Based On Improved Genetic Algorithm For Rayleigh Wave Dispersion Curve Inversion

Rayleigh surface wave exploration has many advantages, such as fast, economical, lightweight, non-destructive, low attenuation, strong anti-interference ability, and it is not affected by the velocity relations of layers. Therefore, it has been widely used in the areas of geophysical exploration, artificial seismic exploration and ultrasonic nondestructive testing, etc. By inversion of Rayleigh wave dispersion curves, we can obtain the following parameters of physical mechanics:shear wave velocity, formation thickness, modulus of deformation, compressive and flexural strength, bearing capacity of subsoil, Poisson's ratio and SPT blow count, etc.Inversion research is the core of Rayleigh wave exploration. Genetic algorithm (GA) is suitable for solving nonlinear, multi-parameter inversion problems of Rayleigh wave dispersion curve because of its less stringent requirements on the initial model, without calculating the derivative and the easy implementation. However, some GA's defect, such as large calculation, inefficient work and premature convergence, make it impeded in the application of Rayleigh wave inversion.In this paper, combined with characteristics of Rayleigh wave inversion, GA was improved as follows to overcome the above shortcomings.(1) The parameter C of Linear calibration was selected dynamicly;(2) Using self-adaptive crossover probability and mutation probability, it can make that the group search towards a good direction more quickly;(3) Reserving one of the same chromosomes and removing others. Some chromosomes with high fitness was selected to mutate for generating new chromosomes which are used to substitute for the removed chromosomes;(4) It can be indicated that there is possibility of reduplicate calculation of fitness by analysing the process of GA. The reduplicate calculation is avoided in programing of GA.The improved genetic algorithm was applied in Rayleigh wave dispersion curves inversion. As for increasing velocity model, fundamental mode dispersion curve was selected to construct objective function. While the model with low-speed sandwich, maximum mode dispersion curve was selected to construct objective function. The results show that the improved genetic algorithm can accelerate the improvement of optimal solution and improve the inversion accuracy efficiently. Finally, using improved genetic algorithm to inverse the measured dispersion curve, we can obtain correct results. It verifies that the improved genetic algorithm is correct and practical.