A Comparative Analysis of Dual Simplex and Memetic Algorithms for Earthquake Location

Earthquake location is a fundamental problem with applications in crustal structure characterization and hydrocarbon exploration. This paper compares two approaches for solving this inverse problem: a memetic algorithm and a dual-simplex-based method. The memetic algorithm combines evolutionary operators-BLX-alpha crossover with adaptive parameter, hybrid adaptive mutation, and tournament selection-with a local refinement step using Weighted Gauss--Newton, applied to 25% of the offspring and accepted only if fitness improves. In parallel, a linearized L1 inversion is formulated using a sensitivity (Jacobian) matrix computed via finite differences, and solved as a linear programming problem using the dual-simplex algorithm. Real seismic data from events recorded by the Seismology Laboratory of UFRN in the Joao Camara region (1988) were used. Results show that the dual-simplex approach converges faster and presents lower computational cost, while the memetic algorithm provides competitive accuracy.