A robust technique for well-log data inversion

Inverse problems associated with many geophysical measurements are often ill-conditioned, nonunique, and multimodal. Consequently, the gradient-type optimization methods to obtain model parameters become ineffective since the accuracy and convergence of these methods depend highly on initial position and search direction in the parameter space. Evolutionary algorithms which employ direct search global optimization technique are well suited for such problems. In this paper, we apply differential evolution algorithm to invert for the Earth formation properties from measured phase and attenuation by an electromagnetic propagation-type sensor and compare the results with those obtained by conventional method. Results indicate that differential evolution algorithm provides robust results although computationally the method is less efficient.