The main purpose of this project is to develop a set of software technologies for numerical simulation and solution of inversion problems. Applications such as the acquisition of seismic images by the inversion of measured field data predate in many years the great current interest in analysis of large data volumes and machine learning. However, these applications remain extremely challenging because they did not benefit significantly from these new developments, given the complexity inherent in the problems, large data volumes and the high computational cost. The software required for these problems is highly specialized, both in terms of the numerical methods and the high-performance computing techniques involved, and requires years of specialized work to be developed. This is a serious barrier to the creation of new methods and innovation, for example, to obtain better images of salt and pre-salt layers commonly found along the Brazilian coast.
This project will develop specialized software technologies capable of reducing the time required for the development of simulation and inversion codes from years to days. High-level programming languages will be designed and implemented that will enable geophysicists to use their area-specific knowledge to rapidly develop and evaluate new algorithms for imaging geologically challenging regions. Discretization methods in finite differences and high order finite elements, optimized for seismic processing, will be developed, as well as techniques of automatic differentiation for the automatic generation of the adjoint models. Another focus of the work will be to achieve a higher level of automation in the spatial discretization of the domain through the generation and adaptation of meshes, which is a task that currently consumes many hours of human work. Finally, modern compiler technologies will be employed to generate highly optimized code for computing platforms from supercomputers to cloud systems.
The position involves taking part in the following main activities:
1. Develop and validate software for modelling acoustic forward problem with numerical techniques such as finite differences or element methods
2. Develop and validate software for solving acoustic inverse problem by using the topology optimization method with the purpose of generating acoustic tomographies
3. Develop a general framework with HPC techniques for acoustic tomography by using the topology optimization method
Requirements to fill the position (Ex: specific experience, minimum or maximum years after concluding the course)
PhD level specialist in Computational Mechanics, Transient Analysis, Finite Differences or Element Methods, Topology Optimization, Transient Adjoint Methods, Analysis of Algorithms, Phyton or Matlab Programming and HPC Processing.
INFORMATION ABOUT FELLOWSHIP
The selected candidate will receive a FAPESP
Post-Doctoral fellowship in the amount of R$ 7.373,10 monthly payed in Reais and a research contingency fund (technical reserve), equivalent to 15% of the annual value of the fellowship which should be spent on items directly related to the research activity, as well as displacement funding, if necessary and applicable. More information about the fellowship is at: fapesp.br/en/postdoc
There is the possibility of offering a Research Internship abroad (BEPE) during part of the postdoctoral assignment, if it is of interest to the project. In this situation, the selection of the institution and the period will be defined by the project coordinator, depending on the purpose of the internship and the needs of the project. http://www.fapesp.br/6557
MORE INFORMATION AND APPLICATION AT http://www.rcgi.poli.usp/opportunities REF 19PDR117