OpendTect Pro Demo: Faults & Fractures Plugin for Fault Detection & Extraction with Machine Learning скачать в хорошем качестве

OpendTect Pro Demo: Faults & Fractures Plugin for Fault Detection & Extraction with Machine Learning

Demo of the Faults and Fractures plugin in OpendTect, showing how to filter seismic data, compute fault likelihood, and apply machine learning predictions for fault detection. The demo also covers automatic fault plane extraction, fault skinning, and visualization with rose diagrams for structural interpretation. #OpendTect #FaultsAndFractures #SeismicAttributes #Geophysics #FaultDetection #MachineLearning #StructuralGeology #SeismicInterpretation -- Transcript -- And I'll start with the demo of the Faults and Fractures plugin so what I have here is a dataset with a number of precooked elements. And the Faults and Fractures plugin is started from this icon over here and it has many different attributes that can be computed it has different filters to pre-process the data or to enhance the data by skeletonization. So thinning the output you can extract fault planes and there are tools utilities like calculating Rose diagrams. Now let's first have a look at a few filters and I've already got a few filters displayed over here. So we can just compare that or switch of the horizons but that it's not in front of us. And I first go and look at this seismic line that is my input seismic line this is how the seismic is what I got this is after dip steered median filtering. So it's just removing the random noise this is The Edge preserved smoothing filter so that is much sharper the edges it has. Also really averaged out everything so this is not really to be used for quantitative interpretation anymore. But it can be really useful for input as input for structural analysis and for computing the fault likelihood for instance. Now what does the fault likelihood look like that looks like this is based on a semblance calculation. And the semblance is calculated in different directions that is different dip directions and also different azimuth directions. So you're really specifying if we look here on how that is done the fault likelihood is this attribute just press go I specify my input. And this could then either be the dip steered median filter or the edge preserved smoothing seismic. And then in the advanced settings you can see that we are Computing these semblances in 25 different directions. And also in the dip between 25 65 degrees in six different steps and the result of that looks like this. Now this in itself is not very useful it becomes useful if I either thin it or I use it as input for automatic fault plane extraction. So if I thin it then it looks like this that is the Thinned Fault Likelihood and you can see because it was used from the EPS it really follows the defaults here very nicely okay that is just a few inputs that you can directly do from The Faults. And Fractures plugin I also have a few predictions from machine learning and I just want to show these as well. So this again is my original seismic this is the EPS smooth seismic this is the tfl from the smooth seismic. And this is a unet that is applied and predicted on this line the fault likelihood so it's a threedimensional application. So I have this everywhere in the 3D seismic so this is before thinning I can then skeletonize it. And get a response which is more like the TFL also have a fault net but only this small portion over here. And if we look at that you can see this is really a much better predictor than the previous unit this is very sharp. And this is even before thinning compared to this one which is the threedimensional unet and this is the fault net that is giving me a better predictor okay back to Faults. And fractures I would like to demo now the automatic fault plane extraction and for that I switch all my Horizon again I switch of this line let's move it over here. So I'm going to compute this only in a small area around this Horizon where we do have these nice faults extract planes is the tool I'm doing that from the fault likelihood. And I'm doing it only in a sub volume run it so from the fault likelihood we are. Now looking for which faults belong together which outputs belong together and can be combined into a fault plane. But this is not really a fault plane yet and let me switch off this one these are if I zoom in you can see these are individual points. And these points have been collected to become a what is called a fault skin and there is kind of grid from which I can create fault planes in seismic interpretation software a fault plane is basically a collection of fault sticks which are. Then interpolated to create a fault plane so if I want to convert this into a fault plane I can. Also use in other softwares I have to extract these and Sample them with fault sticks so that they can. Then become fault planes and we used in other softwares now what we see here is a rose diagram where each of these fault planes Falls we see these sliders.