+ Create new tricision Login 

Login

or


Forgot password?

First visit? Sign-up now!

By continuing you accept the Terms & Conditions and Privacy Policy.

time is up

What are the most pressing machine learning tasks in geoscience?

The exploding field of data analytics (aka data science, aka machine learning) is already part of geoscience. What's so exciting right now is the increasing availability of tools, data, and even skills. What data science or machine learning problems in geoscience seem especially ripe to you? Or, to put it another way: What should we be asking for Watson's help with?


saved
Ideas
Pros and cons
 
Votes
Facies classification from core photographs
by Evan Bianco
One of the most time intensive tasks for a geoscientist is logging core. Automating this process,... more
 
We love rocks, but can do without 3 consecutive months fueled by the coffee at #Nikkelveien13
by Eirik Larsen
 
Human interpretations can be subjective and biased. Image recognition based approaches could help... more
by Lukas Mosser
 
Deep neural networks should do a great task here.
by Clara Castellanos
 
Vast amounts of core photos from Norway are available here: http://hotell.difi.no/?dataset=npd/we... more
by Eirik Larsen
 
8

Brendon Hall, Dan Hallau, Kwinkunks and 5 more

Optimize the seismic processing pipeline
by Kwinkunks
Finding the optimal parameters for the seismic processing flow (nevermind the individual bits) is... more
 
Forward model a bunch of complex scenarios with and without lots of types of noise. Train ML algo... more
by Jonny Ford
 
If we know what function we want to minimize, we can eventually do it. I think the downside is th... more
by Clara Castellanos
5

Anton, Sergey Sergeev, gram and 2 more

Assessing seismic data quality
by Kwinkunks
Would potentially need input from interpreters saying "I can interpret this" or "I am only 40% su... more
5

Matteo, Clara Castellanos, Caio Burin and 2 more

Lithology prediction from well logs
by Kwinkunks
A basic, 'must have' task. Note that this is 'lithology', not 'facies' (a harder problem I think)... more
4

James, Dan Hallau, Fernando Ziegler and 1 more

Well log pick interpretation AI to mimic past top pickers
by Justin Gosses
Many old basins have thousands of wells with tops interpreted by 1-30 people over decades. It sho... more
3
Eirik Larsen
Kwinkunks
Matteo
Estimating Missing Logs
by Eirik Larsen
We're often missing some logs, e.g. the Vs log. We are allready using ML to do this. How can we o... more
 
Before any quality lithology prediction or seismic inversion work to be done, it is indispensable... more
by Anton
 
Add logs where tool failure occured.
by Lukas J. Mosser
2
Anton
James
Data interpolation
by Kwinkunks
Especially spatial interpolation of seismic images, but there are lots of interpolation problems ... more
2
Kwinkunks
Paul Gabriel
Generate Geological cross sections from seismic
by Steve Purves
Could we successfully train something to generate plausible geological cross section diagrams fro... more
2
Kwinkunks
Lukas J. Mosser
Flag hydrocarbon pay zones on logs
by Evan
Using actual perforation intervals to train a model to pick perf zones, identify pay cutoffs, or ... more
 
Detecting bypassed pay could be fantastic. Could we call this wireline classification with "missi... more
by Lukas J. Mosser
1
1
Eirik Larsen
Earthquake prediction
by Eirik Larsen
It would certainly be useful to enable prediction of the timing, location, and magnitude of futur... more
 
That's a very challenging task even for the earthquakes that already happened.
by Curious reader
 
https://www.technologyreview.com/s/603785/machine-learning-algorithm-predicts-laboratory-earthqua... more
by Eirik Larsen
1
Matteo
Interpret Faults
by Steve Purves
Generate fault interpretations from seismic data, with or without the aid of typical fault attrib... more
1
Eirik Larsen
ML as tool to Seismic Attribute interpretation.
by Caio Burin
Different attribute work for different things. But more than one attribute can add information on... more
1
Eirik Larsen
Assessing wireline data quality
by Kwinkunks
Train a model to recognize bad wireline log data. Needs either explicit labels of 'bad data' vs '... more
1
Dan Hallau
Lithology prediction from seismic
by Kwinkunks
Basically just an extension of the prediction from well logs. If you have stacked or pre-stack se... more
1
Eirik Larsen
q
0
Benchmark Datasets
by Lukas J. Mosser
Any machine learning related task, should have a good dataset to work with. In geological models ... more
0
Predict likely outcrop locations. A ML addition to flyoverCountry app
by Justin Gosses
Can you predict where to find quality outcrops in the field before you go. Inputs would be elevat... more
0
Seismic Velocity Model Building and Depth Conversion
by Anton
form 2D to 3D, time and cost-effective (hopefully) procedures to build velocity model, accounted... more
0
Geochemical rock classification
by Antoine Caté
Geochemists use biplot or ternary diagrams to classify rock types (both in the magmatic and sedim... more
0
Seismic Inversion via ML
by gram
Can we teach an AI to invert seismic data? Or migrate seismic data? Or generate a velocity/densit... more
0
Seismic acquisition equipment hazard IDer from satellite data
by gram
Train a bot on satellite data (imagery, elevation, etc.) to highlight problem areas for seismic s... more
0
Integration between Grav/Mag/EM data, Seismic, well log data and core sample
by Caio Burin
Interpretation of subjects in different scales is the most tricky problem in regions where seismi... more
0
Casing failure mu-seismic
by Curious reader
I think a neat problem to attack with ML is to classify a seismic signal with respect to the sour... more
0
Analog media screening, analysis and interpretation
by Evan
Paper, paper, everywhere. Section, logs, maps. Maybe machine learning doesn't need digital data. ... more
0

Comments

Alan J Cohen

best placement of the next wells and frac stages within an existing unconventional Shale play

Mon, Feb 6, 2017

http://www.tricider.com/brainstorming/3Zof8QuO3RZ