> The relatively good news for Boeing is that because the test failed so explosively at just 1% shy of meeting federal requirements, it will almost certainly not have to do a retest. Regulators will likely allow it to prove by analysis that it’s enough to reinforce the fuselage in the localized area where it failed.
I don't understand this. So they did a test, and it failed at 99%. They reinforce that one area where it failed, and show by analysis that the reinforced area would have made it to at least 100%.
OK, but what about the rest of the fuselage? The rest was only tested to 99%. Couldn't there be another place that fails at, say, 99.5%, which wasn't detected in the first test because the place that fails at 99% stopped the test?
Edit: I would guess that during this kind of test they have the plane heavily instrumented with things like strain gauges so they can understand how it reacts and see if it is behaving according to spec. So maybe it is the case that the data from the rest of the fuselage showed that it would have made it to 100% if the test had continued, and only the area that failed was in trouble. That could explain why they aren't require to do another test.
That still leaves me unsatisfied, though. Presumably analysis of the design before the test said it would make it to 100%. If the failure was not due to a manufacturing defect in that particular fuselage, then it suggests that they analysis was not adequate. But then what assurance is there that their analysis that shows that the reinforcement would hold is good?
The truth is that even the reinforcement they are going to do is more for PR than for actual safety. This test was intended to push the plane to its limit. Not close to, but actually up to the limit, so it was actually expected to fail somewhere around this point. What they call 100% for this test is actually 150% of what the plane is expected to experience in a worst-case real-world situation. The difference between failing at 99% and failing at 101% (which is what happened the last time they ran this test) is negligible from a practical point of view. You'll almost certainly get more variance than that in practice just from manufacturing non-uniformities, random environmental factors, age, etc.
Boeing has committed a lot of sins recently, but this isn't one of them.
Doesn't the incidental damage like blowing out a cargo door suggest that, while they expected to push the plane to close to it's limit, that prior to the test they assigned very low probability to this kind of failure? Even if this barely-a-failure isn't on its own evidence of a practical safety risk, anytime you're surprised by new data, it does suggest something about your model needs to be updated.
Sure. They'll update their models saying "yeah, this can happen sometimes at 148.5% of max designed wing load. So... don't fly the plane into any >F5 tornadoes."
These max overload wing tests are scary dangerous, but they don't represent real flight situations. 99% of their target safety factor is within the margin of error for manufacturing variance.
While they'll most certainly investigate the causes of the damage down the fuselage from the wing rupture, it's not likely much of it will be unexpected - if a plane experienced that load, the plane is considered a Loss of Hull even if it lands successfully somehow.
I'm immensely paranoid about flying on planes, I have a terrible fear of heights... and nothing I've read in that article bothers me in the slightest.
It is possible to design systems that fail in predictable ways (tear-strips on consumer packaging are an example of this) but airplanes aren't designed that way. It just doesn't matter how the plane fails when it reaches its structural limits, so no design effort is put into this. So no, it is not at all surprising that the energy dissipated in unexpected ways. That's one of the reasons they ran the last test past 100% so they could find out how it would fail because no one knew.
The whole thing is pretty ingenious, but the way the internal pressure helps vent the can while the tab is acting as a second class lever, then as soon as it’s vented becomes a first class lever to push the scored section of the top into the can — that’s great engineering.
Isn't there an ancient joke about an aircraft manufacturer having problems with the wings breaking off, and some engineer suggesting performating the wing roots, because have you ever seen toilet paper actually tear at the performation?
Not always. The pins that hold the engines on to the wings are "fuses" in that they are designed to break away if the engine is vibrating violently, before it breaks the wing.
That's not true at all. A good example of this would be the requirement that under a blade out condition on an engine it remains contained. They even test this
> What they call 100% for this test is actually 150% of what the plane is expected to experience in a worst-case real-world situation. [..] The difference between failing at 99% and failing at 101%
So, what is it 100% of?
Not that I even care enough about Boeing clickbait to skimread the article, but this discussion makes it seem like "percent" is a really poor fit for whatever this is actually measuring and trying to communicate.
So, first, airliners aren't really expected to experience more than 1.3G's of loading in normal or even unusual operation.
They are expected to be able to survive 2.5G's (limit load) without structural damage. The 777X satisfied this.
They are expected to be able to survive a few seconds of 1.5x this-- 3.75G-- without outright failure but possibly with (irreparable) structural damage. Instead, it catastrophically failed at a loading corresponding to 3.70G's.
They are going to reinforce and fix the specific failure that happened at 3.7G's, and make an argument based on strain gauges and instrumentation that another, distant failure was not likely before 3.75G's.
So, first, airliners aren't really expected to experience more than 1.3G's of loading in normal or even unusual operation.
This is 100% true. But the unexpected happens. China Airlines 006 saw 5G due to pilot error. Everyone survived and the plane (Boeing 747-SP) saw another 25 years of service.
Additionally planes age. What's your failure point after 10-15 years of service?
Another part that's worrisome is the original angle. Originally the failure was reported as a cargo door blowing out. The pictures released show otherwise. So the question remains: who got reporters to go with a cargo door failure and why?
They are going to reinforce and fix the specific failure that happened at 3.7G's, and make an argument based on strain gauges and instrumentation that another, distant failure was not likely before 3.75G's.
That depends on how much you trust Boeing's modeling and assembly skills. IMO the pickle forks on the NG cast at least a little bit of shade on Boeing's reputation.
With regards to the cargo door story - a door did blow out, but it was a passenger door, not a cargo door. Passenger doors are plug-fit (they're bigger than the hole, and pressure holds them in place), so it's not really possible for a passenger door to blow out. Cargo doors are not plug fit, so a latching failure could cause a door to blow out. Everyone in the industry remembers the DC10 cargo door failures.
I suspect the early rumours said a door had blown out, and as that's not really a possible failure mode for a passenger door, it was assumed it must have been a cargo door. Now, in this case a passenger door did blow out, but it wasn't a door failure. The fuselage some distance below a door failed, and split right up past the door. The partially unsupported door then blew out. But the door was peripheral to the story. Boeing could definitely had handled this much better, but it's not really surprising that in the absence of full facts the rumour mill got this one wrong.
Cargo doors are not plug fit, so a latching failure could cause a door to blow out.
Right. A cargo door failure could be simply user error.
Everyone in the industry remembers the DC10 cargo door failures.
And the 747 ones too hopefully (user error and poor design).
The fuselage some distance below a door failed, and split right up past the door. The partially unsupported door then blew out.
Right, the door failure was secondary.
Boeing could definitely had handled this much better, but it's not really surprising that in the absence of full facts the rumour mill got this one wrong.
Someone at Boeing obviously leaked part of the story (the door failed) and Gates ran with what information he was provided. The question remains: why only leak part of that story? Either failure will look bad to laypersons and both would be dismissed fairly quickly by people more familiar with airliners.
Is the working theory that Boeing PR is simply that incompetent or short sighted?
> Additionally planes age. What's your failure point after 10-15 years of service?
In practice there is/are none, because would-be points of failure are continuously replaced at a schedule well ahead of the number of cycles at which they are known or expected to fail. This is true for aircraft of all sizes, not just passenger jets.
In practice there is/are none, because would-be points of failure are continuously replaced at a schedule well ahead of the number of cycles at which they are known or expected to fail. This is true for aircraft of all sizes, not just passenger jets.
Sure, but what does this mean for the 777X? With the NG, even if you reduce the inspection interval to a few thousand cycles you're only looking spending an hour on an inspection.
How often will you have to inspect the skin (and/or whatever else failed or saw damage) on the 777X? How involved will the inspection be?
The thing I can't quite get over is that something led Dominic Gates to run with "the door blew out" as the story. Why? If a torn fuselage skin on a destructive test is so minor, why go with misdirection? Think of it in terms of MCAS. Conceptually relying on multiple AoA inputs is simple, MAX specific training is (relatively) simple. Yet the MAX is still grounded and we're seeing leaks that indicate a suitable fix is dramatically more complex than Boeing has been suggesting.
Note that this test also does not factor in fatigue, maintenance wear and tear, etc. Planes fly for several decades and some will fly with quite a bit of imperfections (i.e. cracks) for a while before it's maintenance day
100% of ultimate load, which is 150% of limit load. The limit load is what goes into the flight manuals and is what the aircraft is certified to withstand. The ultimate load is a safety factor on top of limit load and is what the engineers design for.
The writer has some of the numbers mixed up. Transport category positive load limits are 2.5 to 3.8 (depending on max takeoff weight), not 1.3. I don’t know where the 1.3 number came from, but it might be from the loads required to be demonstrated during flight tests (1.3g pull-up at 1.3x the stalling speed in the landing configuration IIRC).
100% of its expected actual failure point, based on its design.
The plane is rated to 2/3rds of that (I think this is pretty standard in aviation) - the other "100%" - meaning it should never go above that in flight, and if it does it would require a very costly teardown/inspection.
> This test was intended to push the plane to its limit. Not close to, but actually up to the limit, so it was actually expected to fail somewhere around this point.
I would assume that even in a failure scenario the fuselage is expected to fail gracefully instead of catastrophically.
By gracefully I mean having a fuse-like failure mode, where the whe system remains operational while a controlled failure is observed in a specific component. For example, a safety valve door blowout or at best a window/door blowout.
This test shows none of that. This test shows that Boeing's planes fail catastrophically.
1% is well within the normal variation in strength that parts exhibit. Even with modern materials and precision machining, parts will always vary in performance to a degree that is known. That said, predicting critical points is very deterministic process. Strengthening the critical point that failed in this test by a reasonable margin should, for all reasonable likelihoods, eliminate this point of failure. Predicted performance has error bars, and actual performance has scatter. If we bump up the strength, we can make sure the tail end of that scatter has less overlap with the failure strength scatter (theoretically we can never reduce the overlap to zero). All of experimental mechanical engineering is a comparison of statistical odds. It's quite possible that if they tested another identical fuselage, it would have passed the test and this would have never made the news, rightly.
The point that the FAA made is that it doesn't matter. A plane should not be experiencing these conditions ever outside of a test environment. The concern over 99.5% is splitting hairs, since at this point the plane is probably in the middle of a tornado or something
Additionally, this test was done while the plane was over-pressurized, which is not required, which adds a lot of stress to the airframe, and added to the "blow up" factor.
or that the reinforcement doesn't cause it to fail earlier because due to the increase rigidity in that area another nearby areas get more stressed than before
I would expect the analysis is done via FEA. Those models are not perfect but.. Suppose the FEA indicated the highest stress at the same point it failed. Maybe it predicted failure at 102 percent. In that case, doing a reinforcement that allows the whole plane FEA to reach 104 percent should really be sufficient.
I still believe in retesting, but I can see how confidence could be high without it. In this case I'd actually prefer letting it pass vs making a change without retesting since the loads are way outside of normal and it came so close. Best of course is to make the change and test again.
If I was in Boeing's shoes I would be doing everything to ensure the real test is repeated and the test is passed.
They already face a MAJOR plane credibility issue with the MAX and to have another plane enter service without it actually having achieved a physically passed would do nothing for the credibility of that new plane.
While a prove by analysis pass might be fine for the regulators, in terms of public credibility a real pass would be a far better outcome for Boeing.
Absolutely. I think Boeing has proven with these statements that they no longer care about safety or engineering standards. They should have continued to run tests to prove that they meet 150%, make a video of it and continue to run, in order to regain the trust of the public. But they seem unwilling to get back the trust.
Also, I think this also shows that the interaction that Boeing have with the FAA and vice-versa, has been heavily corrupted by something (money?) and FAA is no longer trying to keep Boeing in check. They are trying to make sure that Boeing passes whatever certification even if they might miss some details.
Airbus A380 failed A similar test at 3% less load and Airbus responded in a similar fashion. Boeing has problems, but this is practically a pass.
Airbus was testing a new (higher MTOW) variant of the A380. It's likely Airbus was expecting the premature failure as the fix was quickly deployed. I'm pretty sure you can't say the same about the 777X.
Note that wing spar cracks were later found in pretty much every early build A380 in service. Airbus then spent around 500,000,000 EUR to remedy the problem.
I'm sure Boeing can make the 777X safe, but it's likely a matter of motivation.
that's not how safety margin work, is not a budget leftover set aside for known unknowns, it's there for unknown unknowns.
e: downvoters might want to read Feynman again
"If a bridge is designed to withstand a certain load … it may be designed for the materials used to actually stand up under three times the load … But if the expected load comes on to the new bridge and a crack appears in a beam, this is a failure of the design. The O-rings of the solid rocket boosters were not designed to erode. Erosion was a clue that something was wrong. Erosion was not something from which safety could be inferred"
if you fail a test before your margin you haven't established that your safe up to 90% margin, you've established that the hypothetical on which your design was based are wrong and no further inference might be warranted, regardless of whether you fail at 99% or 60% of where you expected to fail
You aren't paying attention to the phrase "expected load".
If, in the bridge scenario, it cracks under 2.95x the normal load, vs. 3.0x or 3.05x, that is not what Feynman was describing. He was explicitly talking about a partial failure under about 1.0x. Which based on other comments is not what happened here.
except 3.75g here is the expected load°, the certification requirement. the design is engineered to reach that 3.75g°, the safety margin is what lies beyond.
°with the disclaimer that the plane might be not airworthy after landing
That doesn't apply in this case. Modeling will never be a perfect representation of reality. IE you're never going to have tested failure at 100.00% of the expected failure load. If it had failed at 90% I would agree that they should take another look at the design, but at 99% it's within expectations. And you still maintain essentially all of the designed safety margin for the unknown unknowns.
Back in March, the FAA was still insisting the 737 MAX was airworthy, even after it had been grounded by other countries. It seems like at the very least, the FAA is worried about the business interests of manufacturers.
This sounds fine. It's a catastrophic failure test. They push until it breaks, then it broke, and they were very close to their designed target (within 1%). If they had gone 1% past their designed target, they would have been upset, as they could have made the plane lighter. They were 1% short of design limit, they are upset because they'll have to add reinforcement.
Have you seen the wing failure test? I worked for the company that did these tests (they did them before I worked there), and they still talk about how upset the engineers were that it went over spec in strength.
The test on the Boeing 787 broke the load test machine - the composite wing was so strong, they literally couldn't break it.
And let's just put this on record, so people can really shut up about this: The Airbus A380 failed at 145% maximum wing load. That's a full 3.3% below the safety margin. They did not retest, but did subsequently reenforce (just as Boeing is here). The plane was put into service by the FAA.
This post didn't "drag Airbus through the dirt," it proved that there's no issue with the testing process at hand. Airbus did nothing wrong, nor did Boeing in this instance. Both "failed" to hit the target safety margin, took remediation steps, and did not retest. This is the system working as intended.
The fact that you see this as somehow anti-Airbus is evidence of a personal bias.
>And the interior of the plane was pressurized beyond normal levels to about 10 pounds per square inch — not typically a requirement for this test, but something Boeing chose to do.
Hmm. Does it lower structural integrity, or improves it?
100% on the test is 150% beyond what would be expected from the absolute worst case real life scenarios. It would require extraordinary circumstances beyond what has been seen or modeled in reality to push an actual flight to these limits.
It's pretty typical. If the cabin is pressurised to 6,000ft (standard in a 777) and the aircraft is at it's rated celing of 43,100ft, that's about a 9psi difference.
Very impressive. They strengthened the fuselage to meet requirements almost exactly, and fell a tiny bit short. That's an incredible amount of optimization, that means there's hardly any excess material (weight). This keeps fuel costs low and airfares competitive.
This was in the related articles at the bottom. I wonder how the decision is related the test? I wonder how much say legal had. That is, if they end up in court is human-made easier to defend than robot-made?
Also a positive note, the carbon fiber wings bent 3x expected max
>bent the jet’s giant carbon composite wings upward more than 28 feet from their resting position. That’s far beyond the expected maximum deflection in normal flight of about 9 feet
>All this simulated the loads in a flight maneuver where a pilot would experience a force of 3.75 G, compared to the maximum of 1.3 G in normal flight
>All this simulated the loads in a flight maneuver where a pilot would experience a force of 3.75 G, compared to the maximum of 1.3 G in normal flight
The reason fuse has an "expiration date" set by the amount of pressure cycles is because aluminium breaks more due to microcracks and other metallurgical fatigues, than due to ductile failure.
A composite lined fuse could've extended the life of an average airplane long beyond its moral obsolescence, and for this reason airplane makers are afraid adopting it.
What is "moral obsolescence"? If you mean "normal obsolescence", planes are used for a long time, so what would draw the line between normal lifespan and excessive?
It seems to me that if nothing ever fails short of the 100% test its probably overengineered, since the 100% already includes a significant safety margin.
Alternatively if nothing ever fails the test, then they could also just stop doing these tests, since the analysis would already be shown to be sufficient.
What gets me is what is the point of a limit to test to if you are not going to enforce it.
If 1% less is ok (Boeing) test to 1% less,
if 3% less is ok (Airbus) make the limit 3% less.
Not enforcing a limit smells bad and does not fill me with confidence.
Strengthening the failure area will change the failure modes to other areas and may well make it fail at an even lower limit as other areas will be bending more to make up for the reduced deflection of the area that is strengthened.
Not re-testing it after a design change like this worries me.
Wind is never going to make an airplane experience this level of loading. Think about it: an airliner going 900 kilometers per hour is flying into an apparent headwind of 900 kilometers per hour, much faster than any windspeed it is likely to encounter. The fastest winds that an airplane has to deal with are the tradewinds.
The biggest risk to aircraft from high winds is during landing and when on the ground. On the ground these forces are not going to cause a crash (for obvious reasons) and during landing the bigger risk would be a wing strike or ending up next to the runway. In neither case would the strength of the plane as observed in these tests be a huge factor.
What would cause such loading is a long dive or fall and subsequent recovery:
Thanks for clarifying. I felt afraid when I first read this story, because of the recent unpleasantness with Boeing. Now I see that it's unlikely for this level of stress to occur.
"we're not going to re-test. We failed the test but it was close enough"
Yeah that makes me feel safe Boeing, real safe. Specially given your history now.
Planes used to pass these tests with flying colors and they've been pretty safe. Have we forgotten already that planes used to break in half in the sky when there's heavy turbulence? 3.5G sounds like much, but it is also not that much. There's a reason why the test needs to be passed at a certain level, not just below.
Heck otherwise, how about I pay 99% of my taxes, mortgage, pass 99% of my exams, etc. What is this?
Note: before you read and trust the articles too much, airline planes have been WELL above 3.5G already in the real world. The fact that they're not supposed to go there does not mean they never do.
I don't understand this. So they did a test, and it failed at 99%. They reinforce that one area where it failed, and show by analysis that the reinforced area would have made it to at least 100%.
OK, but what about the rest of the fuselage? The rest was only tested to 99%. Couldn't there be another place that fails at, say, 99.5%, which wasn't detected in the first test because the place that fails at 99% stopped the test?
Edit: I would guess that during this kind of test they have the plane heavily instrumented with things like strain gauges so they can understand how it reacts and see if it is behaving according to spec. So maybe it is the case that the data from the rest of the fuselage showed that it would have made it to 100% if the test had continued, and only the area that failed was in trouble. That could explain why they aren't require to do another test.
That still leaves me unsatisfied, though. Presumably analysis of the design before the test said it would make it to 100%. If the failure was not due to a manufacturing defect in that particular fuselage, then it suggests that they analysis was not adequate. But then what assurance is there that their analysis that shows that the reinforcement would hold is good?