Sunday, April 29, 2018


I keep bees now!


Bridge Construction

We had a bridge going over a small stream in the back yard.  HAD.  Two years ago:

Following some rain and flooding, it washed about 70 feet downstream and lodged against the fence:

Because it was heavy (it's about 16x8, made of treated wood, with some steel and 12x12 timbers on the underside -- i have no idea how it floated), nothing i had could drag it back, certainly no without destroying it.

I paid my kids to disassemble it - they spent probably 12 hours each with crowbars and a sledge hammer.  They stacked the wood and put the steel in the barn.  Awesome.
I used some of that wood to build forms for a new concrete footer on each side -- there were some concrete blocks it had been resting on, but this time wanted more.
As usual the hardest part is measuring and leveling stuff.  At least the stream was low, but schlepping stuff back and forth is hard.
I got a harbor freight cement mixer, since i was going to be doing probably 1500lbs of concrete and mixing by hand (e.g. with a hoe and wheelbarrow) is hard.
One notable design flaw with it is that it dumps directly under itself.  Fine if you can straddle the place you're pouring, but not fine on banks of streams.  So i built it a little chute with a stove pipe and 2x4 scraps:
Schlepping the mixer across the stream was no peach either -- it weighs ~100 lbs -- but it got the job done. 
You can see some of the steel beams (from the previous bridge).  I will reuse those and add one other that we found in the pasture.  I don't want to drill through the steel - that sounds hard, so i will just put in little concrete anchors and attach the wood to the concrete.  The steel will be sort of wedged in between providing support, but not really securing the wood to the concrete.

More steal on, and some wood members bolted to the heaviest steel c-channel.  these will secure the decking.

Decking.  You can see through the gaps that I actually added some more wood spanning the gap.  So in addition to the 6 steel pieces, there are 7 wooden beams that go across.  The steel is plenty to support, the wood is mostly just to hold it all together.
 The boys took turns drilling pilot holes and driving screws.  They were a huge help.

 The ramps on either end are pretty simple.  Bolted to the side of the bridge, then i set their feet in concrete (not shown) then put decking over them.

The underside, showing the bolts holding 2x6s to the steel c-channel.

 Load testing the bridge - seems to work.

Saturday, March 31, 2018

Trailer Couple

Somehow we banged up the trailer couple so it wouldn't latch nicely.
Unfortunately, it was welded rather than bolted to the trailer tongue.  So I cut it off with a cut-off wheel on my angle grinder (being careful to remove more material from the couple than the tongue).
It was rusty underneath, so i cleaned it up with a wire wheel 
 and painted it.
The hardest part was drilling 0.5" holes for bolts.  Somehow a cheap bit from a Harbor Freight set of 30 bits worked better than a $17 bit specifically for hard steel.  I think the whole HF set cost less than that one bit.  I kept it pretty oily while drilling for both.

I bolted it on with grade 8 (good stuff) bolts and lock nuts.

Now we can trailer again!

Wednesday, March 28, 2018

school work

I'll have a few manly project posts soon (working on a bridge and repairing a trailer), but much of my time is spent on school lately.  I can't post the code (since it could be used by other students and would constitute an honor code violation), but I can show some of the output.

A brief summary of the output of our AI assignments (for GaTech's OMSCS CS6601):

Assignment 5

On k-means clustering and expectation-maximization.  Required us to do k-means clustering implementation and use it to find ideal "average colors" and the groups of pixels that were closest to those averages.  It is an iterative routine which progressively changes which colors it chooses, and (consequently) which pixels are nearest in color to them.

Starting from this image:

This is what it looks like with 2 means (2 colors, and the pixels closest to them), then 3 means, 4, 5, and 6.

There were some other much trickier pieces, but they were much less visually pleasing.

Assignment 4

Is hard to visualize.  It was about automagically building random forests of decision trees to label samples of data.  We were given large amounts of data to build the model, and our models were tested against other unknown (but comparable) data.  Fitting well without overfitting was key.

Assignment 3

Was about modeling bayes nets and monte carlo markov chains.  It was very hard.

Assignment 2

Was about search - we implemented Uniform Cost Search (Dijkstra), A*, bidirectional, tridirectional and some other combinations.  In so doing we found good paths through a network like this:

Assignment 1

We wrote a game player using minimax with some enhancements to play a game that was a combination of

Friday, March 9, 2018

The way it goes

I wanted to install a ceiling fan where there was only a ceiling light.  No problem!  I've done this many times before - maybe 30-45 minutes is plenty.


As so often happens in projects, this one snowballed and wound up taking over 3 hours.  Follow along with these easy-to-repeat steps and see if you can tell where the problem lies:

Step 1 - Remove the old light fixture

  1. finger screws for the glass bowl -- easy peasy.
  2. a screw or two holding the light housing to the junction box -- boy, this box is pretty wiggly.
  3. the wirenuts and ground wire connecting the light to the supply -- hot wire is off; all is well.
Pro tip: It's a good idea to turn off the breaker, but chances are good your breakers are labeled incomprehensibly or are outright lies.  What I do is turn off the room switch and then quickly tap the hot wire with a finger to see if i get zapped.  I didn't this time!

Surprise! A pleasant surprise!  The power supply run to the light is actually 4 wire romex ("12-3"), which is convenient for fans (one hot for light, one hot for fan, one neutral, one ground).  There was only one switch in a one-gang box, but hey - somebody was thinking ahead.

Surprise! An unpleasant surprise!  While fiddling with the wiggly box, I got zapped by the white wire.  That's bad - the white wire shouldn't be hot... not with the switch off, not with the black and red being unpowered.  Investigation required.

In any case, to hang a fan I'll need to reinforce the box - I can inspect the wiring when i get up in the attic.  The box was using one of those metal hangers that span between the two joists, but it seemed to be quite loose.

Step 2 - Reinforce the ceiling box

Up in attic (getting to the area above the box was a challenge to begin with!) I found that the hanger hardware was poorly installed.  Each side had metal brackets with three holes for nails, but each side was only held on with loose staples.  I pulled them out and put in screws (easier than nails in this case).  That cut down a lot of the wiggling, but it still twisted more than i'd prefer, and i didn't want the fan rocking, so...

I cut a 2x4 to the width of the cavity and screwed it in - some of the angles were pretty hard to get to, but i did it... slightly crookedly.  So i reattached the junction box with some shims to make up the difference.  Perfect!

Step 3 - Run new wires between the ceiling box and the wall switch

Luckily i had an extra length of romex laying around.

Step 4 - Replace the wall switch with a 2-gang box and wire new switches

Luckily I had some extra old work junction boxes and switches laying around.  I'm not sure what was causing the white wire to be hot, but when i was done all was well.  Something must have been shorted in the switch box before i pulled it all out and rewired.

Step 5 - Install Ceiling Fan

Connect the new wires, attach the new base to the ceiling box, connect the blades.  Test (all went fine).


Aspiring do-it-yourselfers beware: this is a disturbingly common occurrence.  Not always ceiling fans, but houses are generally built by drug-crazed vandals.

To answer the question at the beginning: "see if you can tell where the problem lies" the problem was the existence of steps 2,3, and 4.  Ideally you'd only need 1 and 5 (but only 1 of the 3 ceiling fans i did in this batch was actually like that).

Saturday, December 30, 2017


 Ronny and Matt from Church invited some youth to try simple forging.  Ronny brought an anvil and some know-how.  Matt brought a little propane forge.
Starting from some simple rod stock, a couple guys each got to pound on the red hot metal:
to make a stylish little letter-opener:
 I got to hammer too, and made a wee "garlic cleaver":
I didn't do much of the metal - i just supplied some dumb hammering while Ronny moved the work - but I was proud of the handle:

I cut off 5" of an oak branch that had fallen during the recent storm, shaped it with a rasp, then burned it with a blowtorch to get the color.  I sealed it with a quick clear coat spray.  The handle only took about 90 minutes (mostly the rasp), and I was pleased with the comfort and color.

I mounted it on the blade by drilling a hole in the wood (slightly undersized for the tang), heating up the metal to really hot, and shoving them together.  The hot metal burned its shape into the wood and it's holding quite well.

I would definitely do the burned wood again.

Sunday, December 24, 2017

stained glass triakis icosahedron

It was hand-made from cut glass (score and snap), bound in copper foil (rather than lead came), and soldered with 60/40 (tin/lead) solder.The lighting is provided by strip of LED lights - it consumes little power, and could be supplied by a wall adapter or computer’s USB port.I find the shape to be both aesthetically and mathematically pleasing.

On the Making

First make a a lot of uniform isosceles triangles (at least 60, probably more if any turn out badly).  I made mine about 2.5 inches tall and 1.5 inches wide (a very poor-man's golden ratio).
  1. Bind all their edges in copper foil tape, solder the borders
  2. Solder them all together so internal dihedral angles are all exactly 138.189685°
  3. I made a little jig out of coat hangers and hot glue to help get the angles right.
  4. Before closing it up, shove an LED light strip in there.

coat hanger jig calibrated to exactly 138.189685°

On the Triakis Icosahedron

The shape can be seen as a platonic icosahedron with triangular pyramids augmented to each face.


An icosahedron is a polyhedron with 20 faces, 30 edges and 12 vertices. There are many kinds of icosahedra, with some being more symmetrical than others.The best known (and the one this star builds from) is the Platonic, convex regular icosahedron, which has five equilateral triangular faces meeting at each vertex.


Adding a triangular pyramid to each face of a polyhedron, making a particular Kleetope.


In three-dimensional space, a Platonic solid is a regular, convex polyhedron. It is constructed by congruent (identical in shape and size) regular (all angles equal and all sides equal) polygonal faces with the same number of faces meeting at each vertex. Five solids meet those criteria:
Four faces
Six faces
Eight faces
Twelve faces
Twenty faces


Video of lights in action