During the last two months, I have been travelling for three weeks and studying for my exams the rest of the time, so I have not made much progress on the project. Still, I have been able to complete the turning operations on the tailcone and milled the rest of the brackets that will hold the shroud in place around the propeller. Since only some minor machining operations remain on the tailcone, I have also been able to test fit the magnetic coupling and discovered a flaw in my design. More on that later.
Last update left off with me having started to hog out material from the aluminium chunk that will become the final part. A lot of material needed to be removed before the accurate work could begin. Once most of the unwanted stock had been removed, I turned the mating surface for the motor. This is a rather tight fit since the motor will need to be placed concentric with respect to the tailcone and propeller. Also, the distance from the motor mating surface to the surface where the coupling will ride was critical. This was particularly difficult because the boring bar had to reach far out, making chatter under certain conditions. I made a minor mistake when setting the DRO on the lathe for one of the operations, so the inner part of the bore is a bit oversize. Fortunately this is unproblematic, but irritating nonetheless. The rest of the part turned out good, with the exeption of some chatter marks in the chamfer where the inner bore meets the face where the coupling will ride.
Here the motor is fitted to the tailcone. It slips in without trouble, but is quite cumbersome to remove.
Once the inside had been turned, it was time to make the outside taper. This was very tedious work since the feeding had to be done by hand using the compound. Fortunately the beefy boring bar allowed me to take heavy cuts.
Here you can see that the recess for the propeller coupling has been cut. In order to make this cut, I had to grind a trepanning tool with a large tip radius and a lot of side relief. Since my experiece in grinding custom toolbits is very limited, I'm very happy with how this cut turned out. Once this was done, the final cuts on the taper were performed and then the turning operations on the tailcone were completed.
Finally the propeller could be test fitted to the tailcone, representing a major milestone in the project. This is when I discovered the flaw in my design. It turns out that the strong magnets in the coupling exhibit a lot of resistance on the coupling when spun. This is due to eddy currents being induced into the barrier. The induced currents generate their own magnetic field which counteracts the field from the magnets, resulting in loss of efficiency. The intensity of the eddy currents depend on the rotational speed of the coupling, barrier thickness, conductivity of the material and magnetic field intensity. Due to the high electrical conductivity of aluminium, this makes aluminium less suited than for example stainless steel in a magnetic coupling barrier.
After discovering this problem, I did some research on the phenomena. There are some literature on how eddy current losses can be calculated in magnetic couplings, but it is generally not straightforward to calculate. This article [link] present an estimation of average eddy current losses in axial permanent magnet couplings, however it assumes that the coupling is made from tile magnets instead of an array of rectangular magnets which I am using. This expression concludes that the coupling will have an average loss of 300W! Although this is probably conservative since I have less magnetic material in my coupling, it is still way too much.
Another way of estimating the eddy current losses is to model the coupling as an eddy current brake. Eddy current brakes are used in high speed trains and rollercoasters to enable braking without the use of brake pads, relieving the brake from any wearing parts. I found a good example [link] on how to model eddy current brakes in FEMM (Finite Element Method Magnetics), the free program I used to calculate the maximum torque for the coupling. This example assumes that the brake is linear, so it is not an exact representation of my problem. By using the mean relative velocity between the magnets and barrier, I obtained a loss of around 80 W at 1000 rpm. When inputting the maximum value of relative velocity (at the periphery of the coupling), the loss is around 200 W at 1000 rpm. I assume that the real loss is somewhere between these two values, and I have concluded that I will not do any modifications before the tail section is complete and I can measure the real loss. If the loss turns out to be unacceptably high, I can make the barrier thinner. The relationship between eddy current loss and barrier thickness is linear, so by reducing the barrier thickness from 3 mm to 2 mm, I can reduce the losses by 33% at the cost of lower depth rating.
This is how the loss change with rotational speed of the coupling. Note that the rotational speed is in radians per second, so the rpm is roughly ten-fold to the angular velocity in the graphs.
Another magnetics problem I have encountered is that the two magnets that will actuate the hall-effect sensor in the throttle are too weak. This results in that the DPV is unable to turn off via the throttle, as well as being unable to go full speed. Of course this is not acceptable, so a solution was needed. One of my diving buddies who is also a cybernetics engineering student suggested that we could make a circuit that alters the output from the hall-effect sensor so that the controller will output the full range of throttle settings. This solution require some work in developing a useable circuit, but is easy to adjust later. Another solution was to simply cut a recess for the sensor so that it will rest closer to the magnets. This weakens the plastic barrier, so it was rejected. The final idea was to place a thin steel plate behind the sensor in order to increase the magnetic flux that passes through it. Since magnetic flux travels easier in magnetic material such as steel, this could work. I made a quick simulation in FEMM to verify the idea, and it looks promising. I haven't tested it yet, but as you can see from the graph below, the flux through the sensor is significantly increased. The plot shows the flux density relative to the position of the magnets with respect to the sensor.
Above is the 2D density plot of the magnets with the steel backing plate.
Last, here is a picture of the milled brackets that will hold the shroud. Some grinding, deburring and drilling holes still have to be done, but I'm very happy with how they turned out!