The SEIS2LOG Machine Learning algorithm can create pseudo logs from seismic traces.
- The program builds machine learning models from seismic data attributes combined with the available log data. This results in a segy of pseudo logs at every trace.
- Various statistical measures are also generated to evaluate the prediction process along with performing cross validation.
- The approach is able to capture the log (low frequency) trends. Tuning of the hyper parameters is essential to achieving good amplitude predictions.
- As can be seen from the attached images of several horizontal slices, SEIS2LOG is able to identify the relatively low gamma ray values and their spatial trend, which could help in the interpretation of sand body distribution.
- The best way to use the vertical sections of the segy is to display traces in wiggle form with no filling. Each trace now represents a pseudo log, e.g. pseudo gamma ray, and pseudo density.
- SEIS2LOG has transformed the seismic interpretation to a pragmatic geologic interpretation that petrophysicists, geophysicists, geologists, and reservoir engineers can relate to.
Unfortunately, everyone in the industry only focuses on using logs to build rock physics models which are then used in conjunction with inversion to output the usual acoustic /shear impedance and maximum stretch in porosity volumes. They use Bayesian approach to generate fluid factor volumes. These are based on separation of gas from oil from water in sands from shales. A very ambitious undertaking to say the least.
No one is taking logs and generating logs from seismic. SEIS2LOG is not restricted to any log type. We are open to apply it to any log type. It has been tested on gamma ray, density and permeability. The latter, i.e. permeability, should get the reservoir engineers excited. That is what they are after to input to their simulation.
We envisage the value of SEIS2LOG to oil companies would lie in the ability to bring the seismic closer to the geologists who can perform geologic interpretation on a conventional workstation, much like how geophysicist operate. The value of interpreting lithologies straight from seismic, after converting it to pseudo gamma ray for clastics, pef for carbonates, permeabilities for reservoir modelers and for carbonates interpretation of sweet spots.
Applying SEIS2LOG we bring the seismic amplitudes, which is a vague quantity, to the units of petrophysics and lithology, which is understood by, and required for all other disciplines.
For further details on our SEIS2LOG Machine Learning technology and service offerings, please email us at email@example.com and our technical experts are available to assist.