AI for seismic interpretation

A TotalFinaElf researcher showed prototype software at the EAGE that could change the way we look at seismic data. Sismage applies image processing and artificial intelligence to the 3D dataset - to extraordinary effect!

Naamen Keskes - the inventor of Stratimagic has struck again with an increasingly functional tool for stratigraphic interpretation of seismic data. TotalFinaElf’s (TFE) Sismage was originally the core of CGG-Petrosystems-Paradigm’s Stratimagic, but since Stratimagic’s productization, TFE’s researchers have been working away on enhancing the original research tool.

Showman

Keskes, a born showman, started his demonstration at the EAGE by picking a couple of seeds in the deep offshore seismic data volume and hey presto!, Sisimage auto-extracts a whole meander system. Keskes believes that “No one wants maps or blocks - they need [geological] objects.”

Neural net

The neural network technology originally developed for Stratimagic has been extended so that with a small initial learning phase - almost any geological feature can now be auto-extracted from the data. Autopicking faults for instance gives a rosace diagram of strain. A chaotic, high amplitude facies is instantly translated into a myriad of geobodies - and their cumulative volume pops up in a spreadsheet.

DHI

If your game is direct hydrocarbon indication - a special stacking tool tests multiple stacks to emphasis sub-horizontal events associated with potential oil-water contacts - a previously invisible flat spot leaps forth out of the data! Near trace gathers are exploited in conjunction with a database of outcrop analogues and the neural net identifies facies patterns against pre-computed near and far trace behavior.

Vail et al.

Seismic stratigraphy of the Vail school your bag? The seismic is instantly filled with downlaps, toplaps and triple points. A time-stratigraphic chart is extracted from the seismic data and re-injected into the seismic framework.

Tour de force

For Keskes, success comes from combining images, texture and sequence boundaries. A final tour de force was the extraction of color-coded wormy fluvial bodies from the Angolan Dhalia field. Even though some of the demos appear a bit contrived, for instance, the ‘worms’ all seem to have the same lateral geometry, one feels that one is looking at the future of seismic interpretation here; that the amount of information to be obtained from the 3D seismic dataset is without limits.

Click here to comment on this article

Click here to view this article in context on a desktop

© Oil IT Journal - all rights reserved.