Applications of Ant Colony Optimization in determination of source parameters from total gradient of potential fields
2014 (English)In: Near Surface Geophysics, ISSN 1569-4445, E-ISSN 1873-0604, Vol. 12, no 3, 373-389 p.Article in journal (Refereed) Published
A unified interpretation technique, based on the computation of total gradient from first order horizontal and vertical derivatives of a measured anomaly, namely, magnetic, gravity or self-potential, can be used to determine horizontal location, depth, and the nature of the causative source geometry. The total gradient, in general, may exhibit several symmetrical bell-shaped functions of different magnitudes and with different decay rates beta-termed as 'source geometry factor' to describe the nature of the causative sources. beta-values have been presented for idealized source geometries producing three different types of measurements (anomalies). The non-linear inversion of the total gradient with multiple sources of different geometries is accomplished through Ant Colony Optimization (ACO) - a technique with only a few demonstrated applications in geophysical problems. We have also implemented the ACO which, in its original form, is suitable for discrete space or combinatorial optimization in a continuous model space using decimal coding. Moreover, unlike conventional techniques based on pre-specified moving/fixed data length windows, the proposed technique uses much of the total gradient data points except at the ends of the profile (where error in the total gradient can be large). The presence of singularity in the Euler technique in the analysis of a gravity anomaly over a thick faulted slab has been overcome by the total gradient ACO method. A procedure has also been formulated for analysing gravity and reduced to pole (RTP) total field magnetic anomalies over spherical and vertical cylindrical source geometries. To evaluate the efficacy of the proposed technique, ACO results are compared with other modern techniques, namely, Particle Swarm Optimization (PSO), Enhanced Local Wavenumber (ELW) and the Euler method through simulated data contaminated with random noise and three field examples. This study has revealed that the total gradient ACO technique leads to better convergence towards the global minimum and also better inversion results than those obtained by the PSO and other techniques mentioned above. Unlike the ELW and Euler techniques, the total gradient analysis provides enhanced resolution for interfering sources.
Place, publisher, year, edition, pages
2014. Vol. 12, no 3, 373-389 p.
IdentifiersURN: urn:nbn:se:uu:diva-231136DOI: 10.3997/1873-0604.2013054ISI: 000339624200004OAI: oai:DiVA.org:uu-231136DiVA: diva2:743566