4 ms·

I think as you do more pieces you can invest in more automated ways of doing QC as well. As the increased fixed cost for the QC infrastructure is amortized acr

I think as you do more pieces you can invest in more automated ways of doing QC as well. As the increased fixed cost for the QC infrastructure is amortized across the larger number of produced units.

TeMPOraL 3y ago
It all adds up. Scaling up production requires simplifying both product design and the entire process, minimizing the amount of complex manual work and bespoke components. That reduces QA/QC workload as well.
- bumby 3y ago
  A lot of the QA cost is with pedigree and traceability and that doesn’t scale nearly as well. A $1k bolt isn’t expensive because it’s a bespoke design; it’s expensive because you have to maintain tight control from everything from material coupons to storage. This is one area that makes a lot of manufacturers balk at the selling to aerospace companies; the amount of QC work just doesn't make sense for the amount of product they're making. Another area is supplier control, which SpaceX has struggled with in the past. That doesn't scale well either, but maybe they can bring that cost down by vertical integration of manufacturing (although that doesn't really help as much for the pedigree cost).
  - TeMPOraL 3y ago
    Yes, but I believe there are still synergies here. A $1k bolt may be expensive because of QA, but if you're making/buying enough of them, it could make sense to change the QA process to one that has higher false positive rate, but is cheaper to perform. Even if the change makes you throw every other bolt out, most of them perfectly good by the old process, if the new process means dropping per-item-cost from $1k to $250, it's very much worth it. In a similar way, increased scale also makes it reasonable to invest up front to redesign the final product so it doesn't need any of those $1k bolts in the first place.
    - bumby 3y ago
      I think this is reasonable, but there’s a couple important points: Aerospace quality is largely driven by reducing false negatives. You’re generally more ok throwing out something good than accepting something bad. Reducing false negative is what becomes costly. Hiding quality issues by increasing throughout is a classic fallacy in the process control world because it increases waste. Sorry if I misinterpreted your point here. What are you thinking on trends of revamping QA processes? Maybe I’m not imaginative enough, but I’m not sure what kings of different prices controls would bring down costs dramatically. Is there some specific new NDT you’re thinking?
      - TeMPOraL 3y ago
        > Aerospace quality is largely driven by reducing false negatives. You’re generally more ok throwing out something good than accepting something bad. Reducing false negative is what becomes costly. Right. But the special case here is that if you can eliminate part of the process entirely, you no longer have to pay for reducing false negatives in it. > Hiding quality issues by increasing throughout is a classic fallacy in the process control world because it increases waste. Sorry if I misinterpreted your point here. All I'm saying is that trying to increase throughput will force you to simplify and streamline, which often will increase waste, but still significantly lower the total cost. Random example: CNC milling is incredibly wasteful compared to some more complex (and more manual) processes, but it's also much simpler and faster, and this saves you way more money across the board than you lose over wasted material. Same principle applies to QA: if you have two tests with equal false negative rate, with the second one having 2x the false positive rate but being 4x cheaper to run, it's worth it to switch to the second one - while this doubles the waste at this stage, it also lets you double the production (or just save you half the costs). Now, to give specific SpaceX / aerospace-relevant examples, I'd have to dig through some of the things I've read over the years that gave me this overall impression. Having not done that, all I'm saying is that there are strong synergies involved in scaling production up. You don't just get bulk rates and pump stuff out faster - you get better at it faster, which lets you optimize it further and do it quicker. Simplifying design to simultaneously cut costs of parts, assembly and QA, is just one specific case. > Is there some specific new NDT you’re thinking? Not really, not right now. I could look over my notes and some references, and perhaps point something out - but that would really be pretending I know more about manufacturing processes than I do.
bumby 3y ago
Maybe, but I think the tech in that is still some times away. For example, there’s a lot of borescope inspection on a reused engine. Maybe we can eventually train image recognition software to help, but most likely this is still going to be done by humans in the short and medium terms. A lot of QC boils down to validated trust, and I think we’re some ways away from outsourcing that to machines in safety critical spaces like spaceflight. You actually hear a lot of that in the industry. “We don’t make widgets” is a common refrain to illustrate that mass production principles don’t always completely map to the space sector.

I think as you do more pieces you can invest in more automated ways of doing QC as well. As the increased fixed cost for the QC infrastructure is amortized across the larger number of produced units.

TeMPOraL 3y ago

It all adds up. Scaling up production requires simplifying both product design and the entire process, minimizing the amount of complex manual work and bespoke components. That reduces QA/QC workload as well.

bumby 3y ago
A lot of the QA cost is with pedigree and traceability and that doesn’t scale nearly as well. A $1k bolt isn’t expensive because it’s a bespoke design; it’s expensive because you have to maintain tight control from everything from material coupons to storage. This is one area that makes a lot of manufacturers balk at the selling to aerospace companies; the amount of QC work just doesn't make sense for the amount of product they're making. Another area is supplier control, which SpaceX has struggled with in the past. That doesn't scale well either, but maybe they can bring that cost down by vertical integration of manufacturing (although that doesn't really help as much for the pedigree cost).
- TeMPOraL 3y ago
  Yes, but I believe there are still synergies here. A $1k bolt may be expensive because of QA, but if you're making/buying enough of them, it could make sense to change the QA process to one that has higher false positive rate, but is cheaper to perform. Even if the change makes you throw every other bolt out, most of them perfectly good by the old process, if the new process means dropping per-item-cost from $1k to $250, it's very much worth it. In a similar way, increased scale also makes it reasonable to invest up front to redesign the final product so it doesn't need any of those $1k bolts in the first place.
  - bumby 3y ago
    I think this is reasonable, but there’s a couple important points: Aerospace quality is largely driven by reducing false negatives. You’re generally more ok throwing out something good than accepting something bad. Reducing false negative is what becomes costly. Hiding quality issues by increasing throughout is a classic fallacy in the process control world because it increases waste. Sorry if I misinterpreted your point here. What are you thinking on trends of revamping QA processes? Maybe I’m not imaginative enough, but I’m not sure what kings of different prices controls would bring down costs dramatically. Is there some specific new NDT you’re thinking?
    - TeMPOraL 3y ago
      > Aerospace quality is largely driven by reducing false negatives. You’re generally more ok throwing out something good than accepting something bad. Reducing false negative is what becomes costly. Right. But the special case here is that if you can eliminate part of the process entirely, you no longer have to pay for reducing false negatives in it. > Hiding quality issues by increasing throughout is a classic fallacy in the process control world because it increases waste. Sorry if I misinterpreted your point here. All I'm saying is that trying to increase throughput will force you to simplify and streamline, which often will increase waste, but still significantly lower the total cost. Random example: CNC milling is incredibly wasteful compared to some more complex (and more manual) processes, but it's also much simpler and faster, and this saves you way more money across the board than you lose over wasted material. Same principle applies to QA: if you have two tests with equal false negative rate, with the second one having 2x the false positive rate but being 4x cheaper to run, it's worth it to switch to the second one - while this doubles the waste at this stage, it also lets you double the production (or just save you half the costs). Now, to give specific SpaceX / aerospace-relevant examples, I'd have to dig through some of the things I've read over the years that gave me this overall impression. Having not done that, all I'm saying is that there are strong synergies involved in scaling production up. You don't just get bulk rates and pump stuff out faster - you get better at it faster, which lets you optimize it further and do it quicker. Simplifying design to simultaneously cut costs of parts, assembly and QA, is just one specific case. > Is there some specific new NDT you’re thinking? Not really, not right now. I could look over my notes and some references, and perhaps point something out - but that would really be pretending I know more about manufacturing processes than I do.

bumby 3y ago

Maybe, but I think the tech in that is still some times away. For example, there’s a lot of borescope inspection on a reused engine. Maybe we can eventually train image recognition software to help, but most likely this is still going to be done by humans in the short and medium terms. A lot of QC boils down to validated trust, and I think we’re some ways away from outsourcing that to machines in safety critical spaces like spaceflight. You actually hear a lot of that in the industry. “We don’t make widgets” is a common refrain to illustrate that mass production principles don’t always completely map to the space sector.