【作者】 魏磊；

【导师】 戴永寿；

【作者基本信息】 中国石油大学 ， 信号与信息处理， 2007， 硕士

【摘要】 准确的地震子波估计对于高分辨率、高信噪比、高保真度的油气勘探数据处理有极为重要的意义。近年来,统计性子波提取技术在实际地震数据处理中得到了广泛应用。本文重点研究了统计性方法中,地震记录准确模型的建立和基于高阶累积量的子波提取方法。实际地震子波可能是非因果和混合相位的,本文从褶积模型出发,分别应用MA(滑动平均)和ARMA(自回归滑动平均)模型对地震记录进行建模,并采用基于高阶累积量的矩阵方程法和拟合优化法进行子波提取研究。仿真和真实数据处理结果表明,ARMA模型在描述地震记录时具有参数节简、模型更为高效的特点;基于高阶统计量的子波提取方法不仅可以保留地震信号中的幅值、频率、相位信息,而且能够有效地消除高斯色噪声,提取出较为准确的子波。本文进一步研究了将矩阵方程法与拟合优化法相结合的子波提取方法。矩阵方程法是一种线性化方法,在地震记录数据长度较短时具有较大的子波估计误差。而在拟合优化法中,初值范围估计的准确度严重影响参数估计的效率。于是可以将矩阵方程法得到的子波参数用于拟合优化法初始解范围的确定,在此基础上用拟合优化法寻找精确解。该方法综合了两种方法的优点,提高了整体的运算效率。此外,本文改进了拟合优化法子波提取中所用到的非线性寻优方法,将免疫遗传算法与小邻域搜索算法结合,提高了寻优效率和精度。通过上述新方法的应用,子波提取准确度和效率得到了显著提高,同时也预示着高阶统计量方法在地震信号处理领域有广阔的应用潜力。更多还原

【Abstract】 The accurate estimation of the seismic wavelet has profound significance for seismic data processing in the sense of high-resolution, high signal-to-noise ratio and high fidelity. Recently, the statistical methods of wavelet extraction achieved comprehensive application in real seismic data processing. The thesis thoroughly studied the accurate model of the seismic trace and the wavelet extraction methods based on high-order statistics.In practice, the real seismic wavelet is noncause and mixed-phase. Based on the convolution model, both the MA (moving average) and ARMA (autoregressive moving average) models were introduced to fit the seismic trace. Then the cumulant-based matrix-equation approach and cumulant matching method were employed to estimate the wavelet and evaluate the applicability of each model. The simulations and real seismic data experiments demonstrate that the ARMA model provides a parsimonious, more efficient signal modeling in fitting seismic trace than the MA model does, and the cumulant-based methods not only retain the amplitude, frequency and phase information of the seismic data, but also effectively eliminate the colored Gaussian noise and obtain a preferable wavelet with high efficiency.Combining the matrix-equations algorithm and the matching algorithm, a new seismic wavelet extraction method was proposed. The cumulant-based matrix-equations algorithm is a linear wavelet extraction method. When the amount of trace data is finite and even insufficient, the estimated wavelet always has remarkable estimated error and variance. As for the cumulant matching method, the veracity of the initial solution range directly influences the optimization efficiency of the method. Therefore, the combined method first extracted a inferior wavelet estimation by the cumulant-based linear-equations method, then relied on it to fix the initial solution range, finally obtained a fine wavelet via the cumulant matching method. The combined method synthetizes the merits of two wavelet extraction methods, thus improves the whole computational efficiency.Moreover, the thesis ameliorated the nonlinear optimization algorithm of the cumulant matching method for seismic wavelet extraction. The combination of The Adaptive Immune-Genetic Algorithm and the small neighbourhood searching algorithm improves the whole computational efficiency and precision.Through the application of the new techniques, the precision and efficiency of the seismic wavelet extraction gain obvious enhancement. The achievements indicate the broad application future of high-order statistics in the field of the seismic data processing.更多还原