SUBSIM Radio Room Forums - The mathematics of roving searches

SUBSIM Radio Room Forums (https://www.subsim.com/radioroom/index.php)

- Silent Hunter 4: Wolves of the Pacific (https://www.subsim.com/radioroom/forumdisplay.php?f=202)

- - The mathematics of roving searches (https://www.subsim.com/radioroom/showthread.php?t=221505)

1: Yes, traffic is bidirectional.
2: In that test, it was between 8 and 10 knots inclusive. The final test will be from 5 to 25 knots inclusive and will report details based on target speed.
3: I measured the detection radiuses directly in TMO. All measurements where made with the default crew skill in a single mission, clear visibility, wind speed 5 knots. Of the targets I tried (I didn't try all of them) heito maru was the hardest to detect on radar by a small margin (a few tenths of a mile or so) so it is the target I used. All numbers where measured from AOB 0. Values:

Code:

        subSensors[0] = new SubSensor("Radar, SJ-1",      13.5);

        subSensors[1] = new SubSensor("Radar, SJ",         9.3);

        subSensors[2] = new SubSensor("Hydrophone, JP",    8.6);

        subSensors[3] = new SubSensor("Visual, Day",       5.6);

        subSensors[4] = new SubSensor("Hydrophone, WCA",   5.3);

        subSensors[5] = new SubSensor("Visual, Night",     3.575);

        subSensors[6] = new SubSensor("Periscope, Day",    3.3);

        subSensors[7] = new SubSensor("Periscope, Night",  2.05);

I'll put the source up in a couple of days, after I have had time to fine tune and clean it, should you want to look through it.

Quote:

Originally Posted by CaptBones (Post 2338516)

Glad you enjoyed reading it. BTW, if you (or anyone else) knows how to rewrite the "Expanding Square" search to make it work correctly, please let me know...I would use it in both SH4 TMO and OM.

I meant to ask earlier but forgot: What was actually wrong with it? It's been so long since I've used the plot patrol button in any version of the game with any mod that I just have no recollection of how it is. Is it actually a shrinking square?

Quote:
Originally Posted by CaptBones http://www.subsim.com/radioroom/smartdark/viewpost.gif
There are many variables that we don't have in the game, such as multiple sensors with different detection probabilities for different types (sizes) of targets under various weather (signal propagation) conditions.

Quote: A lot of that actually is in the game but it tends to be subtle enough that its effects aren't noticed. The most obvious examples are hydrophones and how rough the sea is and watch crew and light level. Target aspect has an effect, and I have noticed that some targets can be detected further away than others.

Hhhmmm..."apples and oranges" there...I've tinkered with most of those parameters and variables myself in trying to get my crew's behavior and the AI's behavior to more closely match the real world (such as I experienced and remember it). But those aren't the variables I was talking about. What is missing is actual data and data-based algorithms to apply probabilities of detection of different sensors and sensor systems (being employed simultaneously when possible) under various environmental conditions and states of operator proficiency (not to mention countermeasures and counter-countermeasures) to produce the tables graphs and charts that would normally (in pre-computer days) be used to lay out our patrol areas and search patterns. The simplified and simplistic techniques used by the game do not even approach the most basic level of usefulness as compared to real-world detection modeling.

Of course, this is a PC game. It has been made much closer to a real simulation by the amazing work of the modders in this community. But true simulators that apply the aforementioned techniques are multi-million dollar networked computer facilities that occupy thousands of square feet of floor space in multi-story buildings on military and naval bases that I can't even get into anymore without a special escort (sometimes my younger son gets to be that escort; I think he might end up out-ranking me one of these days :arrgh!:).

Quote: Good posts though. Thank you...and BTW, your simulation is producing some very good and very very interesting results. You sure you don't work for ONR in Anacostia?:hmmm:

Yes, exactly. The devs got it exactly backwards; they have you starting on the outside and working your way in. The Expanding Square (which, in real life, can be modified to an Expanding Rectangle if there is information on a more probable target course) is used to search for a target from a known "datum", when the target's course and speed are not known, but one or both are known to be within known limits (i.e., a submarine fired a "steam" torpedo from an observed location and we know it can't go any faster than 8 kts submerged). The starting point is at Datum and the search legs are based on the target's "farthest on circle" depending on the time difference between last target observation/sighting and time of arriving at datum.

Quote:

Originally Posted by CaptBones (Post 2338562)

You sure you don't work for ONR in Anacostia?

Hah, no. I'm just some guy that likes to code for fun once in a while. Not even a major project, just 750 lines or so of code.

Quote:

Originally Posted by ColonelSandersLite (Post 2338558)

Code:

        subSensors[0] = new SubSensor("Radar, SJ-1",      13.5);

        subSensors[1] = new SubSensor("Radar, SJ",         9.3);

        subSensors[2] = new SubSensor("Hydrophone, JP",    8.6);

        subSensors[3] = new SubSensor("Visual, Day",       5.6);

        subSensors[4] = new SubSensor("Hydrophone, WCA",   5.3);

        subSensors[5] = new SubSensor("Visual, Night",     3.575);

        subSensors[6] = new SubSensor("Periscope, Day",    3.3);

        subSensors[7] = new SubSensor("Periscope, Night",  2.05);

I'll put the source up in a couple of days, after I have had time to fine tune and clean it, should you want to look through it.

Are these the detection radii of the various sensors in nautical miles?
Also where are you up to with these simulations?

I am fairly new to the game but I have a search strategy which involves going to where I am ordered, then choosing a small area within that location where a shipping lane or bottle neck might be. Then I arrange a search pattern perpendicular to what I think the traffic path will be. Becasue I am using a radarless S class I travel back and forth along that path diving periodically, ordering all stop, manning the hydrophones myself and continuing on if nothing is found.

The question I was asking myself is how often would I have to dive when traveling at a reasonably economical speed to detect a 12 knot target passing through my least covered area? I chose a 12 knot target since it seemed intuitive that I would find a slower target more easily. Up till now I have been assuming its 20nm but since I am not 100% sure what the hydrophone range is in RFB this has been impossible to improve. If anyone could clear this up I would really appreciate it.

Also I would be more than happy to lend this project a hand in return for that info. I have done numerous simulations in my role as a statistician and, as you might imagine, I also have an excellent grasp of probability.

I go to my patrol area and wait for a ship to come by,sink it and wait for the next one...I average about 30,000 tons per patrol...Anybody want a drawer full of navy crosses???:haha:

Yes, nautical miles.
Honestly, I got distracted by a shiny new thing, and need to come back to this. That shiny new thing led directly to the advanced convoy attack tutorial if I recall correctly.

The question about how often to dive for maximum search is a fairly complicated one and in your case, depends largely on two large unknown variables. Those variables being detection ranges for visual and hydrophones in RFB. Without knowing those, I can't even begin to speculate, except to say that I don't think that WCA should have a detection range of anywhere near 20 miles. I could be wrong on that, but I don't think I am.

The best thing you can do is just find the range for yourself. Here's how I did it:
I used a boat equiped with radar and WCA.
Map contacts on.
Detected target on radar, then turned radar off.
Waited for target to show up on hydrophones.
Turned radar on and marked position on map.
Measured distance on the map.

The same procedure was used for JP, just a different boat configuration.

I would be quite interested in knowing what numbers you get out of RFB.

I should run a test on it, but I suspect that in tmo at least, the answer is really simple. Basically, my suspicion is that you want to use the best sensor you have as much as possible. With radar, this is always going to mean surfaced. With JP and no radar for some reason (damaged radar for example), this would mean staying submerged as much as possible. With WCA, this would boil down to the current visibility level on the surface. I.E., surfaced during the day as long as weather is clear and submerged at night or in otherwise poor visibility.

If minimizing the enemies opportunities to detect you is the priority, you are trading away sensor effectiveness for stealth. Particularly by submerging during the day when sonar isn't your best sensor. In this case, you are reducing your effective search area quite substantially during the day, due to the lower search speed and lower detection radius. This stance would necessitate surfacing at night to recharge, and doing this would significantly reduce your detection radius if you are relying of visual detection.

I just hunt in the shipping lanes. It is easy to find ships with TMO and RSRD.

Quote:

Originally Posted by ColonelSandersLite (Post 2347628)

I should run a test on it, but I suspect that in tmo at least, the answer is really simple. Basically, my suspicion is that you want to use the best sensor you have as much as possible.

Your suspicion would be correct becasue we are talking about circles and its easy to prove that the biggest one is best whether moving or stationary. S = 2rd/A where S = the sweep fraction, r is the detection range, d = the maximum distance of the search, A = the size of the search area.

I found this formula but I don't know how useful it is:
With a random search the probability of any given target being detected is then D = 1 - exp(-pS) where p = the probability of detection of a ship within the detection range r.

I think the answer with radar is simpler than it is with hydrophones. With the radar you can go with continuous sweep and so the only question that remains is how far back and forth do you travel. With the hydrophones you have to choose how far to go back and forth and also when to dive to use them.

About the hydrophone detection; the early war equipment was probably only good to 7,000 or 8,000 yds. In RFB, the detection range is much greater, if you listen yourself. This however is a game bug, you can usually get a big advantage just by using the hydrophones yourself, while the crew performance is modded to be realistic. Don't know how much/when performance improved.

If you are trying to play realistically, your best bet is probably to submerge and go at most economical speed, 3 kn. during the day, and either sit still, or move at 6.5 kn. during the night (if you have enough fuel for that). On my last patrol, I followed the above procedure, only moving about 20 or 30 nm overnight to save fuel. This allowed for a fairly long patrol, while permitting me to cover a good chunk of my zone, bit by bit.

Quote:

Originally Posted by TorpX (Post 2347757)

About the hydrophone detection; the early war equipment was probably only good to 7,000 or 8,000 yds. In RFB, the detection range is much greater, if you listen yourself. This however is a game bug, you can usually get a big advantage just by using the hydrophones yourself, while the crew performance is modded to be realistic. Don't know how much/when performance improved.

If you are trying to play realistically, your best bet is probably to submerge and go at most economical speed, 3 kn. during the day, and either sit still, or move at 6.5 kn. during the night (if you have enough fuel for that). On my last patrol, I followed the above procedure, only moving about 20 or 30 nm overnight to save fuel. This allowed for a fairly long patrol, while permitting me to cover a good chunk of my zone, bit by bit.

I just had a look at the uboat.net article on hydrophones.

"To avoid own noises, a submarine could use underwater sound detector if her speed was up to 6 knots. If a submarine speed was 4 knots, the submarine's underwater sound detector average distance of detecting another object was:

- for a destroyer- 5 to 10 nautical miles,
- for a cargo ship- 3.5 to 7.5 nautical miles,
- for a convoy- up to 50 nautical miles."

Obviously this is for U-boats though and it also doesn't say when. I would like to play as realistically as possible, but I have no idea how far hydrophones can detect ships in game or in real life, just that for the latter it varied a lot with weather conditions.

In my recollection, the 7,000 yd. figure was for detecting a single ship 50%, detection, at war's start, so convoys, and maybe larger/noisier ships would be more easily detected. USN sound gear seemed to behind German equipment, at least at the start.

I can't remember where I saw this, but it was a surprise to me, as I had assumed the range was longer.

Sonar ranges when listen was greater than you think and less than you think depending on sea state and weather.

http://www.hnsa.org/wp-content/uploa...07/11730rb.mp3

http://www.hnsa.org/wp-content/uploa...07/11720ra.mp3 (reference to a 20,000 yard target being heard)

http://www.hnsa.org/wp-content/uploa...07/11722rb.mp3 (12,000 yards)

I love these recordings

Quote:

Originally Posted by merc4ulfate (Post 2347812)

I love these recordings

Thanks for these merc4ulfate, I do too and didn't know these even existed in digital form. 'Roving Searches' assistance from the hnsa, how nice these were saved.

I even learned winter survival techniques (for winter camping) years ago and many other things from old WW II USAAF & USN films.

Happy Hunting!