Inversion Of Dispersion Curve Based On Secular Function Set And Improved Particle Swarm Optimization Algorithm

Surface wave exploration is widely used in energy exploration,geological exploration and other fields due to its economic,fast and environmentally non-destructive advantages.The core link is to extract the dispersion curve from the seismic record and invert it to image the underground structure.However,when the stratum medium is relatively complex,such as the existence of underground cavities or silt layers,the dispersion curve often exhibits phenomena such as "mode-kissing" and "mode-missing",which makes it difficult to make correct mode discrimination.The misjudgment of the traditional inversion method and the insufficient performance of the optimization algorithm are the main reasons for the error of inversion result.The misfit function of inversion and the optimization algorithm are studied for the two types of problems.Based on the secular function set,a new misfit function applied to the inversion of Rayleigh wave’s multi-mode dispersion curve is proposed.Firstly,the extreme value distribution of the secular misfit function is explored through the trial calculation of the two-layer model.And then theoretical seismic records of active source are synthesized by the three-layer stratum model,the four-layer model with low-velocity layer.The theoretical seismic records and the measured data of Shenzhen,Guangdong,and Dongguan have been processed to verify the effectiveness of the misfit function.The results show that the each mode obtained by the inversion are automatically matched and have a high degree of fit,so as to obtain an accurate shear wave velocity structure.The misfit function can effectively solve the mode misjudgment problem of multi-order mode joint inversion.However,the convergence is between the traditional misfit function and the determinant misfit function with multiple local extrema,and the function surface continuity is poor.Inversion effect can be improved by combining a high-performance optimization algorithm.The particle swarm optimization is improved by introducing the local linearization method,and the particle swarm optimization with gradient(PSOG)is proposed.Firstly,the performance of the algorithm is tested by a typical multi-extreme test function,and then combined with the new misfit function to process the synthesized ambient seismic noise.The results of multiple inversions are compared,which show that the PSOG algorithm has both search performance and convergence speed.Compared with the particle swarm optimization,the accuracy and stability have been significantly improved.Finally,the ambient seismic noise of the Suzhou region was processed,and the weak layer caused by the cutting of the ancient river is successfully imaged.In order to further improve the convergence of the new misfit function,a joint inversion misfit function of Rayleigh wave and Love wave dispersion curves is proposed based on the secular function set.Firstly,a two-layer model is used to explore the convergence of the joint inversion misfit function,and then the inversion effect of synthesized ambient seismic noise is compared to verify the superiority of the joint inversion.The results show that the secular misfit function corresponding to Love wave has significant difference in parameter sensitivity compared with Rayleigh wave.The secular misfit function of joint inversion has better convergence.Joint inversion can obtain more accurate results than inversion of Rayleigh wave.Finally,the ambient seismic noise measured in Gaozhou were processed.The location of geothermal resources was successfully predicted.