As title, do you have any idea on what a single value of acoustic_data means? What does a single value of 0 for acoustic_data mean?
From the data page, the acoustic_data input signal is used to predict the time remaining before the next laboratory earthquake (time_to_failure).
And from Bertrand RL's additional info post, the seismic data is recorded using a piezoceramic sensor, which outputs a voltage upon deformation by incoming seismic waves. The seismic data of the input is this recorded voltage, in integers. But what does a value of zero voltage value mean?
\n\nAnd from this pdf, the signal radiated from a seismic source, be it an explosion or a shear rupture, is usually a more or less complicated displacement step function or velocity impulse of finite duration from milliseconds up to a few minutes at the most.
\n\nI am not sure whether this interpretation is in line with the data collected in this competition.
\n\nLet me know your two cents!
", "messages": [ { "id": "497708", "postDate": "03/23/2019 20:20:36", "content": "As title, do you have any idea on what a single value of acoustic_data means? What does a single value of 0 for acoustic_data mean?
From the data page, the acoustic_data input signal is used to predict the time remaining before the next laboratory earthquake (time_to_failure).
And from Bertrand RL's additional info post, the seismic data is recorded using a piezoceramic sensor, which outputs a voltage upon deformation by incoming seismic waves. The seismic data of the input is this recorded voltage, in integers. But what does a value of zero voltage value mean?
\n\nAnd from this pdf, the signal radiated from a seismic source, be it an explosion or a shear rupture, is usually a more or less complicated displacement step function or velocity impulse of finite duration from milliseconds up to a few minutes at the most.
\n\nI am not sure whether this interpretation is in line with the data collected in this competition.
\n\nLet me know your two cents!
", "rawMarkdown": "As title, do you have any idea on what a single value of `acoustic_data` means? What does a single value of 0 for `acoustic_data` mean?\n\nFrom the data page, the `acoustic_data` input signal is used to predict the time remaining before the next laboratory earthquake (`time_to_failure`). \n\nAnd from [Bertrand RL's additional info post](https://www.kaggle.com/c/LANL-Earthquake-Prediction/discussion/77526), the seismic data is recorded using a piezoceramic sensor, which outputs a voltage upon deformation by incoming seismic waves. The seismic data of the input is this recorded voltage, in integers. But what does a value of zero voltage value mean?\n\nAnd from this [pdf](http://gfzpublic.gfz-potsdam.de/pubman/item/escidoc:4017:4/component/escidoc:4018/Chapter_4_rev1.pdf), the signal radiated from a seismic source, be it an explosion or a shear rupture, is usually a more or less complicated displacement step function or velocity impulse of finite duration from milliseconds up to a few minutes at the most.\n\nI am not sure whether this interpretation is in line with the data collected in this competition.\n\nLet me know your two cents!", "votes": null }, { "id": "499272", "postDate": "03/24/2019 15:28:17", "content": "From Bertrand RL's \"Additional info\" post:\nThe seismic data is recorded using a piezoceramic sensor, which outputs a voltage upon deformation by incoming seismic waves. The seismic data of the input is this recorded voltage, in integers.
", "rawMarkdown": "From Bertrand RL's \"Additional info\" post:\nThe seismic data is recorded using a piezoceramic sensor, which outputs a voltage upon deformation by incoming seismic waves. The seismic data of the input is this recorded voltage, in integers.", "votes": null }, { "id": "499393", "postDate": "03/24/2019 18:17:21", "content": "It's the output voltage of a device that responds to physical displacement, so acoustic_data is analogous to an audio amplitude reading at each point in time. It would be nice to know some further details of the device, for example over what range its I/O relationship is linear (if anywhere). I don't know if anyone has improved results via a numerical transform of the signal data.
Yeah I read that too (sorry I forgot to paste that in my original post). But still what does a voltage of 0 value mean? Or is the difference between 2 and 1 is the same as the difference between 12 and 11? I still don't get it.
", "rawMarkdown": "Yeah I read that too (sorry I forgot to paste that in my original post). But still what does a voltage of 0 value mean? Or is the difference between 2 and 1 is the same as the difference between 12 and 11? I still don't get it.", "votes": null }, { "id": "499479", "postDate": "03/24/2019 20:44:22", "content": "This is my understanding, and it may be imperfect so I would listen to other people's views as well:
\n\nacoustic_data overall ~4.5VI think the reason for your confusion is that the mean of acoustic_data is not 0, which we would expect if it was just a facsimile for the distance moved by a static, vibrating material. But a voltage has to be kept over the sensor for it to work, hence why the mean is around 4.5V. Each recording of the voltage/displacement is a snapshot of one instance in time as the acoustic waves pass through the sensor, and we have to piece these readings together so the overall waves can be analysed.
Thanks for your reply, @bigironsphere. I am still on my way to understand the data.
", "rawMarkdown": "Thanks for your reply, @bigironsphere. I am still on my way to understand the data.", "votes": null }, { "id": "501962", "postDate": "03/28/2019 01:56:15", "content": "In case it helps, the accelerometer they were using in 2013 (probably the same one being used today), is a Brüel and Kjær model 4393:\nhttp://www3.geosc.psu.edu/~cjm38/papers_talks/Johnson_etal_GRL2013.pdf\nhttps://www.bksv.com/en/products/transducers/vibration/Vibration-transducers/accelerometers/4393\nand the amplifier is:\nhttps://www.bksv.com/en/products/transducers/conditioning/charge/2635
\n\nThe accelerometer is a piezoelectric device, so doesn't require any voltage to be supplied to it. Zero volts ought to mean zero acceleration. The DC offset of the amp is supposed to only be +-50mV. On the other hand, the maximum output of the amp is supposedly 8V, and there are lots of points over 8 in the data, so I suspect there's more to the setup than I have surmised. For example, the DAC they use after the amplifier may have a DC offset (and maybe a gain too).
\n\nWhy is the voltage data not symmetric about zero? I don't know. It's interesting to note that the mean moves around through the training data (in addition to sawtoothing as the noise increases through a quake cycle). Maybe the amplifier drifts, or maybe there's some other source of low-frequency noise. Amplifier drift seems more likely to me because the accelerometer is only suposed to be good down to 0.1 Hz, and the drift varies over tens of seconds. Another possible source of slow drift might be the servomechanisms they use to maintain the normal force on the side plates.
\n\n What is the low-level noise we see of a few volts in amplitude? It's hard to know the signal-to-noise of the amplifer without knowing the settings they use. But more likely the low-level noise comes from the fault gouge; maybe small rearrangements of the microspheres or cracks forming in individual microspheres?\nFrom Bertrand RL's \"Additional info\" post:\nThe seismic data is recorded using a piezoceramic sensor, which outputs a voltage upon deformation by incoming seismic waves. The seismic data of the input is this recorded voltage, in integers.
", "votes": null, "replies": [ { "id": 499448, "author_name": "pukkinming", "author_url": "", "post_date": "03/24/2019 19:39:44", "content": "Yeah I read that too (sorry I forgot to paste that in my original post). But still what does a voltage of 0 value mean? Or is the difference between 2 and 1 is the same as the difference between 12 and 11? I still don't get it.
", "votes": null, "replies": [] }, { "id": 499479, "author_name": "bigironsphere", "author_url": "", "post_date": "03/24/2019 20:44:22", "content": "This is my understanding, and it may be imperfect so I would listen to other people's views as well:
\n\nacoustic_data overall ~4.5VI think the reason for your confusion is that the mean of acoustic_data is not 0, which we would expect if it was just a facsimile for the distance moved by a static, vibrating material. But a voltage has to be kept over the sensor for it to work, hence why the mean is around 4.5V. Each recording of the voltage/displacement is a snapshot of one instance in time as the acoustic waves pass through the sensor, and we have to piece these readings together so the overall waves can be analysed.
Thanks for your reply, @bigironsphere. I am still on my way to understand the data.
", "votes": null, "replies": [] } ] }, { "id": 499393, "author_name": "bigironsphere", "author_url": "", "post_date": "03/24/2019 18:17:21", "content": "It's the output voltage of a device that responds to physical displacement, so acoustic_data is analogous to an audio amplitude reading at each point in time. It would be nice to know some further details of the device, for example over what range its I/O relationship is linear (if anywhere). I don't know if anyone has improved results via a numerical transform of the signal data.
In case it helps, the accelerometer they were using in 2013 (probably the same one being used today), is a Brüel and Kjær model 4393:\nhttp://www3.geosc.psu.edu/~cjm38/papers_talks/Johnson_etal_GRL2013.pdf\nhttps://www.bksv.com/en/products/transducers/vibration/Vibration-transducers/accelerometers/4393\nand the amplifier is:\nhttps://www.bksv.com/en/products/transducers/conditioning/charge/2635
\n\nThe accelerometer is a piezoelectric device, so doesn't require any voltage to be supplied to it. Zero volts ought to mean zero acceleration. The DC offset of the amp is supposed to only be +-50mV. On the other hand, the maximum output of the amp is supposedly 8V, and there are lots of points over 8 in the data, so I suspect there's more to the setup than I have surmised. For example, the DAC they use after the amplifier may have a DC offset (and maybe a gain too).
\n\nWhy is the voltage data not symmetric about zero? I don't know. It's interesting to note that the mean moves around through the training data (in addition to sawtoothing as the noise increases through a quake cycle). Maybe the amplifier drifts, or maybe there's some other source of low-frequency noise. Amplifier drift seems more likely to me because the accelerometer is only suposed to be good down to 0.1 Hz, and the drift varies over tens of seconds. Another possible source of slow drift might be the servomechanisms they use to maintain the normal force on the side plates.
\n\n What is the low-level noise we see of a few volts in amplitude? It's hard to know the signal-to-noise of the amplifer without knowing the settings they use. But more likely the low-level noise comes from the fault gouge; maybe small rearrangements of the microspheres or cracks forming in individual microspheres?\n