Hacker News new | past | comments | ask | show | jobs | submit login
I defeated a long-broken fridge and became a household hero through 3D printing (arstechnica.com)
120 points by shawndumas on May 29, 2016 | hide | past | web | favorite | 57 comments

The part that I found so relatable here is that 3D printers in libraries are absolutely transformative. It empowers people to do things like fix things on their own but also crazier things like launch a company. A couple of days ago I wrote a blog entry about how we started our company after doing some 3D printing at the library in downtown Chicago.


I've been talking to the head of my local library about this - I think we'll see one installed here within a year or so. (I live in a tiny town, so the main issue is budget.)

thats awesome. If our company manages have some extra cash somewhat soon I am hoping to buy a printer or two for the library we used (harold washington, the main one downtown). I believe they had some sort of rental deal or something. Would love to get them one of the resin based ones.

This looks great - I hope your products stay hacker friendly as you grow.

Definitely. Its amazing how much people (particularly investors) worry about consumer vs maker from a perception perspective. Hoping we will be able to maintain consumer simplicity and elegance but maker hackability and general inventiveness.

This probably all boils down to my lack of experience, but I just recently got a 3d printer, and most of the things I have printed have failed to perform their desired function. This is both things I designed (a replacement battery lid for a remote - the software skips the small details like the notch that holds it in place), and most of the designs I've pulled off of thingiverse (e.g. a water bottle lid - it's too tight and it leaks water).

My only fully successful prints aside from the calibration were the "hello world" roctopus and a toy car.

I expect (hope) that this will improve with time, but so far it's been a little disappointing.

This is why it is still "early days" as far as I am concerned with 3D printing. But it is also about learning and understanding your tools.

Prints that are insufficiently strong may be because the material is not strong, or the way in which it was printed did not favor strength, or the "slicer" converted it into layers in such a way that it was not strong. Sizing is another issue, my Makerbot Replicator does not print dimensionally accurate pieces, what it worse is the X direction is more accurate than the Y direction, the Z direction is spot on. So after printing a number of test pieces where the size it "should" be was known, and the size it was, was measurable. I came up with some "calibration adjustments" that I can put into my slicer program to improve accuracy (Simplfy3D).

There is also an issue of materials, very inexpensive printers might print only with PLA which is not nearly as strong as ABS. In the middle is a PET variation I got from MadeSolid. PLA, ABS, and PET are all shorthand for the chemical formula for the material. Different materials, different properties, these days you can also print in Nylon and other high density plastics. Somewhere I have a picture of a "hula girl" that was printed in PLA and put on a dashboard of a car, she is all melted due to the heat on the dashboard.

So it isn't as simple as click, print, and use. And most printers seem to be a bit cantankerous and picky, but if you take a bit of time to methodically learn what they can and cannot do, they can do some pretty amazing things.

For engineers with new printers I really recommend printing out test pieces to understand your printer. You can test strength (tensile, shear, compression), detail reproduction, dimensional accuracy, and environmental durability. There are a lot of test printable things on the Thingaverse site. Once you know your printer you'll be able to quickly design and print a part that will meet your expectations the first time, every time.

If you want “production quality” parts, I recommend you use a laser cutter, some kind of milling machine, or regular woodworking / machine tools. If you want to make a bunch of the same kind of plastic part, you can mill a mold and cast them, cf. http://lcamtuf.coredump.cx/gcnc/

3D printing is great for making objects whose shapes make them (a) impossible to construct with a subtractive process, and (b) too complicated or subtle for your skills with clay or whatever. For example, various types of mathematical objects, https://plus.maths.org/content/3d-printing http://www.shapeways.com/marketplace/art/mathematical-art/

Keep at it! Part of the problem is designing, part of the problem is printing. If there's a reprap group or a makerspace nearby where there are people with experience, that would help you a lot.

I've tried to keep my designs as simple as possible and it's worked out alright. As you say, more complex things are fairly hard to model. Having calipers so that I could accurately measure things greatly helped me design things for around the house.

I'm probably classified as intermediate in 3d design and printing. It takes a while to get it right, there are so many factors that influence the outcome, so initially it might seem like the failures are random.

Thingiverse models frequently have issues, but you get used to opening them up, checking dimensions and fixing errors in the STLs.

When designing, at some point you get used to the fact that it takes two or three iterations to get it right. I frequently print partial models nowadays (or smaller versions of the final), featuring whatever needs to fit and then going back into the design software and completing the model.

Then there's the whole "what's the printer up to" thing, is the PLA not too old, is my printer properly calibrated, did the slicer do its job properly. That just takes time to understand as well (having a quality printer here helps as well).

Just keep going at it and eventually it all starts falling into place and you get faster and faster at designing as well as incorporating others STLs into your own designs and printing them.

I believe you might be missing the key ingredient of prototyping: iterations.

From all the success stories in the industry, 3D printers combined with competent designers enables very tight iteration cycles, and makes the interations more fulfilling to test :)

What 3D printer do you have? They are not all equal in capability and resolution.

For something like a chunky toy car, precision is not important. For a threaded screw, in particular a bottle cap, you will need good resolution and accuracy.

It's a Lulzbot Taz 5 (they announced the Taz 6 the day after it shipped :(

From what I've read, it's a decent quality printer. I've kept the software in beginner mode so far, so there's probably some settings I'm overlooking.

Looks very capable. Not cheap at all. I haven't heard of it, but then again once I bought my "Up!" printer I stopped looking at what was out there.

My father needed to change the oil in his Toyota truck about 6 months ago, but unfortunately he couldn't find his oil filter wrench. I too thought I could be a hero, and show him that the money I spent on this 3d printer wasn't just for a lark, but it was a useful tool. I found a model of the exact tool needed on Thingiverse, printed the model out (it took about 3 and a half hours if my memory serves me correctly to print the part, all the while my dad was waiting).

Finally, being incredibly proud, I handed my dad the tool, and expected success along with much congratulations from my father for solving his problem.

It was about a millimeter too small to fit around the filter. I had failed. My printer had failed. He went and bought a new oil filter wrench at a car parts store.

One of the problems with STL's is that they're unit independent, they don't care if the part is measured in millimeters or inches. (they also ignore the internal structure of the part, but that's a rant for another time.) And one problem with the 3d printer community is that with all these different printers, there is a wide variety in the tolerances/accuracy of the printed parts. Even with well set-up printers, changing one setting, such as amount of infill, may change the size of the final part. And even if you have a precisely calibrated printer, the person who made the model you're about to print may not.

In this particular case, I don't believe the model was off, but rather to make the tool incredibly strong, I printed with ABS plastic and included extra support and extra shell layers, which I believe may have produced too much material and effectively over extruded plastic, making the part just a bit too small. But I took the part and slammed it into the driveway as hard as I could, and the part didn't even think about breaking, it was quite strong, even if it was useless.

That's why people of the ancient times invented sandpaper.

Surely a flat file would fix that though? If it's just ABS it should file quite easily. The tool is far from useless.

A file is one of the most under-appreciated tools. I find them incredibly useful.

I tried. I worked on it for about 30 minutes and then just gave up. Printed ABS is actually surprisingly difficult to file/sand, it's rather tough stuff. I find it's easier for me to file 6061 Aluminum than it is to file ABS. And as this was a piece with a ~3 inch diameter, and many skinny, flat internal angles, filing wasn't successful.

An STL file is made of floating-point triangles. You get X,Y,Z for each of the 3 corners.

It should be integer tetrahedrons. You'd get X,Y,Z for each of the 4 corners.

Tetrahedrons let you fill the middle. Each tetrahedron should have fill data, for the whole thing or corner-by-corner for gradients. This would allow variable density, elasticity, conductivity, opacity, hardness, etc.

Integer math solves the problem of variable rounding that often screws up STL files.

Being unit-independent is much better than having one particular unit. You may think millimeters are great, but to others this might as well be angstroms or chains or twips or rods or parsecs. Unit conversion introduces rounding error. I could go for an optional free-form text field that suggests a unit to use, with the software defaulting to the suggested unit if recognized. Software should let you define the unit.

I agree with the unit-independent part, and you hit the nail on the head, I would prefer a text field, or anything that's part of the file, to give an indication of what unit of measurement the part was designed.

There's about 15% shrinkage in extrusion printing with ABS. Most of that occurs during printing (which is what causes lamination cracks), and some of it occurs during cooling after printing. It's just not a dimensionally precise process.

I've never seen an oil filter a screwdriver and a hammer couldn't get off...

Not to take something away from what he did and what he learned doing it, but there are plenty of ways to fix that part that would be a good bit quicker.

I'm a big fan of Bondo, fiberglass/fiberglass resin, and my trusty dremel tool. Recently a plastic part to a dryer cracked, and those approaches didn't seem strong enough and would probably make it not fit any more, so I heated up some little nails on the gas stove and pressed them into the plastic, which melted around them, making a very strong part that was no bigger than the original.

And I say this as someone who has a long, long history with 3d CAD modeling.

I have similar experiences. In the case of the OP, it seems that the person found value in being able to simply print a new part whenever needed. With a fiberglass part you need to store the mold somewhere safe. Composites are a great way to build parts (I build lightweight parts for race cars), but they are not as convenient as 3D printed parts.

Not sure what you mean about storing the mold. I just do the fiberglass and resin directly on the broken part to repair things (after cutting away some of the plastic). It's sort of like gluing it, but when you need more strength.

Would he really need to print more parts? I mean, it could break again, but this seems like a one-off fix.

I assumed you were replicating a part by making a mold. :)

3d print and Bondo are a great combination. You can test your idea with the raw part, then dress it up with a skim layer of Bondo and then sand and paint. I've seen prototype assemblies with glorious looking enclosures, and when you flip it over, the inside has that "fabric" look of a 3d printed part.

If it only has to look good, the skim layer can be drywall joint compound, which is super easy to work.

I am, unlike the author, not a mechanical engineer. However, I've been able to teach myself enough about designing and printing in 3d to make household-useful items with my 3d printer. The interesting thing is that my wife and I now take it for granted that I can just print something to fix the problem.

It's not been easy, and I laugh at my earlier attempts, but it's been a fun learning experience. The fun thing is that I've had to re-learn a bit of trigonometry because of my choice in CAD tools (openscad).

I have a repository of the household items I've printed: https://github.com/elliotf/reprap-household-misc

He was a mechanical engineer with some experience with autocad. I have to say this is still something the average person would not be able to do.

I never learned autocad or ME, but I do remember my ME buddies from college taking their required autocad class and taking many hours to design a simple part like a pulley, and then being oh so proud of their creation. But, hey, I am sure it was much more fun than their Fortran class which they also had to take.

My point is it takes special skills, training and a lot of patience to be able to design a part and get all the dimensions right.

It would have been more interesting if he was able to glue the broken part together, then scan it, and then print a copy of the scanned part. Is that something that is easily doable nowadays?

I think the interesting thing is that for small communities (Game Boy revitalisations?), there is a semi-cheap way to build certain plastic replacement parts/mod parts. Sure, the person who made the piece needs to be good, but everyone can profit.

Replacement game boy parts are available on ebay for super cheap.

But people putting a raspberry pi into an old gameboy case don't want gameboy parts, they want modified gameboy parts.

At the moment there will be a youtube video and an imgur gallery of where to apply the dremmel. In future there'll be downloadable 3d printing files.

Yeah game boy might be a bad example. MSX? Or custom Commodore 64 cases (though that could be trickier because of the size)

I was truly convinced of the power of 3D printing a similar way a couple years back.. My 80+yo neighbor lost the back cover of her feature phone. I looked for a replacement but it's an old phone. She told someone at her church, who noted the model number.

The following week, he brought her a 3D printed cover that fit perfectly. Over a month later and she was still talking about it.

When you build a product that appeals to 20yo hackers and 80yo grandmas, you're onto something.

TinkerCAD is really underrated. People who already have good spatial and mechanical skills tend to be CAD snobs and look down on it. But if you’re not by nature a very handy person and just need simple things around the house it will do everything you need. It’s parts tend to have a distinct look. But they work just fine.

TinkerCAD is great in terms of both function and usability. The core problem with it, being both closed and cloudy, is that it can just disappear one day and leave you unable to edit anything you made, or losing it entirely if you forget to export it. Unfortunately, many of the "grown up" CAD packages are the same way, with version-specific DRM formats and cloud activation. They're expensive enough that the customers have some power to demand continued availability and access, but the only way to be safe is either to use non-networked local software or software that is open/free enough that it can be hosted by someone else if the vendor or their post-acquisition owner decides to not give a shit about you anymore.

Modeling needs to get easier, especially for such simple replacement parts tasks.

Any suggestions regarding tools for_capturing_ 3D (or at least 2D) models from photos or videos of existing objects? I imagine if there are multiple reference objects (of known size and shape) visible in the scene together with the object-to-capture, the photogrammetry should be tractable.

A good start would be an app able to produce a 2D CAD file from a prototype shape (e. g. cut from cardboard) photographed on a background of squared paper. Does such a beast perhaps already exist?

There is Autodesk ReMake/ReCap. But I'd say you still need Inventor anyway...

And IMO modeling in Inventor is as easy as it gets. Sure, you need to discover some tricks, but usually you get exact tutorials on YouTube explaining how to achieve what you want

Some people have done similar things using the XBox Kinect, see for example http://makezine.com/projects/from-kinect-to-makerbot/

It's still not easy though!

I used an Xbox kinect as a 3d scanner before, and the resolution of the 3d sensor is wayyyy too coarse-grained for anything really useful, unfortunately.

It works well for human-sized objects, but you can forget it for smaller things (it had trouble scanning my then-reference object: a beer can)

That was a couple of years ago - maybe software has evolved enough to make it useable in the mean time.

3d printers are in a weird transitional state. We are long past the point where every prototype shop has a decent 3d printer, yet we're still (seemingly) years away from the point where every home and business has a 3d printer. A big part of this problem is that there really aren't that many obviously useful things to print.

There are lots of obviously useful things to print, but printing them is expensive and difficult.

I looked into replacing a door shelf for a huge Samsung refrigerator. It's a $2000 refrigerator with $50 parts that constantly break. I could buy a new part every few months, 3D print something suitable, or give up and just replace the refrigerator.

Step one is getting a model. Everything is organically curved. Ugh. This is impossible. Well, putting that problem aside, let's just throw together an approximation so that I can get a price. It's a bucket about 20x30x40 cm, and needs to be about 5mm thick. To survive longer than the original junk, it'll need to be made out of something decent. The first thought is naturally titanium. Woah, expensive! The second thought, reluctantly, is laser-sintered nylon. Woah, expensive! I don't want to just go thinner, but I could carve out some holes or leave ribs. This is getting complicated. Woah, still expensive!

It's a good thing I didn't put the effort into somehow making a model that would perfectly fit the organic shape. My effort would have been wasted as soon as I found out the price.

I bought a new refrigerator. The old one is still here, growing mold, because I can't get it out of my house.

I appreciate the geek aspect to this.

However I do feel he could have achieved the same result a simpler way.

1. Repair the original part in a basic fashion

2. Make a silicon cast of the repaired original part. The original will have been injection molded, so try to use their mold separation lines as a guide for where your separation line should be

3. Spilt out to make a mould

4. Pour in resin to make an very exact replica. The place where the original was broken will be replaced with functional resin

5. De-mold and put in place.

This guy makes his own rubics cube style puzzle via this method and documents it well: https://m.youtube.com/watch?v=i-HXU4cfvdc

How is this simpler? Seems a lot more complex to me.

In his story he sat down at his computer for a while and did what he needed. A couple of trips to town later it was done.

In your story he needs to get a bunch of messy stuff, using tools and processes he has no knowledge of, of which even stage one I don't know where to start.

I don't understand how you think that is simpler!

What you say has merit. My suggestion is not simpler for everyone. There is a free time - experiment space - messiness tolerance elrequirement that you point out.

But if you have free time, have a bit of free space like a shed and can tolerate a bit of mess, then my way could be simpler

For something like this simple plastic non load bearing cap you could just press the broken one in a wad of plasticine then pour molten polystyrene or similar plastic into the mould. A lot cheaper and simpler than 3d printing...

The very first thing I did with my printer is go on a repair binge. Modelling a tiny gear in Blender (having done 0 autocad since college) and fixing a little toy felt like my first "hello world" on a C64. Magical.

I then had a moment of silence for all of the super-glue messes of my childhood that never fixed anything and a touch of envy for my own kids growing up in what to my 10yo self would seem a Star Trek future.

I've got no CAD skills whatsoever - how would I go about creating a replacement for the back of a remote that's cracked?

It's similar to the back of http://www.ebay.com/itm/Panasonic-Remote-Control-EUR644862-C... and I suspect the tiny gap at the top is going to be my downfall.

I fixed 2 remotes with missing battery covers. The secret to getting a good result seemed to be not trying to duplicate exactly what is missing but reimagine the part a little to take advantage of what 3d printing brings.

In both cases, I first modeled as best as I could the parts that I had (not the parts I wanted to print) so that I could see the fit. On one remote, I replaced the entire back, not just the battery cover because I just didn't think my printer could create that thin curved shape. This came out awesome. The second, instead of copying the missing cover exactly, I left much plastic that an injection molder would not (ex between the batteries and set my infill to 90% making a very solid, strong part that snapped in nicely.

Note: Blender has a learning curve like the side of Everest but there are endless excellent video tutorials available on youtube to get you up to speed, so don't despair when you open it the first time and literally can't even navigate the view with your mouse.

Instead of Blender for mechanical parts, try OnShape or Fusion 360.

Duck tape. Sometimes a simpler solution is a better one.

Yep, that's the current solution. Replacing the batteries becomes a pain.

Spoiler! Author is a mechanical engineer.

As the author did state write:

"While I only have experience drawing in 2D, how much harder could it be to add one more D? Fairly hard, it turns out."

Being a mechanical engineer doesn't mean you automatically can draw 3D models (and I should know, I was a mechanical engineer).

How come? I mean you sure can draw isometric views, is that such a huge leap?

Stereotype reinforcement at it's finest. The mechanical engineer designs a part in CAD, 3d prints a prototype and repeats until success is achieved. It's the physical equivalent of the programmer automating a one time task. What a shame no computer simulations were involved.

It would have been easier to go to the dump and spend five minutes looking at the row or two of fridges to find the part or, failing that, find a suitable chunk of plastic, cuts it down and makes the part and spend the rest of the time fixing something else around the house instead of writing an article.

After working with everything from paper to plastic to wood to spring steel at college I'm just gonna kill a few hundred birds with one stone and get a mill. When you consider all the tools it replaces and how many different specialized machines it can stand in for (including a 3d printer) in a pinch if its got a big enough Z it seems like a no-brainer

Applications are open for YC Summer 2019

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact