Hi Guys,

I'm uploading my second version of my spreadsheet. Much more detail in here and still shows potential for OU, so I'm looking for pointers to check where my lack of understanding (or clouded vision!?), is leading me to this conclusion. This is still a 0-layer zed, still without continued pumping of water to sustain the pod's buoyant force. I'm well under way in adding layers and hopefully, once Se3d releases his sim, I can check if my answers add up.

In summary, setup as is, with the theoretical no losses:
  In - pump water = 103.6 J
  Out - Buoyant force pod up - 71.1 J
  Out - drain to refill source - 32.5 J
  Out - Pod sinks - 56.9 J

  Cycle complete = 56.9 J excess

-or-

  In - pump water = 103.6 J
  Out - PE difference in pod - 56.9 J
  Out - drain to refill source - 32.5 J
  Out - Pod sinks - 56.9 J

  Cycle complete = 42.7 J excess

-or-

  In - pump water = 103.6 J
  Out - PE difference in pod - 56.9 J
  Out - Pod sinks - 56.9 J

  Cycle complete = 0.8 J excess (rounding errors?). This value though can be increased by increasing the weight of the pod as although the stroke won't be as far, the work required to fill is reduced as more starting (pre-charge?), water can be in the pod tank.

Tell me where I've gone wrong!?

Amo

Quote from: see3d on October 02, 2012, 12:08:07 AM
Hi Everyone,
This is just a note on the sim progress.  I took a different approach last weekend.  I decided to simplify the math and do more iterations, even though it is a lot slower calculation (think very slow animations, like watching paint dry).  However, I also took an approach that let me calculate any number of layers at the same time (currently limited to 6 for no good reason).  I could not let the builders get too far ahead of me
I still have more work looking at all the edge conditions and testing different cases (and seeing if there is any way to speed it up).  However, I am not stuck inside the complicated math anymore, but moving forward with debugging operational code.   
[size=78%]~Dennis[/size]

Hi see3d,
thank you for update. I'm really interested in your COP numbers >0 layers. I do not know whether it helps but in my spreadsheet  10iterations gave me precision +- 10. 10k iterations +-1 and above 100k provides only precision behind decimal point. I'm using excel just for the 10iterations, to get more precision I put together small C++ program that does any amount of iterations to see "exactly" where is the result converging.
respect,
Marcel

Quote from: MT on October 02, 2012, 03:57:31 PM
Hi see3d,
thank you for update. I'm really interested in your COP numbers >0 layers. I do not know whether it helps but in my spreadsheet  10iterations gave me precision +- 10. 10k iterations +-1 and above 100k provides only precision behind decimal point. I'm using excel just for the 10iterations, to get more precision I quickly put together small C++ program that does any amount of iterations to see "exactly" where is the result converging.
respect,
Marcel

I was surprised about the iterations.  The thing that I thought would take many iterations to settle took only 2.  The thing I thought would need 100 took 1000.  I obviously need to understand why if I am going to be able to speed it up.  I am looking for at least 3 significant digits of precision.  I don't want to see a perceptible wiggle in a straight line on my sim transfer function chart. 

The thing that needed few, was iterating between the water and air stackup to balance the air PSI, air volume, and head differentials across multiple layers.

The thing that took a lot, was my independent variable of the Pod Water Head level.  Those are not the only iterations I do though.  I iterate to find the balance point of the lifting force, lift distance, and loaded riser weight.  That means three nested loop iterations.  Some of my iterations use a binary search instead of linear for speed.  In the worst case, I could escape to C# to speed up the nested iterations.  It is harder to debug though, so I want to get it right in the scripting language first.

I am not ready to publicly state sim COP numbers before I have verified the sim to a build. 

The sim is showing that the input pressure to output pressure gain increases with more levels.  The PSI increases towards the Pod.  The head differentials decrease towards the exhaust.  That is what I would expect with the model dimensions I am using to test.  My single layer model algorithm matched in output to a single layer output in my multiple layer model.  They use different mathematical approaches.

Hi guys,


maybe an interesting observation...


Marcel

Quote from: MT on October 02, 2012, 05:31:42 PM
Hi guys,

maybe an interesting observation...

Marcel

Yes, this is how I run my new sim, with initial incompressible air, except that I have to put in very small units of water at a time.  Then I convert the incompressible air to compressible air and recalculate the air volume again for each small addition of water.  Then back to incompressible air for the next pass.  It really simplified the math.