Adding Flywheel Inertia

I will preface this section with my belief that if you like using the clutch, you will prefer more flywheel inertia.  If you ride just using the throttle, then having a lower-inertia flywheel is probably better/safer.

Although I am loath to add weight to an already heavy bike, it really could use more flywheel inertia.

The ePure Race comes with two removable flywheel weights (shown below).  There's no room to add an additional flywheel weight.  The next easiest thing is to replace the OE steel weight with one made from a denser material.  I looked at a lot of different alloys, sources, and methods to gain additional inertia.  Tungsten is the most effective, albeit expensive choice.  Although pure tungsten is prohibitively expensive, a sintered alloy made of 80% tungsten and 20% copper (W80Cu20) is much more affordable (and has even greater tensile strength than mild steel).

The most economical source I found was via AliExpress.  I ended up buying a plate with nominal dimensions of 100 x 100 x 8 mm at a cost of $129.15 (including shipping and tax).  It took less than 3 weeks to arrive.  W80Cu20 has a theoretical density of 15.15 g/cm3 (whereas steel is 7.83).  The chunk I received measured 100.6 x 100.6 x 8.4 and exhibited a density of 15.4 g/cm3.  I was very pleased with that.  

Removable OE flywheel weights

OE Dimensions

Each OE ring weighs 304 grams, has an OD of 103 mm, an ID of 64 mm, and is 8 mm thick.  There is also a locating feature that is 68 mm in diameter and 2 mm tall.  Six 6.1 mm holes are equally spaced on a 79.5 mm bolt circle.

By my calculation, each ring has an approximate MoI of 588 kg*mm². 

 Finished Product

Right now, my tungsten-copper ring weighs 753.6 grams and has slightly over double the MoI of the original steel ring.  I figured this would be noticeable as EM made the original flywheel weight as multiple pieces to allow for tuning (you can assemble it with 0, 1, or 2 rings).  I will have the equivalent of 3 rings in the space of 2.  I made the ID of mine smaller than the original steel rings.  Material far from the perimeter does not increase the MoI much, and I can always open it up to reduce the weight.  It's probably not possible to make the ID much smaller than this as an outboard bearing-boss would get in the way.  (For reference, that bearing is a 6003 2RS, 17 x 35 x 10.)

As you can see, the OD of my ring is not completely round but it is symmetrical.  Starting with a square chunk of material, I was able to make the OD 104 mm over much of the perimeter.  This reduced the minimum clearance between the ring and the inside of the gearcase to 0.5 mm from the original 1mm.

Tungsten-copper flywheel weight with ID locating feature

Left, W80Cu20 Flywheel weight.  Right, OE Flywheel weight

The Machining Process

Machining was easier than I expected.  I had read that machinable tungsten alloys cut much like cast iron.  It certainly does make the same type of chip as cast iron.  It is not at all gummy like machining pure copper.  It is fairly hard, similar to a medium-carbon steel.

I had intended to machine the entire thing on my CNC mill in order to maintain the best possible concentricity.  Unfortunately, the 1980s-era control in my mill started acting up.  I was able to use it like a manual mill for a while.  Luckily, I was able to pilot-drill the 6 bolt holes and make a nice central hole before it stopped working altogether.  (I wanted to finish this item before repairing the mill's control, so I used manual machines for the remainder of the work.) 

The bolt holes had been piloted with a 3/16" cobalt stub bit at 580 rpm after center drilling.  I finished the bolt holes by reaming in the drill press - working my way up to 6mm diameter one fractional reamer size at a time.  The material reams well.  The central hole was made with a 1" HSS bit at 76 rpm.  I did not use any coolant or lubricant and was very conservative with speeds and feeds (like 0.001" per rev).  I cleaned up the drilled hole with a boring head using a C6 brazed-carbide tool.  This central hole was then used to indicate-in the piecework in my 4-jaw lathe chuck to a TIR of 0.001".  I then opened up the ID to 50 mm and machined the locating feature.

For the OD, I was concerned about making a circle out of a square and the interrupted nature of the cut.  So I lopped off the corners using an abrasive cutoff saw.  This worked great!  (I had also considered using a wet diamond tile saw.)  I found that the material was easy to grind and sand.  I used a belt sander to make it as round as possible by hand prior to mounting the workpiece in a 3-jaw lathe chuck.  Using an aluminum foil shim, I was able to achieve a TIR of 0.0015".

For the thickness dimension, the surface was left as received on both sides. 

0.187" pilot hole

W80Cu20 Chips

After mill-work, prior to lathe finishing


Test running on the stand and in the driveway was smooth - both in terms of vibration and power delivery.  Although this made a noticeable improvement, I would not mind trying even more inertia.  Vendors on AliExpress play games with pricing on this material.  It would be quite easy to spend double what I initially paid. 

Finished tungsten-copper flywheel weight mounted in place.

Even More Inertia!

Having been pleased with the results of my foray into increased inertia, I wanted to try even more.   I could have made another 8mm thick tungsten-copper disk, but using flathead retaining screws allowed the disk thickness to be increased to 10mm.  This increases the inertia by a factor of 1.25 versus an 8mm thick disk (10 / 8 = 1.25).  This is the most additional inertia I could achieve without machining a different flywheel cover. 

The adjacent photo shows the 10mm thick W80Cu20 disk (which weighs 849.3 grams and has an ID of 58.25mm) installed in addition to the original 8mm thick W80Cu20 disk.  The 58.25mm ID is about the smallest that would not interfere with the bearing boss casting.  As with my original 8mm thick inertia disk, the OD is 104mm for as much of the surface as possible.

The M6 flathead screws are 25mm long.

Taken together, both tungsten-copper flywheel weights increased the weight of my bike by pretty much exactly one kilogram (!) so I hope it is worthwhile.

Both tungsten-copper flywheel weights installed.

Coastdown Testing

It was suggested that I measure the coastdown time for my various flywheel configurations.  I like having quantitative data as well as riding impressions and happily complied.  The procedure and results are described below.

With the chain and front sprocket removed, I held the throttle wide open until the motor was running at maximum speed (about 136 Hz or 8160 rpm).  This was with the battery at 54.0 volts, in Green mode.  My automatic regen feature was disabled.

I then allowed the throttle to snap shut (self-close); the motor coasted down to zero rpm and the time between these two events was measured.  Time was repeatable to within 0.1 seconds. 

Purple trace: throttle position. Yellow trace: motor position sensor

I used two channels of a deep-memory digital storage oscilloscope to achieve high accuracy and create a permanent record.  Channel 1 (yellow) monitored one sine wave produced by the motor position sensor (which makes 1 cycle per revolution).  Channel 2 (purple) monitored the throttle position.  The adjacent photo is an example of the experimental record.  Suffice it to say the sine wave was extremely under-sampled.  But I only cared about when the motor stopped rotating, not the frequency of rotation.  I had great difficulty getting clean signals.  I actually connected both 'scope probe grounds.  This usually causes more problems than it solves, but the opposite was true in this case.

Coastdown Results

Note that the coastdown times represent the inertia of the entire system: motor rotor, non-removable inertia disk, primary gears, and even the countershaft in addition to the removable inertia disks being tested. 

All experiments were conducted with the gearcase drained with only residual oil on the primary gears.  Since the gearcase is quite tight, I thought oil would affect the coastdown time. Besides eliminating the complication of working with a wet system, I wanted to maintain a consistent windage by eliminating fluid drag (which would not be constant due to the volume taken up by the various flywheel configurations not being constant).

I did a final coast-down test with with gear fluid using my double W80Cu20 configuration and the time decreased to 13.3 seconds (and was repeatable).  The gear fluid was Valvoline full-synthetic AFT for Japanese cars.  So, clearly, there is a significant fluid drag.

 Hole Saw Trick

The adjacent photo shows a trick I learned from a friend when he heard I was having trouble using a hole saw in metal.  Basically, you need to drill a small hole through the material under the saw kerf to let the chips fall out.  This worked really well!  I only drilled a single hole but two or more would have allowed a fast feed rate.  The hole saw worked so well that if I ever make another flywheel weight, I would buy the proper diameter hole saw to rough the OD as well.

Side views compare the standard flywheel weights with my copper-tungsten versions.

I think the substantial increase in inertia accelerates clutch wear and I will be experimenting with the 10mm thick disk alone in 2024.

This assembly weights 3126 grams (1 kg more than stock).