【摘要】：在地震數(shù)據(jù)處理中,通常受到復(fù)雜因素的影響其在空間方向上的采樣是不規(guī)則的,而這種不規(guī)則的采樣對(duì)后續(xù)的地震資料處理影響嚴(yán)重。數(shù)據(jù)規(guī)則化的作用是通過在不規(guī)則采樣網(wǎng)格中,使用估計(jì)空間頻率的傅里葉方法進(jìn)行實(shí)現(xiàn)。在一個(gè)不規(guī)則坐標(biāo)網(wǎng)格中,傅里葉基函數(shù)的非正交性導(dǎo)致“譜泄漏”的問題:能量從一個(gè)傅里葉系數(shù)泄漏到另一個(gè)傅里葉系數(shù)中。本文研究了基于傅立葉變換的地震數(shù)據(jù)規(guī)則化的反泄漏算法,通過遞歸減法的ALFT算法消除了不規(guī)則地震數(shù)據(jù)所引發(fā)的頻譜泄漏現(xiàn)象,可實(shí)現(xiàn)精準(zhǔn)分析相對(duì)應(yīng)的傅里葉系數(shù),進(jìn)而實(shí)現(xiàn)不規(guī)則網(wǎng)格數(shù)據(jù)處理和規(guī)則重建。本文采用區(qū)域加權(quán)方案準(zhǔn)確估計(jì)傅里葉權(quán)值;應(yīng)用頻率域過采樣技術(shù)消除邊界吉布斯現(xiàn)象;利用非規(guī)則快速傅里葉變換代替?zhèn)鹘y(tǒng)的非規(guī)則離散傅里葉變換降低運(yùn)算成本;應(yīng)用分塊策略將大數(shù)據(jù)分塊處理提高運(yùn)行速率;應(yīng)用分布式并行策略用以存儲(chǔ)及處理超大型數(shù)據(jù)。在本文中,介紹了基于CUDA架構(gòu)的GPU/CPU并行加速技術(shù),針對(duì)較高的輕量計(jì)算任務(wù)如矩陣乘法部分采用GPU算法進(jìn)行改進(jìn),改進(jìn)后算法并行度高,處理速度快。在反泄漏傅里葉變換算法中,通過將非規(guī)則傅里葉變換運(yùn)算部分按照傅里葉因子與輸入數(shù)據(jù)的關(guān)系拆分成矩陣乘法形式進(jìn)行表達(dá),并將拆分后的矩陣乘法部分傳入GPU端進(jìn)行處理;在GPU端應(yīng)用share memory對(duì)數(shù)據(jù)處理進(jìn)行進(jìn)一步的加速;應(yīng)用reduction改進(jìn)算法在GPU中求取最大值。多種改進(jìn)以及優(yōu)化技術(shù)應(yīng)用在反泄漏傅里葉變換方法中,很大程度上提高算法處理效率以及結(jié)果模型的精度,通過GPU端相應(yīng)優(yōu)化,其算法的加速比可以達(dá)到76倍以上。理論模型和實(shí)際數(shù)據(jù)處理結(jié)果可以滿足工業(yè)化的需求,驗(yàn)證了方法的有效性和合理性。
[Abstract]:In seismic data processing, the sampling in spatial direction is irregular under the influence of complex factors, and this kind of irregular sampling has a serious effect on the subsequent seismic data processing. The function of data regularization is realized by using Fourier method to estimate spatial frequency in irregular sampling grid. In an irregular coordinate grid, the nonorthogonality of the Fourier basis function leads to the problem of "spectral leakage": energy leaks from one Fourier coefficient to another. In this paper, a regularized anti-leakage algorithm of seismic data based on Fourier transform is studied. The frequency spectrum leakage caused by irregular seismic data is eliminated by recursive subtraction ALFT algorithm, and the corresponding Fourier coefficients can be accurately analyzed. Furthermore, irregular grid data processing and rule reconstruction are realized. In this paper, the region weighting scheme is used to accurately estimate the Fourier weight, the frequency domain oversampling technique is used to eliminate the boundary Gibbs phenomenon, the irregular fast Fourier transform is used to replace the traditional irregular discrete Fourier transform, and the operation cost is reduced. The block processing of big data is applied to improve the running rate, and the distributed parallel strategy is used to store and process super large data. In this paper, the GPU/CPU parallel acceleration technology based on CUDA architecture is introduced. The GPU algorithm is used to improve the high lightweight computing tasks such as matrix multiplication. The improved algorithm has a high degree of parallelism and a fast processing speed. In the anti-leakage Fourier transform algorithm, the irregular Fourier transform is expressed in matrix multiplication form according to the relation between the Fourier factor and the input data. And the split matrix multiplication part is passed into the GPU terminal for processing; The data processing is further accelerated by using share memory at the GPU end, and the maximum value is obtained by using the improved reduction algorithm in GPU. Many improved and optimized techniques are applied to the anti-leakage Fourier transform method, which greatly improves the processing efficiency of the algorithm and the precision of the result model. The speedup ratio of the algorithm can reach more than 76 times through the corresponding optimization of the GPU terminal. The theoretical model and the actual data processing results can meet the needs of industrialization and verify the validity and rationality of the method.
【學(xué)位授予單位】：東北石油大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2017
【分類號(hào)】：P631.44

【參考文獻(xiàn)】

相關(guān)期刊論文 前8條

1 石穎;王維紅;李瑩;井洪亮;;基于波動(dòng)方程三維表面多次波預(yù)測(cè)方法研究[J];地球物理學(xué)報(bào);2013年06期

2 高建軍;陳小宏;王芳芳;馬劍;;不規(guī)則地震道數(shù)據(jù)規(guī)則化重建方法研究[J];地球物理學(xué)進(jìn)展;2011年03期

3 高建軍;陳小宏;李景葉;;三維不規(guī)則地震數(shù)據(jù)重建方法[J];石油地球物理勘探;2011年01期

4 石穎;劉洪;鄒振;;基于波動(dòng)方程表面多次波預(yù)測(cè)與自適應(yīng)相減方法研究[J];地球物理學(xué)報(bào);2010年07期

5 孟小紅;郭良輝;張致付;李淑玲;周建軍;;基于非均勻快速傅里葉變換的最小二乘反演地震數(shù)據(jù)重建[J];地球物理學(xué)報(bào);2008年01期

6 孟小紅;劉國峰;周建軍;;大間距地震數(shù)據(jù)重建方法研究[J];地球物理學(xué)進(jìn)展;2006年03期

7 劉喜武,劉洪,年靜波;非均勻地震數(shù)據(jù)重建方法及其應(yīng)用[J];石油物探;2004年05期

8 劉喜武,劉洪,劉彬;反假頻非均勻地震數(shù)據(jù)重建方法研究[J];地球物理學(xué)報(bào);2004年02期

，

本文編號(hào)：2408182