For the last quarter century, I’ve been lurking on every forum not only for each car or bike I own, but every car I’ve designed parts for. Mostly it’s to check that I’m not responsible for some series of horrific failures (I’m not! So far…), but I’ve never found one post, not one, praising the elegant and efficient design of a fuel pipe bracket or whatever. Not even the AC pump replacement bracket for the V6 Exige, and that was really neat. What I have found is so many questions that could be answered by reading the manual, so many terrible pieces of advice, so many wrong conclusions, and the one thing that keeps cropping up again.
That thing is the idea that cars are designed to fail the second the warranty runs out. That phrase “designed to fail” really gets to me. I take it very personally when someone suggests that I have deliberately set out to ruin their car the second their eligibility for free repair runs out.
When we start designing a car we have a set of durability targets that we have to beat. These are never defined as “must fail after,” but instead are “must still exceed this spec at the end of testing.”
[Dave Larkman is a mechanical design engineer who had a 25-year career at Lotus Cars and Lotus Engineering (the consultancy business that worked for other OEMs), eventually becoming Lead Engineer of Powertrain Design. He has also been a semi-pro drifter, rides sports bikes, and used to feel ashamed about his taillight collection until he found Jason Torchinsky on the internet. Wait, why am I writing this in third person? It’s me, Dave, writing my own bio. – DL [Ed note within ed note: You know Dave from his excellent article “I Was So Bored At Work I Redesigned A Tiny Engine Part For Fun And Accidentally Saved 22,000 Pounds Of Aluminum.” -DT]]
Many Parts/Systems Are Designed For Conditions Far Tougher Than Your Commute
A good example is a 1,000 hour idle test on engines. Idling is really tough on engines because at idle, the oil pump is doing its worst work, so any slightly moody bearing surfaces are going to get destroyed. I’ve seen a poorly heat-treated camshaft lobe wear out in just six hours of idling, to the extent the valve wasn’t opening at all, but the pass criteria on that test is that at the end of the test, the bearing faces must still be within the size and surface finish tolerances of a brand new part. As new, at the end of the test.
That test seems weirdly arbitrary, doesn’t it? 1,000 hours is a nice big round number. Engine testing is the result of a thing called Design Failure Mode Effect Analysis (DFMEA), where we sit down before the start of the project, write down everything that might go wrong, work out how severe it is, and how likely. A slight chance of a minor annoyance gets a low score, any chance at all of a safety or emissions issue and it gets a high score. Then we have to work out what testing we’d have to do in order to ensure that the design we come up with won’t suffer from the likely or severe failure modes.
The actual testing regimes are usually whatever we’ve proven to have worked well in the past. A 1,000 hour idle test sounds like something someone thought up because it’s a suspiciously round number, then found that poor parts fail, but good parts will pass. Job done.
A thermal cycle test is where we run an engine on a dyno at full load until the exhaust manifold is glowing red hot, then we turn off the fuel and spark, but keep motoring the engine at max revs to blow cold air through it, cooling it all down far faster than you ever could by driving it. Then repeat again and again, red hot to room temperature and back again. It’s horrific on exhaust components.
It’s also a 450 hour test, which sounds like the sort of number that’s not just plucked out of the air. Someone at some point either found that 1000 hours of this was impossible to pass despite the parts being OK for production, or that 100 hours was inadequate. Or maybe it’s just how long it takes to do 10,000 cycles. Either way, it tends to fail parts that won’t get failed by any other test, which is why it’s so useful. When your exhaust design stops failing at thermal cycle, you can relax.
Note that the target here isn’t to delay the failure until after the warranty period, it’s so that you can go to court with your DFMEA, drawings, and testing plan and say: We thought of this, we designed a solution, we tested for this, we can prove we took reasonable steps to ensure that this failure wouldn’t happen.
We don’t have an upper target for durability. Cars are almost infinitely repairable; with careful use and the right maintenance, you can get to 500,000 miles on pretty much any car, even one that I’ve worked on. But that maintenance is going to cost, which isn’t going to be acceptable once the value of the car falls below the cost of a new set of spark plugs (which is what happened to one of my old E30 BMWs).
We can make engineering decisions to keep that maintenance cost low, but every time it’s going to increase cost, and probably mass. I worked on an engine on which you could swap out the cylinder liners in half an hour — great for keeping maintenance costs low, but an utter irrelevance for pretty much every car owner on the planet. All engineering is a compromise, and anyone who says otherwise is trying to sell you something that’s very, very expensive, and also still a compromise.
The liners were made of glass, in case you were wondering; you had to pop them out for cleaning every few minutes of running time.
That’s a lot of text in one go, let’s break it up with a picture of some weird engine parts, all of which made it through testing first time without a single issue. Remember folks: Every bracket is already a design failure. (Brackets are inelegant).
These tests cost tens of thousands in prototype parts (look at how much aluminium they had to machine away on that big curved bracket – it started off as a rectangular billet), then more tens of thousands in labor to build, run, strip, and inspect. As a design engineer, you don’t want to waste that much money by getting it wrong, but that’s not the worst part of failing the test – those parts will also be fitted to the rest of the test engines, and test cars — maybe hundreds of them. If your prototype part isn’t durable, then that entire phase of testing can’t be done until you’ve fixed it; the knock-on cost is millions, or tens of millions.
So you do the analysis, you benchmark successful designs, and you hope that everything you design will be good as new at the end of whatever horrific accelerated life testing they throw at it. Sometimes it goes wrong, and there’s a combination of tolerance and assembly and testing that overloads something, or creates a weird resonant frequency, or maybe the guy doing the heat treating put a “before” cam in the “after” pile, but for the most part, you design for success, and expect to get it.
If you have a wacky new idea, you test it on research engines for years before risking it on a production project.
Design Engineer pay isn’t great, so we’d be open to the idea of a massive bonus for creating dealer service work, if such a thing existed. But often we don’t even work for the same OEM as the dealer network. With the amount of consultancy work and badge engineering going on, the link between the people who design your car and the people who will service it is tenuous. In fact, the only dealer feedback we get is “make servicing easy,” which we totally ignore just to infuriate the fifth owner sixteen years later (actually we agonise over that too – I still feel bad about the requirement to use a
crow’s foot to get to one of the inlet manifold nuts on supercharged 2ZZ Exiges, and that was over twenty years ago).
There is no motivation at all to design-in a lifetime limiting feature; it’s bad for us personally, in an immediate being-shouted-at way, and if that failure can cause an engine to stop, or a loss of vehicle control, then we know the guys who risk their lives testing the cars. Sometimes we are those guys.
While I always considered it deeply cool to get to do some testing or mileage accumulation, always in the back of your mind is “what if something important fails right now?” The pre-production cars have to pass a huge amount of testing before they’re allowed out in public, but when you’ve spent weeks in DFMEA meetings, it sticks in your head. You really don’t want anything to break, ever.
Designing Parts To Fail At A Specific Time Isn’t Easy
But. What if we were evil? What if there was some kind of incentive to create failures and drive up reliability fear in customers to increase extended warranty sales? Presumably in a way that wouldn’t also utterly tank actual vehicle sales. Could we do it?
I’ve designed parts to fail. They were shear pins, the things that let collapsible steering columns not spear you in the chest, but in my case, they were steel pins half a mile under the sea in an oil well drill pipe. So, I do the analysis, design the notched pins, design the housings to load them in shear, then test a bunch of them and get an unworkably huge range of results. Getting those pins to be strong enough not to break until we needed them to was a nightmare, and that was when we could create an overload to suit us. Trying to get that huge bell curve of actual part durability to start after the right time or number of cycles is pretty much impossible, and for that bell curve to be narrow enough for one of the wheels to fall off a significant number of customers’ cars the day after the warranty runs out is just impossible.
Then you have the problem of the customers’ use case. Say it’s 100,000 miles or three years, are those miles on a track or sat idling in traffic? Are those years parked at the coast going rusty or baking in the sun? Or I guess both if you aren’t in the UK. Designing something to fail in large numbers immediately after that time/distance, but not a minute/mile before, is just impossible when you have no control at all over the environment or customer abuse.
Even if it were possible, which it definitely isn’t, the OEMs would have to test everything not just to exceed the required durability, but then carry on testing to destruction. That would cost hundreds of millions more, and seriously delay the introduction of new designs by years, which would cost hundreds of millions more on top of the hundreds of millions you already spent making the project late, and all for no benefit whatsoever.
It Does Happen, But It’s Not Intentional, And It’s All Just A Compromise
If something goes wrong on your car, just out of warranty or not, it wasn’t intentional. We hoped it would provide so many years of reliable, loyal service that you’d tell all your friends to buy one too. We bet our careers on it. We get paid little enough that we drive fourteen-year-old Toyota-badged Subarus and really could have done without having to shell out for a new wheel bearing again this year; we feel your pain.
Top graphic images: DepositPhotos.com
The post No, Car Companies Don’t Design Parts To Fail The Second The Warranty Is Over appeared first on The Autopian.
