Receiver Function Inversion Using Genetic Algorithms

We use a genetic algorithm to attack the receiver function inversion problem where the objective is to retrieve shear wave velocities of rock layers as a function of depth. The receiver function inversion is a non–linear, multi–modal, and multi–parameter optimization problem suited for genetic algorithms. We implement a three operator genetic algorithm consisting of selection, single–point crossover, and mutation. We find that a modified selection algorithm based on ”elitism” works better than simple roulette–wheel selection for this problem. The results show that a genetic algorithm based parallel search method can find solutions that can fit the data well. The method is powerful in finding a good subset of optimal models from a randomly chosen population of models sampling various portions of the entire model space. We include a priori information in the inversion to direct the search towards a desired set of solutions. This helps to eliminate many multi–modal solutions.