Simulated Annealing for Airborne EM Inversion

P014 Simulated Annealing for Airborne EM Inversion C. Yin* (Fugro Airborne Surveys) & G. Hodges (Fugro Airborne Surveys) SUMMARY The traditional algorithms for AEM inversion take the negative gradient of an objective function as searching direction and run downhill search. Such algorithms are strongly dependent of starting model and are easily trapped in local minima. Simulated annealing exploits the analogy between the annealing process of a melted metal cooling and freezing into crystal of minimum energy and search for a minimum in mathematical optimization. It employs a random search that accepts both model updates decreasing and model updates increasing system