The Novel Seismic Wavelet Estimation Based on ARMA Model and Chaos Genetic Algorithm
Yongshou Dai, Junling Wang, Weiwei Wang, Shaoshui Wang · 2008
On the assumption that the seismic wavelet is noncausal and mixed phase, ARMA (autoregressive moving average) model was utilized to describe the seismic wavelet, the ARMA cumulant matching method was exploited to estimate the wavelet parameters, and the matching error was proposed to feedback as the evaluation of the matching result. This approach could improve the wavelet estimation precision with high computational efficiency by avoiding the use of cumulant matching method under MA (moving average) model description. But this method also leads to a highly nonlinear optimization problem. To provide effective and reliable optimal solutions and overcome the problem of local minima in the optimization stage, the improved chaos genetic algorithm was introduced in this paper. In this algorithm, the high dimensional chaotic-mapping-three-layer feedback neural network was absorbed in genetic algorithm. Theoretic analysis and numerical simulation demonstrate the feasibility of the wavelet extraction approach. Compared with the normal seismic wavelet extraction, this approach can improve parameter estimation precision efficiently and stably.