{ "id": 572867, "title": "Questions on placement of sources and sensors.", "url": "/competitions/waveform-inversion/discussion/572867", "author_name": "doteeee", "post_date": "2025-04-11T19:10:17.858000", "votes": 13, "comment_count": 15, "views": 0, "content": "

After spending some time on understanding waveform inversion, got few questions on the placement of sources and sensors.

\n

We see multiple sources and receivers in the waveform dataset.
\nwhat could be the distance between sources and receivers.?

\n

I donot have good understanding of problem at hand, but i feel the distance between source-source, sensor-sensor and source-senor might play in determining velocity maps

\n

Thanks,

", "messages": [ { "id": 3176830, "postDate": "2025-04-11T19:10:17.857Z", "content": "

After spending some time on understanding waveform inversion, got few questions on the placement of sources and sensors.

\n

We see multiple sources and receivers in the waveform dataset.
\nwhat could be the distance between sources and receivers.?

\n

I donot have good understanding of problem at hand, but i feel the distance between source-source, sensor-sensor and source-senor might play in determining velocity maps

\n

Thanks,

", "rawMarkdown": "After spending some time on understanding waveform inversion, got few questions on the placement of sources and sensors.\n\nWe see multiple sources and receivers in the waveform dataset.\nwhat could be the distance between sources and receivers.?\n\nI donot have good understanding of problem at hand, but i feel the distance between source-source, sensor-sensor and source-senor might play in determining velocity maps\n\nThanks, ", "votes": 12 }, { "id": 3176894, "postDate": "2025-04-11T21:43:48.467Z", "content": "

In the OpenFWI dataset and the test set for this competition, there are 5 sources and 70 receivers all uniformly distributed on the top surface. The source distance is 140m and the receiver distance is 10m. Please refer to OpenFWI paper table 3 for more details: https://proceedings.neurips.cc/paper_files/paper/2022/hash/27d3ef263c7cb8d542c4f9815a49b69b-Abstract-Datasets_and_Benchmarks.html

\n

You are absolutely right that the source-source and receiver-receiver distances play a critical role in velocity map reconstruction. In the OpenFWI paper and this competition, we simplify the problem by considering fixed source/receiver numbers and positions.

", "rawMarkdown": "In the OpenFWI dataset and the test set for this competition, there are 5 sources and 70 receivers all uniformly distributed on the top surface. The source distance is 140m and the receiver distance is 10m. Please refer to OpenFWI paper table 3 for more details: https://proceedings.neurips.cc/paper_files/paper/2022/hash/27d3ef263c7cb8d542c4f9815a49b69b-Abstract-Datasets_and_Benchmarks.html\n\nYou are absolutely right that the source-source and receiver-receiver distances play a critical role in velocity map reconstruction. In the OpenFWI paper and this competition, we simplify the problem by considering fixed source/receiver numbers and positions. ", "votes": 3, "replies": [ { "id": 3179714, "postDate": "2025-04-15T16:05:27.190Z", "content": "

thanks @hanchenwang114 , went through paper. looking forward to build some good solution to the problem, thanks for the competition.

", "rawMarkdown": "thanks @hanchenwang114 , went through paper. looking forward to build some good solution to the problem, thanks for the competition." }, { "id": 3179872, "postDate": "2025-04-15T19:57:43.633Z", "content": "

So is it, in single dimension:
\nreceivers 0-6, 5m, source 0, 5m, receivers 7-20, 5m, source 1, 5m, receivers 21-34, … source 4, 5m, receivers 63-69?

\n

Given that the total length is 700m, but there's 5 not 6 sources, and 70 not 71 receivers? It's not yet fully clear to me how sources and receivers align.

", "rawMarkdown": "So is it, in single dimension:\nreceivers 0-6, 5m, source 0, 5m, receivers 7-20, 5m, source 1, 5m, receivers 21-34, ... source 4, 5m, receivers 63-69?\n\nGiven that the total length is 700m, but there's 5 not 6 sources, and 70 not 71 receivers? It's not yet fully clear to me how sources and receivers align. ", "votes": 1, "replies": [ { "id": 3179884, "postDate": "2025-04-15T20:13:51.340Z", "content": "

The same grid point can serve as both a source and a receiver. There are 70 grid points in the horizontal dimension, and each grid point refers to a 10 meter distance in space. Each grid works as a receiver, and thus, there are 70 receivers (let's say index [0, 69]) with 10 meter interval in between, starting from grid index 0 to 69. There are 5 sources (index [0,4]) with source 0 on grid 0, source 5 on grid 69, and the rest 3 uniformly distributed in between.

", "rawMarkdown": "The same grid point can serve as both a source and a receiver. There are 70 grid points in the horizontal dimension, and each grid point refers to a 10 meter distance in space. Each grid works as a receiver, and thus, there are 70 receivers (let's say index [0, 69]) with 10 meter interval in between, starting from grid index 0 to 69. There are 5 sources (index [0,4]) with source 0 on grid 0, source 5 on grid 69, and the rest 3 uniformly distributed in between. ", "votes": 1 }, { "id": 3179938, "postDate": "2025-04-16T00:01:10.597Z", "content": "

Blame a 78 year old brain - but I cannot get your numbers to match a sketch I am drawing. A 700 meter line having 5 objects placed uniformly has the sources 175 meters apart. 6 sources placed uniformly would be 140 meters apart. Here is the ChatGPT that supports my drawing.

\n

*If you want to place the objects including the endpoints of the 700-meter line uniformly, the spacing between them is calculated as:
\n\"\"

\n

The coordinates (in meters) will then be:

\n
m\n\nm\n\nm\n\nm\n\nm\n
\n

These are the positions of the 5 objects along the line from 0 to 700 meters.

\n

If you include the endpoints of the 700-meter line, placing 6 objects uniformly means dividing the line into 5 equal segments. The spacing between objects is:

\n

\"\"
\nThus, the coordinates of the 6 objects along the line are:

\n
m\n\nm\n\nm\n\nm\n\nm\n\nm*\n
\n

Based on examples from host git and some of the images it's a reasonable conclusion that it's 5 sources and they are located 175 meters apart.

", "rawMarkdown": "Blame a 78 year old brain - but I cannot get your numbers to match a sketch I am drawing. A 700 meter line having 5 objects placed uniformly has the sources 175 meters apart. 6 sources placed uniformly would be 140 meters apart. Here is the ChatGPT that supports my drawing.\n\n*If you want to place the objects including the endpoints of the 700-meter line uniformly, the spacing between them is calculated as:\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F1062256%2Fe035be8d67f4c9178b3fe25dede7f6d6%2FScreenshot%202025-04-15%20at%2020-08-23%20Dataset%20File%20Error%20Debugging.png?generation=1744762153020895&alt=media)\n\nThe coordinates (in meters) will then be:\n\n 0 m\n\n 175 m\n\n 350 m\n\n 525 m\n\n 700 m\n\nThese are the positions of the 5 objects along the line from 0 to 700 meters.\n\nIf you include the endpoints of the 700-meter line, placing 6 objects uniformly means dividing the line into 5 equal segments. The spacing between objects is:\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F1062256%2F01d437e9c9a1ef79e6b27db2f825d1d4%2FScreenshot%202025-04-15%20at%2020-10-11%20Dataset%20File%20Error%20Debugging.png?generation=1744762259417576&alt=media)\nThus, the coordinates of the 6 objects along the line are:\n\n 0 m\n\n 140 m\n\n 280 m\n\n 420 m\n\n 560 m\n\n 700 m*\n\n\nBased on examples from host git and some of the images it's a reasonable conclusion that it's 5 sources and they are located 175 meters apart.", "votes": 4 }, { "id": 3179982, "postDate": "2025-04-16T01:52:17.297Z", "content": "

Yeah, echo the confusion/question:

\n

Receivers: 0 to 690 meters, or 0 to 700? (10 meters apart or evenly spaced ~10.14 meters apart to reach 700 meters?)

\n

Sources: 0 to 560, 0 to 690, or 0 to 700? 140, 172.5, or 175 meters apart?

", "rawMarkdown": "Yeah, echo the confusion/question:\n\nReceivers: 0 to 690 meters, or 0 to 700? (10 meters apart or evenly spaced ~10.14 meters apart to reach 700 meters?)\n\nSources: 0 to 560, 0 to 690, or 0 to 700? 140, 172.5, or 175 meters apart?", "votes": 2 }, { "id": 3180466, "postDate": "2025-04-16T16:54:19.937Z", "content": "

Hope this figure helps you understand the configuration. For non-integer source positions, we round them to int.
\n\"\"

", "rawMarkdown": "Hope this figure helps you understand the configuration. For non-integer source positions, we round them to int. \n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F6128879%2F82e7dd142a314d25c4648c8e802f1d8c%2FScreenshot%202025-04-16%20at%2009.36.31.png?generation=1744821431428622&alt=media)", "votes": 5 }, { "id": 3180468, "postDate": "2025-04-16T17:00:40.663Z", "content": "

hmmm not seeing a figure :)

", "rawMarkdown": "hmmm not seeing a figure :)" }, { "id": 3180532, "postDate": "2025-04-16T18:34:05.260Z", "content": "

I guess I'll keep asking, sorry, since we're talking about a highly precise technical simulator, and both non-ints are 0.5 remainder:
\nRound towards middle, round towards outside? (Symmetric). Or round down, or round up (asymmetric)?

", "rawMarkdown": "I guess I'll keep asking, sorry, since we're talking about a highly precise technical simulator, and both non-ints are 0.5 remainder:\nRound towards middle, round towards outside? (Symmetric). Or round down, or round up (asymmetric)?" }, { "id": 3180634, "postDate": "2025-04-16T22:11:21.147Z", "content": "

The sources are rounded to the nearest integer grid, which makes the source positions in the form of index [ 0, 17, 34, 52, 69].

", "rawMarkdown": "The sources are rounded to the nearest integer grid, which makes the source positions in the form of index [ 0, 17, 34, 52, 69].", "votes": 2 }, { "id": 3186689, "postDate": "2025-04-25T03:50:33.783Z", "content": "

Do we know the value of the source term? Or do we need to approximate it from the data like taking the double derivative of the pressure data at the source

", "rawMarkdown": "Do we know the value of the source term? Or do we need to approximate it from the data like taking the double derivative of the pressure data at the source", "votes": 1 }, { "id": 3186705, "postDate": "2025-04-25T04:09:47.307Z", "content": "

The source term is an amplitude 1 Ricker wavelet with 15 Hz peak frequency. You may approximate it from the data if you want. The Ricker wavelet itself is kind of standard. You can refer to the MatLab code named in the OpenFWI paper for how to generate the wavelet.

", "rawMarkdown": "The source term is an amplitude 1 Ricker wavelet with 15 Hz peak frequency. You may approximate it from the data if you want. The Ricker wavelet itself is kind of standard. You can refer to the MatLab code named in the OpenFWI paper for how to generate the wavelet.", "votes": 1 }, { "id": 3219294, "postDate": "2025-06-07T12:54:48.120Z", "content": "

Are all sources firing simultaneously (or synchronously ?) , or one by one ??

", "rawMarkdown": "Are all sources firing simultaneously (or synchronously ?) , or one by one ??", "isDeleted": true }, { "id": 3219325, "postDate": "2025-06-07T13:56:27.703Z", "content": "

Sources are fired separately, and we get the waveform record per source, per sensor, per time step.

", "rawMarkdown": "Sources are fired separately, and we get the waveform record per source, per sensor, per time step." }, { "id": 3219792, "postDate": "2025-06-08T10:11:12.403Z", "content": "

Also , I am assuming all instruments were placed at surface , so can I take z=0 (in grid coords) or should take 1 or 2 ?

", "rawMarkdown": "Also , I am assuming all instruments were placed at surface , so can I take z=0 (in grid coords) or should take 1 or 2 ?\n\n", "isDeleted": true } ] } ] } ], "comments": [ { "id": 3176894, "author_name": "Hanchen Wang", "author_url": "", "post_date": "2025-04-11T21:43:48.467000", "content": "

In the OpenFWI dataset and the test set for this competition, there are 5 sources and 70 receivers all uniformly distributed on the top surface. The source distance is 140m and the receiver distance is 10m. Please refer to OpenFWI paper table 3 for more details: https://proceedings.neurips.cc/paper_files/paper/2022/hash/27d3ef263c7cb8d542c4f9815a49b69b-Abstract-Datasets_and_Benchmarks.html

\n

You are absolutely right that the source-source and receiver-receiver distances play a critical role in velocity map reconstruction. In the OpenFWI paper and this competition, we simplify the problem by considering fixed source/receiver numbers and positions.

", "votes": 3, "replies": [ { "id": 3179714, "author_name": "doteeee", "author_url": "", "post_date": "2025-04-15T16:05:27.190000", "content": "

thanks @hanchenwang114 , went through paper. looking forward to build some good solution to the problem, thanks for the competition.

", "votes": 0, "replies": [] }, { "id": 3179872, "author_name": "Robert Hatch", "author_url": "", "post_date": "2025-04-15T19:57:43.633000", "content": "

So is it, in single dimension:
\nreceivers 0-6, 5m, source 0, 5m, receivers 7-20, 5m, source 1, 5m, receivers 21-34, … source 4, 5m, receivers 63-69?

\n

Given that the total length is 700m, but there's 5 not 6 sources, and 70 not 71 receivers? It's not yet fully clear to me how sources and receivers align.

", "votes": 1, "replies": [ { "id": 3179884, "author_name": "Hanchen Wang", "author_url": "", "post_date": "2025-04-15T20:13:51.340000", "content": "

The same grid point can serve as both a source and a receiver. There are 70 grid points in the horizontal dimension, and each grid point refers to a 10 meter distance in space. Each grid works as a receiver, and thus, there are 70 receivers (let's say index [0, 69]) with 10 meter interval in between, starting from grid index 0 to 69. There are 5 sources (index [0,4]) with source 0 on grid 0, source 5 on grid 69, and the rest 3 uniformly distributed in between.

", "votes": 1, "replies": [] }, { "id": 3179938, "author_name": "PC Jimmmy", "author_url": "", "post_date": "2025-04-16T00:01:10.597000", "content": "

Blame a 78 year old brain - but I cannot get your numbers to match a sketch I am drawing. A 700 meter line having 5 objects placed uniformly has the sources 175 meters apart. 6 sources placed uniformly would be 140 meters apart. Here is the ChatGPT that supports my drawing.

\n

*If you want to place the objects including the endpoints of the 700-meter line uniformly, the spacing between them is calculated as:
\n\"\"

\n

The coordinates (in meters) will then be:

\n
m\n\nm\n\nm\n\nm\n\nm\n
\n

These are the positions of the 5 objects along the line from 0 to 700 meters.

\n

If you include the endpoints of the 700-meter line, placing 6 objects uniformly means dividing the line into 5 equal segments. The spacing between objects is:

\n

\"\"
\nThus, the coordinates of the 6 objects along the line are:

\n
m\n\nm\n\nm\n\nm\n\nm\n\nm*\n
\n

Based on examples from host git and some of the images it's a reasonable conclusion that it's 5 sources and they are located 175 meters apart.

", "votes": 4, "replies": [] }, { "id": 3179982, "author_name": "Robert Hatch", "author_url": "", "post_date": "2025-04-16T01:52:17.297000", "content": "

Yeah, echo the confusion/question:

\n

Receivers: 0 to 690 meters, or 0 to 700? (10 meters apart or evenly spaced ~10.14 meters apart to reach 700 meters?)

\n

Sources: 0 to 560, 0 to 690, or 0 to 700? 140, 172.5, or 175 meters apart?

", "votes": 2, "replies": [] }, { "id": 3180466, "author_name": "Hanchen Wang", "author_url": "", "post_date": "2025-04-16T16:54:19.937000", "content": "

Hope this figure helps you understand the configuration. For non-integer source positions, we round them to int.
\n\"\"

", "votes": 5, "replies": [] }, { "id": 3180468, "author_name": "PC Jimmmy", "author_url": "", "post_date": "2025-04-16T17:00:40.663000", "content": "

hmmm not seeing a figure :)

", "votes": 0, "replies": [] }, { "id": 3180532, "author_name": "Robert Hatch", "author_url": "", "post_date": "2025-04-16T18:34:05.260000", "content": "

I guess I'll keep asking, sorry, since we're talking about a highly precise technical simulator, and both non-ints are 0.5 remainder:
\nRound towards middle, round towards outside? (Symmetric). Or round down, or round up (asymmetric)?

", "votes": 0, "replies": [] }, { "id": 3180634, "author_name": "Hanchen Wang", "author_url": "", "post_date": "2025-04-16T22:11:21.147000", "content": "

The sources are rounded to the nearest integer grid, which makes the source positions in the form of index [ 0, 17, 34, 52, 69].

", "votes": 2, "replies": [] }, { "id": 3186689, "author_name": "chewy", "author_url": "", "post_date": "2025-04-25T03:50:33.783000", "content": "

Do we know the value of the source term? Or do we need to approximate it from the data like taking the double derivative of the pressure data at the source

", "votes": 1, "replies": [] }, { "id": 3186705, "author_name": "Hanchen Wang", "author_url": "", "post_date": "2025-04-25T04:09:47.307000", "content": "

The source term is an amplitude 1 Ricker wavelet with 15 Hz peak frequency. You may approximate it from the data if you want. The Ricker wavelet itself is kind of standard. You can refer to the MatLab code named in the OpenFWI paper for how to generate the wavelet.

", "votes": 1, "replies": [] }, { "id": 3219294, "author_name": "", "author_url": "", "post_date": "2025-06-07T12:54:48.120000", "content": "

Are all sources firing simultaneously (or synchronously ?) , or one by one ??

", "votes": 0, "replies": [] }, { "id": 3219325, "author_name": "CPMP", "author_url": "", "post_date": "2025-06-07T13:56:27.703000", "content": "

Sources are fired separately, and we get the waveform record per source, per sensor, per time step.

", "votes": 0, "replies": [] }, { "id": 3219792, "author_name": "", "author_url": "", "post_date": "2025-06-08T10:11:12.403000", "content": "

Also , I am assuming all instruments were placed at surface , so can I take z=0 (in grid coords) or should take 1 or 2 ?

", "votes": 0, "replies": [] } ] } ] } ], "raw_markdown_by_id": { "3176830": "After spending some time on understanding waveform inversion, got few questions on the placement of sources and sensors.\n\nWe see multiple sources and receivers in the waveform dataset.\nwhat could be the distance between sources and receivers.?\n\nI donot have good understanding of problem at hand, but i feel the distance between source-source, sensor-sensor and source-senor might play in determining velocity maps\n\nThanks, ", "3176894": "In the OpenFWI dataset and the test set for this competition, there are 5 sources and 70 receivers all uniformly distributed on the top surface. The source distance is 140m and the receiver distance is 10m. Please refer to OpenFWI paper table 3 for more details: https://proceedings.neurips.cc/paper_files/paper/2022/hash/27d3ef263c7cb8d542c4f9815a49b69b-Abstract-Datasets_and_Benchmarks.html\n\nYou are absolutely right that the source-source and receiver-receiver distances play a critical role in velocity map reconstruction. In the OpenFWI paper and this competition, we simplify the problem by considering fixed source/receiver numbers and positions. " } }