Research On Chaos-Genetic Algorithm And Its Application On Seismic Wavelet Estimation

Accurate and effective seismic wavelet estimation has an extreme significance in the seismic data processing of high resolution, high signal-to-noise ratio and high fidelity. The emerging non-liner optimization methods enhance the applied potential for the statistical method of seismic wavelet extraction. Because non-liner optimization algorithms in the seismic wavelet estimation have the defects of low computational efficiency and low precision, Chaos-Genetic Algorithm (CGA) based on the cat mapping is proposed which is applied in the multi-dimensional and multi-modal non-linear optimization.Although Genetic Algorithms(GAs) not only can search the overall parameter globally, but also can search a single parameter deeply. GAs can not maintain the diversity of population when dealing with large-scale and compex optimization problems, thus GAs easily get into the premature and the local optimal. Aimed at the defects of GAs and the characteristics of cumulant matching objective function, the thesis introduced the idea of chaos optimization, and compared the distributed characteristics of logistic mapping, tent mapping with ones of cat mapping. Thus, CGA based on the cat mapping is designed and developed. The algorithm used the initial sensitivity of the cat mapping to expand the scope of the search, and used the ergodicity of the cat mapping to search the chaotic variables. Thereby, reduces the data redundancy, maintains the diversity of population, and solves the problem of local optimum effectively. Theoretical analysis and high dimensional function simulations demonstrate that the algorithm speeds up the evolution of populations, solves the problem of the premature and low accuracy caused by GAs, and has better search efficiency and accuracy.CGA based on the cat mapping was used to optimize the MA(Autoregressive Moving Average) and ARMA(Autoregressive Moving Average) model parameters. Compared the experimental results by CGA with the ones by Adaptive Immune-Genetic Algorithm(AIGA) and Improved Genetic Algorithm, the experimental results show that the wavelet estimated by the algorithm in the thesis is more similar to the real wavelet. The simulations and real seismic data experiments results demonstrate that wavelet estimation using fourth-order matching and CGA is applicable and stable.The thesis combined CGA based on the cat mapping with the model order determination. Experiments show that the parameters of the seimic wavelet are more accurate, the cumulant matching error is small, and the extracted wavelet is highly accurate under the premise of the accurate model order determination. With the effect on the signal-to-noise ratio, data length and other factors, the process of matching method can appear the instance of wavelet model badly-parameterized. Simulation experiments show that CGA based on the cat mapping could extract the wavelet effectively, and the extracted wavelet has good similarity with the ture wavelet. Thus, the wavelet with high precision and high efficiency is estimated by the algorithm.