Thin Bed Predication Using Spectral Inversion And Inversion Method Analysis

This paper discusses the construction process of objective function in spectrum inversion, and the inversion method is studied. Through the parity component of reflection coefficient analysis, even component can greatly improve the ability to distinguish thin beds, and odd component weakens the ability to distinguish thin beds, using this feature, from the frequency domain convolution theory, the objective function of two beds model is established. then based on the principle of similar, multi-bed model of objective function is established too, through research, I found the objective function is very good convergence and binding, can reduce the inversion of ambiguity.In the inversion method, the simulated annealing, random mountain climbing , Monte Carlo are studied, and analyzes the advantages and disadvantages of the three methods, in order to ensure correctness, I used the Monte Carlo method of inversion. Monte Carlo has high accuracy, but slow. In the process of the inversion ,the simulated annealing algorithm is also reference method, first big step length modified model and then small step length changes.In the trial of spectrum inversion model, based on random search using the Monte Carlo method respectively, the two bed model, multi-bed model, beds of thin bed model, and wedge-shaped model were tested and found spectrum inverse identification can distinguish a thin bed, the inversion result of reflection coefficient is accurate.Then the article discussed the main factors affecting the spectrum inverse, including the noise, wavelet , boundary and Initial model effect, when the model joining noise with the wrong wavelet , the inversion results is bad. According to the analysis of the process gradually, the inversion of seismic data with noise in the high frequency is unstable, should be limited proper bandwidth inversing. In spectrum inversion, the wavelet is hypothesis already known or extraction well ,wavelet is very sensitive to spectrum inversion , poor wavelet the results will also is very poor, so, in the inversion ,improving the quality of wavelet extraction is important, In the boundary, considering the short time Fourier transform of boundary effect, we have to change the rectangular window into arc window, the influence of boundary effect will be reduced. At the beginning of inversion, it needs initial model, good inversion model not only can reduce iterative times, improve efficiency, also can achieve good results.Next, I process actual data of a region by spectrum inversion. We can find the resolution of seismic profiles is improved , and the information of high frequency part also has greatly increased.Finally,paper study the pulse spectrum inversion ,through the given initial pulse reflection coefficient, it continuous iterative modified the size and location of the pulse reflection coefficient, and make it close to the formation of the true reflection coefficient.