Comment by supernova87a | Hacker News Reader

Comment by supernova87a | Hacker News Reader

Parent Full thread

Maybe someone with more industry knowledge or interesting sources can corroborate --

But I believe that one of the things that happened to the military aviation industrial complex was that in the WW2, early Cold War era, planes and military development programs were very much driven by individuals or very small teams responsible for designing the requirements, overseeing the planning, and the outcome of new aircraft. That was how you got such distinctive (and long lasting designs) that served well the needs of the military, even in use up to today. Let's call it (using today's fashionable term, "passion" about the aircraft, and unwillingness to put out a piece of crap.) You could name who was responsible for an aircraft design.

But in the 1970s, 1980s, something happened where the size of projects or the professionalization of military projects caused them to have distributed (read, lack of) ownership by very critical people, and instead by an anonymous/dispersed "committee style" management. Also, the experts within the military often left for better pastures at the contractors (NG, LM, etc).

Add to that the requirement (by Congress, regulation) that you couldn't just use your expert / private connections to other experts to "get things done" and bring in contractors who had expertise, and instead had to farm out RFPs to every company, get their bland input, and fill out a heat map of who did what best, and cobble together a solution "mutually acceptable to all stakeholders".

So no longer was an aircraft project owned by someone who had very certain opinions about what was needed, and not needed, and who something should be designed for. But instead, projects follow a certain recipe, hoping that form produces function, yet not actually produce useful output because of what happens when you lack ownership. You cannot today name who is responsible for an aircraft design (I would say). It is the metaphorical side project among many, of a large dispersed group of people.

It seems to me something was missed in the migration to this new method of working.

But as I said, I would love to hear corroborating observations about whether this was a major factor.

At least three things happened:

- congressional oversight increased, with the result that the purchasing authority is no longer generally the end user. We aren’t just buying a ship (plane, gun, etc.), we’re maintaining a logistics and manufacturing industrial base to produce that type of item (and maintain jobs in congressional districts). So maybe we should buy a few more tanks/ships/bombs than we need so that we don’t forget how. (A bit dubious, but it’s one of the reasons we make the funding decisions we do.)

- risk tolerance shrank. Especially in war time there was a high tolerance for risk if the payoff was big enough. These days our risk tolerance is basically zero. I once had to get an admiral to sign off on a risk hazard assessment with an estimated likelihood of ~1e-18 (you’re more likely to win the powerball jackpot twice). There was a cap that might pop off an item at high speed during flight, and if it hit someone they would probably miss a day of work. By the likelihood/consequence framework for risk management, that meant it required an admiral’s signature. You can’t move fast in an environment like that.

- complexity increased. Old schools systems like a WW2 jeep were designed and prototyped in 40 days because they were incredibly simple and made liberal use of COTS parts. There’s an argument to be made that maybe we don’t need such complex systems to meet requirements, but I don’t think anyone is arguing that we should be replacing all the F-35 fleet with a WW2-style aircraft. (Although there have been proposals to build super-simple aircraft for CAS missions, etc.)

I am not sure about "incredibly simple" in general though. Look at how insanely sophisticated the P-51D fighter was, for example. But sure, something in engineering as we see it today, and around it, has been lost.

Or the B-29. That was arguably one of the most sophisticated (and expensive) aircraft of the war.

> - risk tolerance shrank.

There's a lot of truth here and it plays out in even more ways than that. I look back through documentation of legacy systems and see so many things designed based on "engineering judgement". Sizing for that bearing? Engineering judgement. Thickness of other minor structure? Engineering judgement.

Now, the load on every little thing requires a 50 page engineering calculation report with figures, charts, and tables, that takes 3 months to prepare and another month to get reviewed by everyone. The littlest things now take forever.

There's a lot of things changed since the era of 'engineering judgement'.

In the 50s and 60s when 'engineering judgement' was common, it was a signal to a skilled machinist that "You have more experience with the material, so use your best guess because I don't know what the right answer is".

Things have changed a lot!

1. Material quality is much higher. Which is to say that batch-to-batch variance is vastly reduced. This means that it's worth investing the effort in detailed understanding of the material because that material is more predictable. Which led to ...

2. We understand a _LOT_ more about material science. Our knowledge of materials is vastly higher, so things that used to be 'best guess' are now 'do this because it will reliably work'. e.g. Spallation and galling used to be poorly understood issues that were worked around based on personal experience and guesswork. Now we understand them very well, and any competent manufacturer will clearly explain exactly how to (eg) install their bearing in a way that prevents such issues.

3. Expertise shifted. The 'machinist' is no longer a 20-year experienced highly trained person, but a CNC operator, who probably won't be within eye-sight of the running machine. So the practice of 'increase the RPM until it chatters slightly then back off' doesn't cut it any more. The answer needs to be known up-front to go into the G-code, not 'feel'. This has led to the obvious cycle where more responsibility moved to the engineering end, which increased the demand for exact knowledge, which reduced the requirements on machinists, which led to lower skill, which further moved responsibility to engineering.

So I totally get pining for the era of 'engineering judgement', but it died for a reason and it's not really likely to come back!

I would've thought your first 2 items there would help increase the use of engineering judgement by reducing pitfalls previous generations encountered. Your third item is definitely a significant factor.

It's not so much that I yearn for the era of engineering judgement so much as I yearn for refocusing on work that's important. Not every little decision needs to written up in professional report, circulated around through multiple drafts, signed off by 4-5 people, briefed out in slide deck form, then never read again.

Ok, boomer.

That would be nice, but you just wait until the root cause analysis of a catastrophic failure gets traced to a backside that was inadequately obscured by "engineering judgement".

A month for review? Where is this glorious Valhalla?! Navy flight certification checklist is a 2-3 column full page of densely packed names for the approval authority on all the different systems. It may take a month to circulate your proposal to those folks just to get a final list of the subset who will need to ultimately sign off. We had cases where item X was used on USAF aircraft Y, but it could not be used on Navy aircraft Z (which was the same airframe) without literally tens of millions of dollars in testing. Not because the loads on the platforms were substantially different, but because the test standards between USAF and USN were different and the approval authorities wouldn’t sign off otherwise. (See also: zero risk tolerance.)

A month for review is an achievable-but-best-case scenario for projects that need to get out the door ASAP, but its not unusual for these reports to be in review for many months. I know of one project report that's been "in review" since 2015.

> So maybe we should buy a few more tanks/ships/bombs than we need so that we don’t forget

Ironically most modern weapons programs do the exact opposite.

Costs increase.

So they cut orders.

Which just makes the per unit costs increase even more.

So they cut orders again.

Repeat until program fails.

Right, the death spiral. But I’m actually talking about the infuriatingly counter-intuitive headlines about buying tanks that the army says we don’t need, or ships that we know will never work as planned. There are only so many places that can build such things, and if we let them go to mothballs then we lose all the manufacturing experience to do it at all—which is a risk for national defense. So the Navy says “I need $40m for planes!” but Congress takes $100M from the plane budget line to keep open the shipyard in Sen. So-and-so’s district. Seems terrible, but that might be the only place on earth that employs the folks who can build such a ship and that keeps it alive. I’m not saying it’s worth it…just that it’s worth considering.

If you just want make-work, why not build schools, parks, and public hospitals instead?

Or useful infrastructure, at least.

It’s not about make-work. It’s about the intersection of what the military says it needs, what congress folks can get away with, and what contractors lobby for to “maintain the industrial base”.

Building parks & hospitals doesn’t contribute to national defense, and it doesn’t develop or maintain the skills of welders who can join absurdly thick plates in the hull of a ship, or folks who can work with composites and thin aluminum, etc.

I’m not arguing that the system is working or that we have the right balance. But it’s not a simple “don’t build bombs, build hospitals” dichotomy.

> Building parks & hospitals doesn’t contribute to national defense

Neither does spending astronomical sums on a system that never reaches operational readiness.

Unless it maintains the ability to produce systems for national defense.

Risk tolerance is not just engineering, but also how comfortable we are risking human lives.

In a "good war" like WW2 we should recognize that fighting that war is worth the risk and that casualties are going to happen. We don't need our own aircraft to compromise their performance to be 99.9% reliable instead of 99%, because we strongly believe we need to e.g. beat Hitler.

That’s true, but the uniform service folks are much more willing to take risk than they’re allowed to by the oversight process. That’s a big part of why it takes so long to develop and field new stuff. I understand why we do it, but I’m not convinced that we have the balance right.

On risk tolerance, in WW2, the accident rate for a single seat fighter, the P-51, was 274 per 100,000 flight hours[1], for the last 10 years, that rate for f-16's is below 10, and for fatal accidents it's 0.76[2]. Although I'm not sure it's fair to make a 1-1 comparison as the modern accident rating system includes a few scenarios, the orders of magnitude difference is a great illustration of your point!

[1] http://pippaettore.com/Horrific_WWII_Statistics.html [2] https://www.safety.af.mil/Portals/71/documents/Aviation/Airc...

Exactly. Those accident rates drove the creation of the modern flight clearance system that creates such an onerous risk management process today. It was very successful in reducing casualties, but the consequence is slowing down to an absolute crawl.

COTS = Commercial off-the-shelf

CAS = Close air support

> - complexity increased

The US almost never operates in contested airspace. For that, the F-35 (as the F-22 before it) is overkill.

Do you see a need for manned weapon systems that can survive engagement with other manned weapon systems in the near future? I'm thinking semi-autonomous drones can replace a lot of such systems and, if you can build them quickly and cheaply enough (as in it takes a $1M missile from a $50M plane to down a $1M drone), then you can build lots and lots of them to overwhelm your opponents.

Risk tolerance, even outside of wars, has shrunk quite a bit. People used to do things that are considered practically suicidal now. In 1956, only 2% of Ford customers opted to have seat belts in their cars, and even for the 2% who did have them, those cars were complete death traps compared to modern ones.

The F-35 broke the most important of Kelly Johnson's rules.

15. Never do business with the (damned) Navy!

Reply on news.ycombinator.com