University of Patras
Προσδιορισμός πεπερασμένων σεισμικών ρηγμάτων από γεωδαιτικές παρατηρήσεις με βάση την αριθμητική-στοχαστική μέθοδο της τοπολογικής αντιστροφής (αλγόριθμος TOPological INVersion, TOPINV)
Abstract
dc:descriptionVarious geophysical processes are described by differential equations which may lead to systems of redundant, highly non-linear systems of ordinary equations. In the latter, constant terms derive from measurements and represent stochastic variables, in addition, no approximate values of the solutions of these equations are available. An example is faults reactivated during earthquakes which, in their near-field, produce permanent ground displacements which may be measured by geodetic techniques, usually GPS. Each fault is approximated by a planar surface in an elastic half-space and is defined by 9 variables.No formal methods to solve such problems exist, and various techniques, mostly optimization methods are adopted. The most frequently used methods are (1) to a priori constrain some variables and optimize only 2 or 3, usually through search techniques and using a simple cost function, the mean weighted misfit. And (2) to use a small sample of points in a search space, so that these points and their misfits can be easily treated as 3-D data, and then optimize the search in the vicinity of points with minimum misfit, i.e. identify a point solution. These techniques may lead to solutions trapped in local minima, to correlated solutions or to solutions with poor error control.To overcome these problems, a numerical-topological, exhaustive grid search-based technique in the Rn space is proposed (n the number of unknown variables). This technique is in fact a generalization and refinement of techniques used in lighthouse positioning and in some cases of low-accuracy 2-D positioning. The basic concept is to identify the optimal n-dimensional compact grid space(s) containing the unknown “true” solution(s). This is possible first by transforming the model space into a discrete n-dimensional Grid G and the observation equations into inequalities on the basis of an optimization parameter k and of their standard errors. Then, starting from an objectively selected value of k and using a Boolean approach all gridpoints of grid G are tested, and those satisfying the observations inequalities are identified. If these points define compact spaces, they represent potential solutions approximating the “true” solution. The process is repeated for gradually smaller values of k until the optimal minimum n-dimensional volume containing the true solution is obtained. From the corresponding optimal set of gridpoints a stochastic solution is computed as the first and second statistical moment of the population of the gridpoints.The overall method has been tested with synthetic data using and accuracy-oriented approach, in which the “true” solution is a priori known, on the basis of the precision, accuracy and compactness of the solution.Also, the method was applied in the 2003 Leucas earthquake, Mw 6.2, for modeling a double seismic fault (i.e. inversion of 18 unknowns) providing unbiased results and consistent with the seismological evidence.
Degree
thesis:*- Grantor dc:publisher
- University of Patras
- Year dc:date
- 2016
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Saltogianni, Vasiliki
- Σαλτογιάννη, Βασιλική
Subjects
dc:subject × 8Rights
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/39435