Episode 5: Tackling nitty-gritty details, FEA, and the difficulty of urban boat building

Chris Sullivan
Jul 21, 2017 · 11 min read

I just realized it’s been six months since I published my last update (it feels like no more than two), and almost a year since I started this blog. I think my original plan had me sailing to Cabo by now.

“How am I so far behind?” he asks himself rhetorically, before the helpful little voice in his head chimes in, “…you should have tried designing a friggin’ parametric chair first.”

Fig 1: My original “parametric chair” software modeling idea. Could leverage powerful computers to adjust any of at least 3 inputs. Game changer.

I won’t bore you with excuses, like helping coordinate my brother and best friend’s bachelor parties and weddings in a two month window, on top of an intense quarter-long work project… Hamilton pulled off a wedding while fighting the Revolutionary War, so no excuses, play like a champion.

So now, back to the story:

We left off with my having finished the stability analysis necessary to convince myself of basic seaworthiness for a theoretical design. I had also integrated a whole new host of seating arrangements, railings, bow coverings, etc. A lot had happened.

I was pretty pumped about all the progress, but much of it had been hurried design work so I could get to the stability analysis. So for the next few months, the scant free time I did have was spent cleaning up a million small details I had rushed before.

It was all I had time for, but it was also a really good strategy for feeling productive with a hobby while swamped by life: “Today, I will accomplish at least this one small improvement for my boat design” was the goal I’d set out for myself, and every time I did it, I felt a small sense of accomplishment, like I wasn’t completely neglecting my project.

At one point, it had been so long since I had looked at the design that I literally had forgotten where I left off. Notes are helpful, but when you’re dealing with work-in-progress code, context is king. So I decided to use the opportunity to look at the design from the perspective of someone with no project familiarity. How would I explain this to that person, and in what order? In doing so, I would figure out where I had left off.

I’d need some kind of instruction manual.

Boat assembly schematics quickly evolving into Lego-like instructions.

I scripted a sequence of assembly instructions, and pretty soon they reminded me of the Lego instructions I spent so much time staring at as a kid. You don’t often think about how much time someone must have spent contemplating the best way to show you how to build something… the right view perspective, how much to show in one diagram without overwhelming, or the assembly order that is easiest to follow, not necessarily time optimal. It’s an art form I’ll never again take for granted, no matter how annoying the IKEA furniture. Good schematics are hard.


Once I had fixed all the little issues hanging over from my previous push, was back up to speed and had some basic assembly schematics, it was time to decide what to do next.

Modeling a boat as a bunch of interconnected, force-propagating nodes (right). On the left, you can see how the force of winds will spread throughout the sail. Eventually, a lot of that force gets transferred down the mast to the boat structural elements.

One thing I still needed to do was FEA. Finite Element Analysis is a pretty broad technical discipline, but in engineering contexts it usually refers to calculating load distributions throughout some structure using small (finite element), numerical, iterative approximations. Much as I had started with “basic principles” for stability before finally confirming with proper analysis, it’s now time to confirm the structural integrity of the design I’ve more or less been eyeballing so far. Some people call that “Architecture”. Engineers call it “madness”.

We do this by approximating solid objects as large mesh networks of interconnected “nodes”, and basically treating the force interaction between nodes as a big spring system. That sounds a bit silly, but the force propagation characteristics of most things can be approximated very accurately this way. If you try ripping cloth apart, the threads holding it together will act just like a grid of springs with particular stiffnesses; they’ll stretch a bit as they distribute load amongst each other, then they’ll deform, and finally they’ll snap. How much force it takes just depends on the stiffness of the springs, and where the forces are applied.

All the action occurs where the mast and the keel meet the hull structure. The mast will transfer energy from the sails into the boat frame, while the keel transfers energy from the water into the frame. Whether or not the boat splinters into a million pieces depends on the magnitude of those red sections, and the properties of the wood + fiber glass.

My boat will behave the same way. If I model the wood parts as a nodal mesh, and then describe to an FEA solver how they mate, how stiff they are, and where loads from the sail will traverse down the mast into the frame, I can understand how well my boat will resist deforming and (hopefully) avoid structural failure.

Once you get good at this, and have the proper software configuration in place, the FEA process gets pretty turn-key. I’m starting to look into the best way to automate this in python, since I’m trying to keep this design and analysis completely human-free. Most FEA is still done with a lot of manually software interfacing, at least to set up, but once you’ve done it once, scripting it to run a million different ways is easy. Hopefully I can dedicate an entire blog post to this in the future when it’s complete.


In addition to starting FEA investigation, I had been toying with the idea of scale model v3 for a few months. Honestly, I just didn’t want to sink the time into one again. With every model, there comes a few hours of computer prep work, adjusting the design from its real proportions to the nearest plywood dimensions available at a given scale. Then there’s a few hours of laser cutting or CNC milling. Hours of sanding, assembling (super gluing my fingers together), and some simple finishing. All told, it is probably 10–15 hours of valuable free time sunk, after which I inevitably feel frustrated with imperfections of the model and design.

I fielded opinions: One coworker advised me to just skip the model this time. His logic was pretty fair: You’ve come a long way with the design, and anything you may need to fix now, you can address on the fly with the real boat. Just cut a few parts at a time, and adjust things as you go. We call this the “F**k it, we’ll do it live!” approach.

Weighing the benefits of more scale models

Another coworker told me to stop being such a whiner, and learn to love the process more. Do the damn model. People who get really good at this sort of thing love every step of process. Me? I hate assembling small models of things. Always have. Surgeon hands have not I, nor a craftsman's attention to detail. I will need to work on the latter for this to have any chance of success.

Yet still, I was definitely leaning more towards the first perspective, when a fairly impassable obstacle came along to derail my “Do it live” plans…

No one wants to rent a work space to a guy building his own hobby boat

I’d been looking for places to do the assembly and fiber glassing since shortly after my last blog post. I had decided that taking up friends’ offers for garage or yard space was a great way to jeopardize friendships. Instead, to date I have reached out to over a dozen local boat yards, wood shops, and warehouse managers. Their responses can be summarized by the following list of “reasons [real reasons]”:

  1. Sorry, our boat yard is currently full [Why would I rent yard space to someone not spending thousands of dollars more on my services?]
  2. Our warehouse doesn’t typically engage with personal projects [Oh, so you’re not a well-funded San Francisco startup?]
  3. Sorry, our arts/crafts shop doesn’t have the space for 20-foot projects [We’re actually being honest with you]

A third coworker summed it up rather depressingly well for this urban studio-dweller, showing me an email one of his sailing contacts forwarded, after he had reached out for space recommendations on my behalf:

“No one does this sort of thing who doesn’t have their own property to work within.” — A Realist


Suffice to say, rental space difficulties had given me the arm-twisting I needed: If “doing it live” wasn’t going to be an option until I sorted out a place to build, I might as well iterate on the scale model one more time. Plus, there must be things lurking beneath the surface of my evolved design, possibly serious flaws, that only building it would reveal.

10–15 hours of painfully difficult assembly later… boy, was I right.

1/16th” hull planks (right), 1/8th” general components (top left), and 1/4" structural elements (bottom left). Nowadays I print part numbers on each because I can’t remember what’s what. I generate a PDF list of parts/number, and diagrams to put them all together.

Breaking each of the larger parts into sub-pieces millable in an 8'x4' bed had resulted in two major complications: First, the sheer number of parts was getting to be more than I could realistically keep track of in my head. Depending on the exact design, my software now outputs 60–100 unique parts for assembly. Whereas before I could look at any part in the mill and think “Ah yes, the bottom-most port hull plank”, I now found myself thinking “Where the hell does port_hull_15 go??”

This may not seem like a big deal, but when you are trying to laser cut 75 pieces, across three sheets of varied-thickness plywood, in two hours of TechShop reservation time, it’s unnerving. You want to be sure you have cut the right piece to the right thicknesses, and you don’t want to spend all night doing so. Also, things stop passing “sniff test” inspections that had caught mistakes before. If you’re not even certain what piece you’re looking at, how could you know it looked obviously wrong?

I was going to have to stop thinking about this project as something I could just show up to the shop and manufacture all in one night. From here on out, I’ll focus on more manageable chunks each trip, and get used to manufacturing cycles being several days.

I was also going to have to double my efforts on part labeling and schematics. They were the only thing that saved me from losing my mind this round, but they were full of errors, and a lot of “unpleasant user experience”, as we say in the commercial software world.

The other interesting complication that revealed itself this iteration was the introduction of compounding errors from assembling sub-parts.

Trying to place one piece perfectly (A, where blue is ideal and grey is actual placement), versus the larger cumulative inaccuracy of trying to place three pieces (B)

Every part is imperfectly cut to some degree, and every part is imperfectly assembled. When my boat consisted of large, continuous parts, the manufacturing and assembly errors were minimized. But by breaking every part into smaller pieces I could actually cut at full scale, I introduced sources of compounding error for assembly. Cut imperfections, combined with misalignment in one sub-piece, exacerbates errors and misalignment in the next, and so on.

For a scale model, these imperfections are magnified: The work is done with comparative haste, less precise tools, and the small size of each part makes it more difficult to place precisely than the real part. Whereas two millimeter cutting imperfection and two millimeter misalignment on a plank four feet long is negligible, at ten centimeter scale, it’s disastrous.

Two perspectives of the previous model v2 (smaller) next to this model, v3 (larger). In addition to the design increasing in actual length by ~20%, the scale model was almost twice that of the last iteration.

This iteration of the scale model was a lot more frustrating, and would have been a complete disaster if my girlfriend hadn’t volunteered to provide some extra hands to help place and glue the parts. The model is still too small to warrant more realistic drilling/nailing, but that’s an important fact remember as well. The model introduces complexities and difficulties that won’t be present at full scale. While full scale will have it’s own host of bigger issues, at least it will be easier to line up parts accurately.

So the big lessons were (A) precise cutting and well-measured assembly is going to be even more important than previously realized, but also (B) don’t lose your mind over the imperfections of a scale model. And despite all my griping, the final model actually looks like quite a nice improvement from the previous one. There were a bunch of other good lessons learned from this iteration:

  1. The grain direction of the wood is not currently considered when I lay out parts for cutting. It makes for an ugly aesthetic with wood when adjoined parts have perpendicular grain direction. I’ll need to add an additional constraint to the layout algorithm to ensure parts are cut with wood grain all the same direction. Tom noticed this immediately when I showed him the model.
  2. I added additional spacing between the floor planks to accommodate the keel housing. While it looked good in computer modeling, the real model shows this is a dangerous hole for someones foot to get stuck. Don’t want to break any ankles, will need to tighten that up.
  3. In general, the parts I added recently are particularly vulnerable to the misalignment problem discussed above. While this should be lessened in real construction, I should still go back to see what I can do to simplify these parts and reduce their dependence on optimal assembly.
  4. I’m also feeling less certain about the structural stability of the seating, as well as the bow covering and hull in general. While I can probably fiberglass or MacGyver my way out of this, I again want to consider smarter design options. This is where FEA will be very informative.

For the time being (while I hopefully sort out this warehouse problem), I have plenty to iterate on. There’s another notebook page full of “To-do’s” I’ll spare you.

Screw it, I’ll settle for an open field.

One idea I’m contemplating is to rethink the assembly a bit, to minimize the amount of time I’ll need a large space. It’s actually already pretty close to something I could assemble as three separate sub-sections first. Then I could do most of the work in a much smaller space. Once the bow, stern, and midsection are assembled, glassed, and safely stored away, I could find somewhere for final assembly and glassing, needing considerably less time.

And while all this is going on, I’m also starting to sail more regularly. I figured I should get ASA certified before this baby meets the water, and the refresher courses have thus far been a good source of practical inspiration.

Plus, feeling a 23 foot sailboat creak and groan as it accelerates under high wind load has been a great reminder to go back and double, and then triple check my hull force distribution calculations…

A slightly larger, and significantly more battle-tested boat design on the SF bay. And some doofus in a red hat getting sun burned.

)

Chris Sullivan

Written by

Aerospace engineer, history nerd, basketball aficionado.

Welcome to a place where words matter. On Medium, smart voices and original ideas take center stage - with no ads in sight. Watch
Follow all the topics you care about, and we’ll deliver the best stories for you to your homepage and inbox. Explore
Get unlimited access to the best stories on Medium — and support writers while you’re at it. Just $5/month. Upgrade