Inversion of seismic refraction data using genetic algorithms

Abstract

The use of genetic algorithms in geophysical inverse problems is a relatively recent development and offers many advantages in dealing with the nonlinearity inherent in such applications. However, in their application to specific problems, as with all algorithms, problems of implementation arise. After extensive numerical tests, we implemented a genetic algorithm to efficiently invert several sets of synthetic seismic refraction data. In particular, we aimed at overcoming one of the main problems in the application of genetic algorithms to geophysical problems: i.e., high dimensionality. The addition of a pseudo-subspace method to the genetic algorithm, whereby the complexity and dimensionality of a problem is progressively increased during the inversion, improves the convergence of the process. The method allows the region of the solution space containing the global minimum to be quickly found. The use of local optimization methods at the last stage of the search further improves the quality of the inversion. The genetic algorithm has been tested on a field data set to determine the structure and base of the weathered layer (regolith) overlaying a basement of granite and greenstones in an Archaean terrain of Western Australia.

What if your Inversion has no Numerical Target?

“…The GA modifies the solution set by applying parameter swapping between highly-ranked models, generating new sets of models that progressively converge towards the target. A general overview of GAs is provided by Goldberg (1989), and details specific to our inversion method can be found in Boschetti et al (1996). As in biological evolution, an element of randomness exists in the generation of new models, so that unexpected results may suggest new possibilities outside the experience or expectation of the geologist.…”

Self-Potential Inversion Using Genetic Algorithm

“…Over the last decade, genetic algorithms (GAs) have been extensively used as search and optimization tools in various problem domains (Wilson et al, 1994). The primary reasons for their success are their broad applicability, ease of use and global perspective (Goldberg, 1989;Billings et al, 1994;Boschetti et al, 1996;Stoffa andSen, 1991 andBoschetti et al,1997). Genetic algorithms mimic natural selection and biological evolution to achieve their power, and their operational characteristics are typically analogous to the evolution theory.…”

Interactive modelling for natural resource management

“…GAs are today an established technique, with wide range of applications to both theoretical and industrial problems. We refer the reader to (Davis, 1991) for basic description of genetic algorithms and to (Boschetti et al, 1996), for a more detailed description of the specific GAs implementation used in this work. The mathematically oriented reader can also refer to (Boschetti and Moresi, 2001) for a discussion on the implications of subjective evaluation in both the search space landscape and convergence speed of a GA.…”

“…Secondly, some noninteractive GA implementations naturally process the ranking of the solutions fitness rather than their exact numerical values. This is done in order to reduce the probability that the algorithm converging rapidly to a local minimum (see (Boschetti et al, 1996) for a technical discussion on this topic). Employing the user ranking of the models' output in the Evaluation GUI thus allows use of a standard GA algorithm without any need to pre-process the subjective evaluation.…”