Overview of the Faults and Fractures plugin in OpendTect, featuring fault likelihood, basement fracture attributes, fracture density and proximity, and edge-preserving filters. The presentation also highlights machine learning fault prediction and diffraction imaging for high-resolution fracture and fault detection.
#OpendTect #FaultAndFractures #SeismicAttributes #FractureAnalysis #MachineLearning #DiffractionImaging #StructuralGeology #SeismicInterpretation
Duration: 8:34
--- Transcript ---
So Faults and Fractures is a plugin to OpendTect Pro it combines all the attributes filters and utilities that we have for Faults. And Fractures analysis in OpendTect the main attribute here is Fault likelihood and it's derivative thinned fault likelihood unique to the dGB solution is. Also that we have a solution for basement fractures and in certain basins occur as very steeply dipping Reflections.
So you can't use the fault likelihood attribute but we have an attribute that is based on fault likelihood that actually follows these very deep steeply dipping reflectors we have basement fractures likelihood. And its derivative the thinned basement fracture likelihood then we have a number of attributes which are. Also quite useful for fracture analysis two of them are fracture density and Fracture proximity we have automated fault plane extraction tools.
And we have a very powerful Edge preserving smoothing filter very often people say yes faults Thinned Fault Likelihood is a very nice attribute. But it produces so many small elements that is there a possibility to filter it so can we get rid of the smaller things. And in OpendTect yes you can do that the workflow is through the automatic extraction of fault planes because in that automatic extraction of fault planes you already set a minimum size for the fault planes that will be extracted.
And that is one element of the workflow and the second element is that we have once the fault planes are extracted a number of Sliders that can be used to minimize the number of elements that you really want to convert into fault planes. And also which you want to resave as Thinned Fault Likelihood volume and these sliders I can just use to reduce the very small ones get rid of them I can use them to only select faults in a certain direction or only dipping in a certain dip range. And then when I'm done and I've selected all the ones that I want to keep I can save these skins again as TFL as thinned fault likelihood volume.
Now there's a visualization tip here if you display TFL on sections it's often very nice to look at. But if you display this on time slices or along Horizons the visualization is a bit hampered because the TFL attribute is. So thin it's just one sample thick that you can hardly see it and then we use this trick we use our attribute engine to beef up the TFL values Again by Computing an RMS in a very small volume surrounding each TFL value.
And that gives very nice displays in the horizontal slice domain here we see an example of a filtering the Thinned Fault Likelihood. So on the left we have left upper corner let me put a laser pointer here in this image we see the input seismic here we see the TFL as computed by the algorithm. Then we go to this filtering option we get rid of the smaller elements and you can see here on this time slice it's difficult to see where these faults are.
So then we beef it up with the RMS filter TFL and then we can see all the larger faults. And we have compared to this one removed a lot of the smaller stuff that we didn't want to see examples of fracture density on the right. And Fracture proximity on the left fracture density gives me the sweet spots on the in the data input is the TFL response this is with different cutoff values.
And here we see where we have most of these TFL values and because that is where it is mostly broken it's. Also that is where we find the largest number of fractures the opposite of the fracture density is fracture proximity here we see where we have clean seismic data without any TFL responses. So that is what you would like to drill for instance in geothermal development in the Netherlands you want to stay away from the faults.
And you want to stay only and drill only in the undisturbed area yes because of seismicity problems. Now an example of filters the first one is our edge preserved smoothing filter this is my input seismic as you can see it's quite noisy. And here we see the edge preserves smoothing result look at this fault and also here these faults over here toggle it a couple of times input EPS filtered.
Now this is compared to our standard dip steered median filter which also has Edge preserving properties it removes random noise. But it is not smoothing the data as much this I could still use for quantitative interpretation because everything is well preserved in terms of Quant of amplitude Behavior it's just a little bit shuffled.
And the edges are not as well preserved as in the EPS volume but it's still Edge preserving either input we can use for fault likelihood or in Fault likelihood computations. Now then a few slides on machine learning and Diffraction Imaging because machine learning obviously we have. Also tools to predict faults and some of these tools are delivered with the software in the form of pre-trained models I will show you just one example or two examples here on the left we have input seismic. And this is the application of a unet fault predictor in threedimensional fault predictor and this is after thinning.
So what happens here is that this model takes a block of seismic data it passes it through the trained model. And the model knows how to transform this into fault likelihood and the block is a similar size as this one. And we do that block by block we splice the blocks together and we have our machine learning fault prediction volume here we see another model with which is interpolating seismic traces fault net is our best fault predictor it's a threedimensional unit.
And this output here is not thinned yet so this is the fault likelihood as predicted by the machine learning model. Then we also offer a Diffraction Imaging Service Diffraction Imaging is or this service is offered in association with Moser Geophysical Services Tijmen Jan Moser is really a guru on this in this field. And the technique that he has developed is splitting the seismic input pre-stack depth migrated seismic input into a reflection seismic.
And Diffraction images and these Diffraction images are very high resolution they pick up the fractures and faults at the highest possible resolution the technology developed by Tijmen Jan. And his colleague Mike Pelissier is also integrating interpretation Services they call it customization to interpretation and they use the processing. And they tune it in such a way that you see the optimal results from an interpretation perspective this type of technology is ideal for detecting fractures in carbonate in Basements.
And in stratigraphic features to highlight these as well.