How do you design something as complex as a submarine

I’m watching Das Boot for the first time and am impressed by how complex the submarine is. And I’m sure there are other feats of engineering that are much more complex.

So I’m wondering, how do engineers know down to the most minute details how to build something that large, complex and interdependent? Can you just tell a team of 5,000 engineers ‘design a submarine’ and then it will be designed in a way that works well, down to the most minor details?

How do they even know if the parts will all work together, or will withstand real world testing? Do they just build a model, then fix the kinks, then do that again and again until they have a working model? Or do they just assume everything will work together? Or do they just tweak an existing model?

Does the engineering team get broken up into various groups and subgroups? How would the teams be broken up?

I have no background in engineering, so I don’t know how any of this works.

This is a damn good question; I, too, would like to know the answer. All I can say with regards to submarines is that I was inside the captured Nazi submarine that’s at the Chicago Museum of Science and Industry, and I learned there that being in a Nazi submarine was, like most things associated with Nazis, horrible.

They had something like an 85% mortality rate.

I’m not a physical objects style engineer, but I do note that most buildings simply reserve a space and expose some tubes for things like the washing machine, dish washer, etc. when the building is under construction. They don’t need to know exactly what machine nor exactly what size machine will go there, nor do they need to know anything at all about how it works internally.

My presumption would be that they do something similar, where there’s a lead engineer who gets some basic dimensions for each thing and the I/O (water, fuel, electricity, etc.) that’s necessary for it and does up all of his plans with the expectation of a perfect cuboid, slightly large than the dimensions he was given, that can accept cables and piping from any direction. The lead engineer will have some basic parameters like that overall, people need to be able to move around the ship and the whole thing can’t weigh more than X, can’t supply more than Y water, Z fuel, nor W electricity in total, and so he’ll work with all of the available cuboids he was given, see if it all fits under some arrangement, and if not then he’ll need to call around to different makers and ask them to custom make something.

On the maker’s side, they won’t know anything at all about the submarine, they just know that they were told to make their widget no larger than some specific dimensions and have an I/O of whatever. And while they’re building their widget, they’ll do the same thing. There will be a lead engineer who knows what all of the major subcomponents are, maybe takes some off the shelf parts and orders others to be designed from scratch by his minions. He’ll give them specific dimensions that they have to hit, and it’s on them to make sure that they hit the requirements. If somehow they can’t, then it will have to go back up to the lead designer and he might have to rejigger everything and see if any other components are coming in smaller than originally expected, etc.

Back in the 40s, following that step maybe they would have gotten a model maker to actually go and craft little versions of everything (so they don’t have to treat it all as cuboids) so that they could glue it all together in their model submarine shell and see if it all fits nicely and if there’s any spare room for more goodies. These days, they’d probably ask the maker to send them files for some sort of 3D program (and probably would have skipped the initial step with cuboids).

But even then, assembly (particularly for your first version of the submarine) is going to be a major undertaking because you’re suddenly going to find out that measurements were wrong, that the position of the I/O ports are all in completely horrible locations, that a device gets super hot and needs an extra 4 inches of insulation around it, that some machine is ridiculously failure prone, etc. They’ll have to build custom stands and mounting points for everything, figure out how to feed wire and tubing throughout the ship and where to install maintenance access points, etc.; they’ll need to request that some stuff be replaced with different hardware; and they might end up performing some jury-rigged modifications on the hardware they got.

Through all of this, they’ll probably end up developing more detailed designs, they’ll build sub-blueprints for small parts for installation - like mounts for different parts - and so on so that the next time they build the ship they can do so more quickly. But likely each ship will end up slightly different due to assembly-time problem-solving.

But so overall the strategy is: Visualize a bunch of cubes. Find things that fit in those cubes or order something that will. Repeat as necessary for subcubes, sub-subcubes, etc.

The biggest issue with a sub is most of the guts have to be there before you put it together, or they fit through a hatch 30" inch in diameter.

To put Sage Rat’s excellent description in a word, it is modularity and hierarchy. And specialization. I don’t know about submarines but I do know about the design of billion transistor microprocessors. You start at the top level with modules like processors, memories, cache controllers, I/O controllers, etc. You have teams designing each one to a spec. You have other specialized teams routing signals between these and making sure the global timing works. Then you have other teams working on verification and manufacturing test. It might take hundreds of people, but it is not chaotic and everyone knows what he or she is doing.

I’m sure submarines, like microprocessors, are not all that different from the ones that came before, so they start with a known template and evolve the design from there.

You take an earlier model of submarine and improve it. Most technology is a result of evolution, not intelligent design.


The designers are not starting off with a blank slate. They are starting off with the detailed technical knowledge and specifications of previous submarines, and the records of how they performed under real life conditions. Then they are making incremental improvements.

Usually the lead engineers would have many years, maybe decades, of experience in submarine design, and the manufacturers and builders likewise would have plenty of experience in building earlier submarines.

This. Start with the big picture - what you want, and what you want it to do. Figure out what systems you need to accomplish those goals. Break everything down bit by bit, until you have individual parts. Throw that design document at an engineer or five and say, “GO!”

An engineering degree gets you part of the way. You spend four+ years learning the basics of machine design, fluid mechanics, thermodynamics, heat transfer, metallurgy, and so on. Or you go for a couple more years to get a master’s degree in a more specialized area. Or you go for several more years to get a Ph.D. in a very specialized area.

“Institutional knowledge” is another part of the puzzle, and comes with experience. Some of that is in the heads of the engineers. A senior engineer from Ford will know from experience roughly how far apart to place the attachment points for a door panel, dashboard, or headliner so as to minimize buzz, squeak and rattle, something that submarine designers won’t know. He’ll know from his engineering classes that a submarine propeller blade will utilize a “wing” cross section to produce thrust, but he won’t know exactly what shape is most efficient, least noisy, or least prone to cavitation under various conditions - something that experienced submarine designers will know.

Some of that institutional knowledge is proprietary, i.e. not all submarine designers will know it. If your company has invented an alloy that’s particularly resistant to seawater corrosion, or developed a propeller that’s ultra-quiet, or a novel welding process that makes explosion-resistant hull joints, you’re not going to share that with your competitors.

Some of it is incremental development: you take one of your submarine designs that worked pretty good in the past, and you improve on it.

Some of it is independent design and testing. The components of a submarine are not 100% independent, so you can parcel out the work to separate design teams and bring it all together at the end. You can find examples of this in other industries. Here, for example, is a lab test of the Airbus A380’s landing gear: the wheels are are spun up to touchdown speed, and the whole landing gear is slammed to the ground at “hard landing” vertical speed, with all of the momentum on it that a real A380 would have during a hard landing. Likewise, here’s a laboratory brake test, in which they simulate braking the A380 to a stop in an overweight condition with no reverse thrust. How does the brake behave? Does it disintegrate? Does it catch fire? If so, does the fire spread? There will be a full-up aircraft test like this later on, but you don’t want that to be the first time you’ve tried using the brakes; it’s possible to gain confidence first with test cell work like this. There’s a lot of the same sort of thing you can do for a submarine.

And no matter what you make, there’s a lot of testing/redesign iteration along the way. Decades ago you’d make scale models or full-sized prototypes based on your engineering degree and institutional knowledge, test them, and then refine your design based on the test results. This cycle is much faster and cheaper these days thanks to computers. You can design a part or assembly on a computer, and then use finite element analysis to assess various performance aspects. Is it strong enough? Are there stress concentrations that need to be alleviated with a little more/less material in certain spots? Are the heat transfer properties adequate/optimized? More/thinner/thicker cooling fins needed? Problematic vibrations that might be fixed with more/less mass here and there? Is it managing air/water flow/pressure as needed? You can do all of this in the computer and come up with a design that’s very likely to be very good the first time you make it.

Computers are also great for coordinating the work of all the subgroups. A central archive can hold the master plan for the submarine that hosts all of the subsystems being designed by the semi-independent groups, so that everyone can see how the different subsystems are going to mesh in the final assembly and work in a cooperative manner.

And in the end, you make your full-scale, fully-assembled prototype, and you test it as incrementally as you can. At this point it’s unlikely that you’d have to scrap the entire design, but the computer analysis is never perfect; you’ll always find little things that need adjustment. For airplanes, you start with a high-speed taxi test, in which the plane doesn’t even leave the ground. Then you take off for a very basic flight, never even raising the landing gear. You build confidence in the plane as you put it through harsher and harsher tests (e.g. a high-speed flutter test. Similar testing happens with submarines and other large machines.

I have a degree in structural engineering, but never really practiced, so that that as you will.

As others pointed out, a lot of it is breaking it down into systems and subsystems and hierarchies. For example, a submarine is essentially just a big pressure vessel. Similar to a tank that one might use to store propane or an aerosol can. Except the pressure is on the outside. There are pretty well established mathematical formulas for calculating how thick a material needs to be to support such and such pressure pushing in on it for a vessel of so and so size.

Of course the trick is you can’t cast a big U-boat as one monolithic entity. And even if you could, you would still need holes for hatches, screws, periscope, exhaust, torpedoes, venting for ballast and whatnot. So now engineers break down these different systems, figuring out the requirements for each.

A whole lot of project planning and management goes into it as well. That is to say, figuring out the dependencies of tasks and sequence for which the can be addressed.
Also note, that as technology matures in any particular area, they don’t need to redesign everything from scratch. Like when they design and build a new airplane, most of the time, they are designing around existing engines and the people who design engines design them for fairly standardized engine housings.

There are actually a fair number of videogames like Kerbal Space Program, Factorio, Oxygen Not Included and Space Engineers that give you a sense of how to design and build complex systems. Often what makes them fun is when a weakness is exposed causing a “cascading failure”. Like when my outhouse breaks down in ONI and soon my base is drowning in it’s own vomit.

Each small team of engineers works on a small part of a submarine.

Each new submarine is a modification of a submarine previously built.

And proof of this is the set of odd features found in any nth generation design, which were good ideas (or maybe bad ones) when added originally and which have been kept in because taking them out would be risky and might delay the project.
Just like the odd features found in most evolved plants and animals.

And other teams coordinate stuff. Also, some blocks are designed already: you most likely don’t need to design… toilets, bunks or GPSs, just pick which ones to use.

I was a tiny part in a project to design a new motor. The first few years were going to be spent in design, make, redesign, make again… and this had to be coordinated through multiple factories in several countries. The design team was a handful of people, all sitting in the same room, but they had to coordinate among themselves (if you change one piece, you have to check the designs of at least all the pieces it’s in contact with) and with the factory teams (can you do it, if I change this detail? How long will it take to get the machines adapted?). That motor was eventually, hopefully, going to be slotted into place in bigger designs he way one slots dishwashers, but that requires the designers for the slotable item to take sizing standards and connectors into account.

Former submarine officer here. After I graduated from college with an engineering degree, the U.S. Navy spent 18 months training me just to safely *operate * a nuclear-powered submarine (much less design anything). I can’t tell you how many times I was awe-struck at the ingenious engineering that was evident in the design.

As others have indicated, you have engineering design teams that specialize on each small part of the sub, and they build on previous designs.

Some submarine design improvements involve a great deal more effort, such as when the first nuclear power plants were designed for submarines. The Navy (and its contractors) tried out a variety of competing designs before settling on a standard, which has since been modified more incrementally.

This model for building on previous designs is one reason why the U.S. Navy does not want to ever stop building nuclear submarines. Otherwise all of the expertise and institutional knowledge would be lost as people retire (and pass away). This is also why the Navy has supported two shipyards capable of building nuclear submarines: General Dynamics Electric Boat in Groton, CT and Newport News Shipbuilding in Virginia.

On a related note, while submarines are complicated, they are not nearly as complicated as manned spacecraft and rockets. After the Apollo program ended, we allowed much of the associated expertise and institutional knowledge to wither away. This is one reason why there haven’t been any manned missions beyond low Earth orbit (LEO) since then. I’ve heard it said that we couldn’t send a man to the moon right now even if we had the will and desire to do so. After nearly 50 years, much of that knowledge and expertise would have to be recreated nearly from scratch.

There’s a Power Point for everything. You just need to know where to look.

As others have said, hierarchy and specialization are key.

There are tradeoffs here, though. As Sage Rat pointed out, for something like a building, most of the components are left unspecified. Standard electrical sockets are put in, plumbing is routed, but beyond that there is flexibility in where things are put. Buildings have a great deal of margin available–they are massively overbuilt, and so not very sensitive to moving weight around or other variations.

On the other end of the spectrum, we have rockets, which are exquisitely sensitive to mass, power, temperature, vibration, and so on. Virtually every component can, potentially, cause a failure in the entire system. This makes rockets very expensive–you can’t just buy an off-the-shelf whatever and plug it in; it has to be qualified for use and integrated into the full system. Rockets have almost no margin to them; if a particular component is twice the mass that it should have been, it probably just won’t work. It has to be redesigned.

Somewhere in the middle we have things like submarines and automobiles. They are not as free-form as buildings, but they have a decent amount of leeway when it comes to mass and other things. If the engine produces 10% less power than it should; well, that’s unfortunate but it probably means the vehicle is just a little bit slower than we’d like. Contrast with a rocket where if the engines underperform by 10%, then it doesn’t make orbit and has to be destroyed.

The more margin you have, the more flexibility you have in the hierarchy and the cheaper it gets, because it’s much easier to make tradeoffs. Design variations have only local effects. But as margin decreases, the hierarchy becomes rigid, up to the point where a single variation anywhere in the system affects the entire system. Design becomes very expensive.

The Navy actually uses relatively few commercial off-the-shelf items on submarines (with some exceptions, like computer laptops).

To discuss your three specific examples:

Submarine toilets are connected to sanitary holding tanks that can be pressurized to overcome the surrounding sea pressure to empty the tanks (while far out to sea). The toilets have a stainless steel ball valve at the drain that can withstand this pressure. The toilets themselves are also made of stainless steel (as are the sinks and showers). Porcelain doesn’t work all that well in a vessel that might take battle damage from exploding depth charges.

Similarly, with the possible exception of prisons, nobody would have a need for a bunk with as little space as on a submarine. :wink:

Finally, GPS units utilized by the military are military hardware, not civilian models. The Global Positioning System (GPS) (originally the Navstar GPS) is a military satellite system that is owned and operated by the U.S. Air Force. Should the U.S. government choose to do so, it can selectively deny access to the system or degrade the signal at any time. An encrypted signal can be maintained for U.S. military units and our allies. You need a military-grade GPS unit to be able to utilize the GPS network in such a scenario.

Besides that, you want a GPS unit that can work reliably in a submarine environment, and which can be utilized with receivers mounted on submarine masts and/or antennas.

This specialized equipment is one reason why military hardware is so expensive.

Definitely. Even though a submarine has millions of details, most of them are in blocks designed by generations of engineers before.

These might not be commercial to civilians, but that doesn’t mean that I need to go request a fresh redesign of a GPS system just because I’m building a new submarine design. There is, presumably, a catalogue of “off-the-shelf” items that exists solely among the military world.

New submarine classes come around so infrequently these days (like once a generation) that virtually every integrated system on a new class of submarine is indeed redesigned.

But again, you are building on previous designs. While a Los Angeles-class submarine has an S8G reactor (i.e. 8th generation reactor designed by GE), the Virginia-class subs have an S9G reactor. It’s not like you’re starting from ground zero.

Also, many systems are upgraded along the way, especially with the electronics (including software upgrades), as well as systems that can easily be swapped out, such as towed-array sonars and weapons.

For example, the Los Angeles-class submarines started out with Mark 48 torpedoes, then were upgraded to carry Mark 48 ADCAP torpedoes. Virginia-class subs also currently carry Mark 48 ADCAP torpedoes. These torpedoes are themselves upgraded on a continual basis.