I've got polar plots out of the Weaverbird data.

They are located here: https://drive.google.com/drive/folders/1d5lF1XT4DlbX9-ie7WZpWFcVDLl9FPbm

I changed the settings. The Weaverbird results should be public now.

I combined the port and starboard data. My strategy has been to first group the data by TWS (4 knts, 6 knts, etc.) then slice those groups into subgroups by TWA. For example, the “4 knt” TWS group would contain sentences with TWS between 3.5 knts and 4.5 knts. The “6 knt” group would contain data rows with TWS between 5.5 knts and 6.5 knts and so on. Eight groups in total [ 4,6,8,10,12,14,16,18-20]. Each TWS group was furthur broken down into TWA subgroups. For example the “40 deg” subgroup within the “6 knt” group would contain sentences with TWA between 35 and 44 degrees. The “50 deg” subgroup would contain TWA between 45 and 54, etc. Fifteen subgroups in all: [40, 50, 60...180]. Each group X subgroup (8 X 15=120 cells) cell contained varying numbers of STW, boat speed values. The median STW was gotton for each cell and smoothing function applied to each set of 15 values within each the 8 TWS groups.

Eight polar plots (wind from the east) for each TWS group showing medians of the per cell data and smoothing function line. Final polar plot has all the smoothing function lines combined. The more data that's collected the better the polar plots will look.

My strategy is very sample size dependent. After slicing and dicing the data into 120 groups, even starting with around 10,000 values some cells are very data poor...< 10 points.

I'm wondering if information is being lost by culling the 5 Hz original data. 10 points at 1 Hz would be 50 points at 5 Hz. If the 5 Hz data is available for download (perfect if in the same csv format) I'd run them.

rself

I am happy to report that I have been able to create two more pages on the website which show the principal dimensions of the boats tested to date, and the polars generated from the raw apparent wind data for those boats. They are located just below the "Sailing performance" page in the "Junk Information" menu. When we have True wind polars I will create a page for them.

You will see that the polars are very "spiky", which shows how hard it is to get flat out performance data for every wind direction and wind speed. I have concluded that it will be necessary to edit and fair the curves, either by hand, or by using some clever software to do it, if we can track down or create a suitable piece of software. 

Any ideas or suggestions?

Anonymous wrote:

Robert, I have now looked at your csv file and it all looks good to me. Can you explain in more detail how you are processing it? It would be good to be able to replicate what you are doing.

My processing is mostly read and write. The only manipulation I do is to remove data rows with missing awa , stw, or aws data. If date or time is missing the row is kept. I've only seen missing stw data the few times I did quickie looks in a spreadsheet prog.The number of missing rows in the 8 file weaverbird data ranged from 0-69....pretty insignificant given row counts were hundreds to thousands.

I can pretty much create output in any format you want. For example, most likely polauto is format specific so I could create a csv just like the originals with the tws and twa columns filled in.

Assuming the file format for all the data you have is identical to the weaverbird format then at this point it's "manufacturing". The download/upload time is longer than an actual run time to create a csv.

Let me know the format and I'll create a weaverbird file for you to test.

rself

Robert, I have now looked at your csv file and it all looks good to me. Can you explain in more detail how you are processing it? It would be good to be able to replicate what you are doing.

The 5ps means the file was read into Polauto at 5 lines per second. This roughly corresponds to one wind speed and direction reading per second. Boat speed is being transmitted less frequently as described on the web page. That's why we need a better multiplexer.

Polauto offers a range of data reading speeds from 1 line (NMEA sentence) per second up to 100 per second. I used 100 to start with, because it is quick, but then I realised it was missing out about 95% of the data. The multiplexer outputs about 5 to 7 lines per second. 

There are typically 90,000+ lines of data per tack per boat, so it takes about 6-8hrs to process and produce the polar for one tack on one boat at 5 lines per second. At 1 per sec it would take 30 to 40 hrs, and it wouldn't improve the results. I started including the 5ps in the title to distinguish these files from ones run at 100 per second which I use as a quick check after a run.

I thought it would be interesting to find a way to make a comparison of the results and polars for different boats which is more meaningful. Obviously the size of the boat affects the speed it will go, and this is related to the square root of waterline length. In order to make some allowance for a boat having short or long overhangs, I used a length equal to 1/3 of the overall length plus 2/3 of the waterline length. As the boats tested so far are scattered around the 25ft length, I used this as the "standard" length, and "adjusted" the boats speeds up and down pro rata, i.e. Correction = 5/ sqrt L.

In the same way, I looked at the sail area to displacement ratio, which also has a big effect on performance, and took a value of 18 as being mid range, and adjusted the boats up or down to that. Correction = 18/(SA/D ratio)

I then combined (multiplied together) these two correction factors to produce a single correction factor. These correction factors are listed in the data sheet for the tested boats which you can see here:

https://db.tt/0VyMHbZGUW

This table of boat data can also be accessed from the Sailing Performance web page. Most of the data was gathered from sailboatdata.com, except for the junk rig sail areas which come from the owners.

We can have a long and fascinating debate about these correction figures, and I now think the SA/D correction is probably too big, but I have decided to focus for the moment on fairing the raw data polars because the filtering process just seems to make the polars spikier, which is the opposite of what I had hoped it would achieve.

In any case, the critical component for comparing windward performance between boat/rig combinations is the angle (range) for optimum VMG, which is not affected by these calculations as they only affect the boat speed.

The new log is said by JGTech to be on it's way, and can be combined with the new float when both are ready. We will then have to arrange a calibration session in still current free water. We will also have to install and set up the new multiplexer (which I have not ordered yet as still researching options). Hopefully it will all be done by next spring so we can start testing again then.