[gridspy]

737-Max crashed because of the "lets analyse this rather than test it" thinking that you described earlier. That and an overly cosy relationship with the FCC and a culture of silencing cautious engineering voices.

It's seemed cheaper to tack larger engines and new software onto the old air-frame than to redesign the aircraft to be mechanically stable and update the cockpit.

Not surprising that the same company (Boeing) that campaigned for analysis over testing in spaceship design (they claimed SpaceX couldn't do commercial crew) also failed to design a working and stable aircraft.

I'd argue that the bureaucracy involved in "signing off" an new aircraft design contributed to these problems. Designing a aircraft or spacecraft from first principles and then refining through testing does have its advantages.

My point about aircraft was that the lift vector is perpendicular to the wings, even if the aircraft flips (rolls) upside-down there is still a force that pushes you toward the deck if you are inside the plane. In the StarShip there is a similar upwards force from the engines (when lit) towards the nose regardless of which direction the ship is oriented with respect to the Earth.

The reason why flying a plane in clouds is so dangerous is that your sense of up and down in a plane is invalid. You'll always think the deck is down without the aid of either a view of the horizon or the correct use of functioning instrumentation.

[GizmoSwan]

Nope. A purposely designed aircraft is simulated is not one that is elongated with flight control data being patched.

Example of how it works: https://www.darcorp.com/advanced-aircraft-analysis-software/

The bureaucracy had nothing to do with it. They did not want to spend the money to create a new craft. The long bodies aircraft was not airworthy physically speaking and so the software was designed to limit its pitch to prevent it from stalling.

All aircrafts can stall and some angle of attack but a long+narrow aircraft is way more sensitive.

Design of an aircraft from scratch would have flagged it as poor design right away.

[Cogito]

As pointed out in some other comments, SpaceX do extensive simulations, they almost certainly have the best rocket simulations in the world (see some of the published talks on youtube to see the state of their research).

We know that their simulations/calculations were excellent for these last couple of Starship tests because the tests themselves were excellent. There were a number of things being tested, including things like flying multiple Raptor engines, turning those engines off, switching between different fuel tanks for launching and landing, the bellyflop maneauver, the landing flip maneauver, engine relight from the header tank, and finally landing. Almost all of these were perfect, from what information we have.

There have been two failures, and from what we know these were not related to simulations at all - they seem to be system/integration issues, though we don't really know much about the engine failing to relight this test.

On SN8 they had a pressure issue feeding the engines from the header tanks. It may have been possible to anticipate this or test it independently, but it was probably the least important of all the things being tested. Given that they have multiple new (and updated!) test articles rolling off the production line they clearly made the choice to conduct the test without spending years trying to validate every single component. Again, they obviously did a lot of excellent engineering despite not undertaking a long pre-flight validation, as the test was so succesful.

[gridspy]

Both you and I agree that aerospace companies should do large scale vehicle simulations, which they do.

In addition SpaceX does complex fluid dynamics simulations to design and test their engines. That's likely part of why they have successfully made the "holy grail" of Full-Flow staged combustion work in the Raptor. [0]

None of the failures so far could have been detected through simulation since they were complex integration issues. At some point the cost of simulation is vastly larger than the cost of just creating a test article.

Components like drogue chutes, explosive bolts, failsafe valves and abort systems are typically validated both analytically (by reviewing designs) and are often hard to fully validate by testing.

Since validating components is so expensive, it leads companies that rely on analyitical validation (like Boeing) to become overly attached to previously analysed and working components because analysis is so time consuming and expensive. You end up with modern rockets built with space-shuttle engines as components (SLS) and old jets with new engines tacked on (747 Max).

When you talk about explosive deconstruction of a test article after something new was learnt you fail to realise the learning (not the landing) was the goal. If SN9 had landed correctly, it is very likely they would have deconstructed it anyway. They just deconstructed SN12, 13 and 14 to make way for SN15 which has a very improved design.

0: https://www.youtube.com/watch?v=LbH1ZDImaI8 "Is SpaceX's Raptor engine the king of rocket engines?"