accelerator are applied at the same time, what if the fault is not in the accelerator pedal, nor in the driver panicking, but in the brake pedal sensor? This could lead to a tragic loss of power when the driver needs to accelerate out of danger, such as on a railway crossing.

So one layer of complexity involves where do you set the limits, how much analysis do you perform and how many other systems do you compare?

But there is more complexity, oh yes, much much more.

What if the various systems are not entirely in tune with each other? For instance when braking, as the speed drops the gearbox changes down and requests the engine speed rises to match, so the throttle is opened.

Usually the various signals are perfectly matched and this works seamlessly, but what if the signal from the gearbox results in a momentary surge of power from the engine?

So clearly the teams developing and tuning the brakes, gearbox and engine have to work together to ensure that under every different level of braking and speed combination, everything matches up. And that is a lot of work.

However, it gets more complicated. Many companies buy in certain systems, maybe the ABS from Bosch, the gearbox from ZF, possibly even the engine might come from another company, or another division in a different country. And even within those teams, parts of the computer control software might be outsourced to other divisions or companies, bringing a further layer of remoteness to the design team.

So where is this all leading?

Well, to greater complexity and less understanding of how each component system relates to other systems. The potential for unintended influences between systems will always be present.

I believe that it is now impossible to accurately assess how such a car will react in all conditions.

This is true not only for cars, but in many of the systems we rely on today, from automatic number plate recognition, military automatic targeting and smart weapons, the DNA database and even the way we use the internet.

The potential for technology to assist is immense, but it has to be understood that we have now lost control of every single detail. So how far do we let the machines dictate to us, and how much override can we allow to fallible humans?

The answer to this question will dictate the future of motoring, and before we set about blaming the auto makers consider how complex their task has become and what the weakest component in the system really is.