Carolina Ribeiro Xavier

Genetic Algorithm for the History Matching Problem

Abstract In this work we present a study of genetic algorithms for the automatic history matching problem of reservoir simulation. The history matching process is an inverse problem that searches a set of parameters that minimizes the difference between the model performance and the historical performance of the field. This model validation process is essential and gives credibility to the predictions of the reservoir model. We studied a Parallel Genetic Algorithm implementation, several tests were performed and the preliminary results are presented and discussed in this work.

Multi-level parallelism for the cardiac Bidomain equations

Cardiovascular diseases are associated with high mortality rates in the globe. The development of new drugs, new medical equipment and non-invasive techniques for the heart demand multidisciplinary efforts towards the characterization of cardiac anatomy and function from the molecular to the organ level. Computational modeling has demonstrated to be a useful tool for the investigation and comprehension of the complex biophysical processes that underlie cardiac function. The set of Bidomain equations is currently one of the most complete mathematical models for simulating the electrical activity in cardiac tissue. Unfortunately, large scale simulations, such as those resulting from the discretization of an entire heart, remain a computational challenge. In order to reduce simulation execution times, parallel implementations have traditionally exploited data parallelism via numerical schemes based on domain-decomposition. However, it has been verified that the parallel efficiency of these implementations severely degrades as the number of processors increases. In this work we propose and implement a new parallel algorithm for the solution of cardiac models. By relaxing the coherence of the execution, a new level of parallelism could be identified and exploited: pipelining. A synchronous parallel algorithm that uses both pipelining and data decomposition techniques was implemented and used the MPI library for communication. Numerical tests were performed in two different cluster configurations. Our preliminary results indicated that the proposed algorithm is able to increase the parallel efficiency up to 20% on an 8-core cluster. On a 32-core cluster the multi-level algorithm was 1.7 times faster than the traditional domain decomposition algorithm. In addition, the numerical precision was kept under control (relative errors under 6%) when the relaxed coherence execution was adopted.

Comparing Genetic Algorithms and Newton-Like Methods for the Solution of the History Matching Problem

In this work we presents a comparison of different optimization methods for the automatic history matching problem of reservoir simulation. The history matching process is an inverse problem that searches a set of parameters that minimizes the difference between the model performance and the historical performance of the field. This model validation process is essential and gives credibility to the predictions of the reservoir model. Derivative-based methods are compared to a free-derivative algorithm. In particular, we compare the Quasi-Newton method, non-linear Conjugate-Gradient, Steepest-Descent and a Genetic Algorithm implementation. Several tests are performed and the preliminary results are presented and discussed.

DECoDe - Differential Evolution Algorithm for Community Detection

Community structure of networks, i.e., groups of nodes densely connected inside the same group and weakly connected outside the group, are one of their most important property and there is great interest in the investigation of methods that are able to automatically detect such divisions. This work presents a novel method for community detection based on Differential Evolution, the Differential Evolution Algorithm for Community Detection (DECoDe). Differential evolution is an evolutionary algorithm frequently applied to continuous problems and this work presents a novel approach which adapts the algorithm to discrete problems, allowing the solution of the community detection problem. Several tests were executed with real networks and the results show that the presented approach is able to find consistent community structures, when compared to other methods in the literature.

Performance evaluation of a reservoir simulator on a multi-core cluster

Reservoir simulators are one of the most important tools on reservoir engineering since they allow the prediction of real reservoir’s behavior. However, in order to deal with medium and large scale problems it is necessary to use parallel computing. This work presents the development of a reservoir simulator, based on a two-phase flow model of porous media, and its parallelization. The implementation of the simulator was based on an IMPES scheme and the PETSc library, which uses MPI for data communication between processes, was employed to solve the system of equations. The performance analysis was made in a parallel environment composed by a cluster of multiprocessor computers and the results suggest that the performance of parallel applications strongly depends on the memory contention in multiprocessor computers, such as the quad-cores. Thus, parallel computing should follow certain restrictions regarding the use and mapping of tasks to compute cores.

Automatic history matching in petroleum reservoirs using the TSVD method

History matching is an important inverse problem extensively used to estimate petrophysical properties of an oil reservoir by matching a numerical simulation to the reservoirs history of oil production. In this work, we present a method for the resolution of a history matching problem that aims to estimate the permeability field of a reservoir using the pressure and the flow rate observed in the wells. The reservoir simulation is based on a two-phase incompressible flow model. The method combines the truncated singular value decomposition (TSVD) and the Gauss-Newton algorithms. The number of parameters to estimate depends on how many gridblocks are used to discretize the reservoir. In general, this number is large and the inverse problem is ill-posed. The TSVD method regularizes the problem and decreases considerably the computational effort necessary to solve it. To compute the TSVD we used the Lanczos method combined with numerical implementations of the derivative and of the adjoint formulation of the problem.

Performance Evaluation of Modularity Based Community Detection Algorithms in Large Scale Networks

Community structure detection is one of the major research areas of network science and it is particularly useful for large real networks applications. This work presents a deep study of the most discussed algorithms for community detection based on modularity measure: Newman’s spectral method using a fine-tuning stage and the method of Clauset, Newman, and Moore (CNM) with its variants. The computational complexity of the algorithms is analysed for the development of a high performance code to accelerate the execution of these algorithms without compromising the quality of the results, according to the modularity measure. The implemented code allows the generation of partitions with modularity values consistent with the literature and it overcomes 1 million nodes with Newman’s spectral method. The code was applied to a wide range of real networks and the performances of the algorithms are evaluated.

Analysis Spreading Patterns Generated by Model

Spreading have been studied in networks from a wide range of contexts, such as social, biological and technological. Models for spreading simulation can be applied to real world networks in order to investigate how spreading phenomena occurs from different perspectives. An usual approach is to analyst a diffusion process by assessing the number of reached nodes and the depth of a propagation. This work describes the spreading processes by identifying their patterns, characterized by the canonical name of the propagation trees started by each seeder. Diffusion was investigated in four real world networks considering Independent Cascade Model (ICM). The results show that, as observed in real world scenarios, the occurrence of complex cascades is quite rare and the majority of propagation trees are very simple.

Populational Algorithm for Influence Maximization

Influence maximization is one of the most challenging tasks in network and consists in finding a set of the k seeder nodes which maximize the number of reached nodes, considering a propagation model. This work presents a Genetic Algorithm for influence maximization in networks considering Spreading Activation model for influence propagation. Four strategies for contructing the initial population were explored: a random strategy, a PageRank based strategy and two strategies which considers the community structure and the communities to which the seeders belong. The results show that GA was able to significantly improve the quality of the seeders, increasing the number of reached nodes in about \(25\,\%\).

Investigação Sobre Robustez de Comunidades em Redes

Resumo—Em sistemas complexos modelados como redes, onde os nós representam os indivíduos e as arestas representam os relacionamentos entre os indivíduos, uma das principais questões a serem exploradas é a robustez, ou seja, a capacidade de uma rede suportar ataques a seus elementos. Além disso, o estudo da estrutura de comunidades de redes é bastante importante para a compreensão da topologia da rede em uma organização local. Neste trabalho, estuda-se o efeito da estrutura de comunidades na robustez de redes reais. Para isso, é feita uma investigação comparativa entre o comportamento exibido por redes e suas comunidades correspondentes submetidas a ataques. Os resultados mostram uma forte analogia entre o comportamento exibido nas redes completas e suas comunidades isoladas.