CCES Unicamp

Vibration readings and high computational power allow us to ‘see’ the bottom of the ocean

In order to see what is there bellow the bottom of the ocean, such as the pre-salt, scientists have learned to read the vibrations that propagate through the soil and rocks, just as a submarine captain reading the sonar, or a doctor reading an ultrasound. But to create images by interpreting the way waves generated by explosions, impacts, or tremors penetrate the seabed isn’t simple, requiring sophisticated devices and a lot of math.

Researches conducted by the High Performance Geophysics Lab (HPG), in the University of Campinas (Unicamp), Brazil, develop methods and the technology to make these readings more accurate. HPG papers published in recent years involving capturing, processing, and interpreting these vibrations already show results, and have been made available to Petrobras, some already in use, others still being tested. “One of the main contributions brought by these researches is that it reduces uncertainties”, explains HPG researcher Jorge Henrique Faccipieri Junior. “The pre-salt is an example of that. It has a complex geological environment, which demands advanced seismic processing techniques.”

HPG is based on the Center for Petroleum Studies (Cepetro), directed by Professors Martin Tygel (Cepetro) and Edson Borin (Institute of Computing – IC), both head researchers in the Computational Geophysics line of research, at the Research, Innovation and Dissemination Center (RIDC) for the São Paulo Research Foundation (Fapesp).

Furthermore, Faccipieri says the researchers developed by HPD allow more precision in the readings of vibrations passing through subsoil, reducing the risks of errors in locating oil wells, in determining the depth of the reservoir, and in estimating the amount of oil, for example. “Errors in determining these types of information may represent loss of tens to hundreds of millions of reais”, warns the scientist.

The vibrations generating these images may be natural or artificial. “Our studies focus on data from artificial sources”, says Faccipieri. “In the case of seabed data, the vibration source is a compressed air cannon; and onshore, explosives or seismic vibrators are used.” Each region examined goes through hundreds of data collection cycles, and each one requires detonating the vibration source and recording the environment’s response through receivers.

“When the signal spreads, depending on the physical properties of the environment, it may be reflected or transmitted. The ones transmitted can be captured by the receivers”, explains the researcher. The time-lapse between the wave emitted and the reflected vibration is one of the attributes that allow us to generate images representing the subsoil.

An article written by Faccipieri and fellow researchers, entitled ‘Stacking apertures and estimation for reflection and diffraction enhancement’, published by Geophysics (2016), show means of enhancing the reflection or diffractions (diversion) of waves through the subsoil. “The reflection provides key information on the geological structures, its forms and positions. The diffraction is important to determine the average speed of propagation of these waves with greater precision”, states the scientist. He adds that it’s pretty difficult to distinguish one event from the other. “In this paper, we suggest a method for separating them for later handling.”

In addition to these geophysical difficulties, Faccipieri points out that there are complications with processing the gathered data: “The separation process of the reflection and the diffraction requires great computational power, and optimized software able to run tasks simultaneously.” The research group, thus, developed special data processing techniques to handle the task, shown in ‘PY-PITS: A Scalable Python Runtime System for the Computation of Partially Idempotent Tasks’, an article presented in a conference in the United States, also in 2016.

Related posts

Construction of reduced-order models for fluid flows using deep feedforward neural networks

cces cces

CCES researcher elected for the board of the main computing association in the world

escience

São Paulo School of Advanced Science on Learning from Data

cces cces