Inversion of Seismic Prospecting Parameters using Improved Genetic Algorithms

Most compute of geophysical inversions appear as objective function optimization with multi-peak, it is difficult that we apply the nonlinear inversion theory to seismic inversion. This study has considerable academic value as well as wide prospects. This paper deeply investigates Genetic Algorithm on the subject. With fully absorbing and digesting research achievements of domestic and overseas scholars about nonlinear optimization technique, the paper puts forward several improved measures as follows: A. Good individual of parent participate the competition of filial generation in the operation of cross and mutation. B. The improved algorithm uses the parameter of the model digit's variance to adjust the probability of mutation. It improves the performance of the algorithm by adjusting the probability of mutation. C. Preserving the best individual, the global optimal solution would be obtained as soon as possible, which can verify the convergence of the algorithm. D. For overcoming cheated problem of genetic algorithm, the paper introduces Boltzmann existent mechanism in genetic operating so that bad offspring also can be accepted, which keeps variety of the population to avoid the loss of the optimal solution. E. The paper sums up diverse conversion system from binary system to decimal system, which may compute the same example using different conversion system and different precision, till the satisfactory solution is obtained. This paper tries to find the appropriate coding system that pays attention to the short length of chromosome (It needs smaller memory space) as well as the high precision in seismic inversion. A large number of examples and experimental analysis verify the improved algorithm is effective and practical.