You are Home   »   News   »   View Article

Full Waveform Inversion moving ahead and Acceleware

Thursday, October 1, 2015

Seismic interpretation algoirthms and software company Acceleware says that there is a growing interest in full waveform seismic processing.

Acceleware, a company based in Calgary which produces software and algorithms for seismic interpretation and processing, reports that there is growing interest in 'full waveform inversion' seismic processing.

Full waveform inversion is a highly compute intense process which converts seismic recordings into a subsurface model in a single process - or in other words getting quantitative information about subsurface properties directly from seismic data.

It works by developing an initial subsurface model, and working out what seismic response you would get if the real world was the same as the model. (This is known as 'synthetic seismic').

The software compares the actual seismic recording with the synthetic seismic, and keeps tweaking the model until they match.

The technology was originally developed in the mid-1980s, but it is only recently that the computing power has been available to do it fully.

Even with today's technology, the interpretation / processing work uses a lot of hardware resources, the company says. The work can still involve a lot of manual effort - efforts are being made to reduce this, and also to improve accuracy of the results.

Acceleware has developed a powerful FWI platform (AxFWI) to run the research code provided by customers. It is designed to allow researchers to customize the specifics of the algorithm to the dataset. Acceleware applies its high performance computing expertise resulting in large computational savings.

Acceleware also develops algorithms for another seismic processing technique called 'Reverse Time Migration (RTM)', a process to convert seismic data (recorded in time) to seismic data by depth, or calculating how fast the seismic data will go through different areas of the rock (velocity modelling). It is suited for very complex geologies, such as areas with complex salt.

It models the seismic wave going both downwards and upwards, including through complex propagation paths (which are normally just treated as 'noise' in the imaged data).

Technology for RTM is maybe ten years ahead of where it is now for full waveform inversion (FWI), says Geoff Clark, chief executive officer of Acceleware. 'RTM is a mature product,' he said.

Also, RTM is a significant component of FWI.

In April 2015, Acceleware announced an agreement with Spanish oil major REPSOL to work together developing custom reverse time migration (RTM) seismic imaging software.

The deal is worth an expected $2.1m, with $1.3m payable in the first year, the remainder over a 3 year maintenance period.

Acceleware also makes software for 'forward modelling', which means modelling what seismic response you can expect to record from a seismic survey you are planning.

Acceleware' s software algorithms are used as part of software by several other companies, including Paradigm, Tsunami Development and Geotomo, and sold to some oil and gas companies (such as Repsol) who want to use it as part of their own tools.

Microchips

A key factor with computing intense seismic processing is the choice of microprocessors.

Acceleware was one of the first companies to do computing on GPUs (graphics processing units), microchips designed partly for computer games, but which are also very good for seismic processing calculations.

Microchip manufacturer NVIDIA has acquired 5 per cent of Acceleware, and it sees the oil and gas industry as one of the biggest users of its 'GPU' graphical processing unit microchips, Mr Clark says.

Intel has developing a similar microchip called 'Xeon Phi', and also microchip manufacturer AMD are starting to increase focus on this market.

GPUs have thousands of 'cores' or single computing components on them, which enable the processing task to be broken into thousands of small problems and processed at the same time.

A typical processing task will run 7-10 times faster on a GPU than on an equivalent CPU, Mr Clark says. The microchip will probably cost double, but that means you are getting about 3.5 times more processing power for the same cost. GPUs are usually thought to have twice the operating cost, because they use more power.



Associated Companies
» Acceleware
comments powered by Disqus

CREATE A FREE MEMBERSHIP

To attend our free events, receive our newsletter, and receive the free colour Digital Energy Journal.

FEATURED VIDEO

How much time are you wasting on data management?
Ross Philo
from Energistics

DIGITAL ENERGY JOURNAL

Latest Edition Jan 2017 issue
Jan 2017

Download latest and back issues

COMPANIES SUPPORTING ONE OR MORE DIGITAL ENERGY JOURNAL EVENTS INCLUDE

Learn more about supporting Digital Energy Journal