科研之友
微信
新浪微博
Facebook
分享链接
摘要
Random noise attenuation is one of key problems in seismic data processing. The existence of random noise greatly reduces the signal-noise ratio (SNR) of seismic records. Although there are lots of filtering methods available to attain this end, it is inconvenient to select a more appropriate tool for random noise attenuation, of which the characteristics change with the fields of seismic data collection. The understanding of how random noise is generated is the first requirement to solve this problem. We model seismic random noise on land to analyze the characteristics of noise generated by different sources in seismic records. Taking the noise collected in the mountainous region in Southern China for example, the noise sources include natural sources such as wind friction over the ground surface, tree vibrations and rustles caused by wind loads, and cultural sources including running machines, footsteps of people and animals around the geophones and traffic, factories, people's daily lives in the distance. For convenience of computation, it is assumed that all of the sources contribute as point-sources in their designated areas, the function of each kind of noise source is decided according to the corresponding theory, including wind load theory, effect of mountain on wind speed, transverse vibration of beam, aeroacoustics, pseudoharmonic signal and so forth. The noise propagates by wave equation and random noise record is the superposed wave field. The theoretical model of random noise is built, the factors which influence noise characteristics are analyzed in theory, e.g. wind speed, surface roughness, mountain size etc. When the source functions are finalized, all kinds of noise wave-fields can be obtained by solving wave equations. The synthetic records, the single channel waveforms and their frequency spectrums of each kind of noise are shown. The results show that the noise caused by branches and leaves of trees rustle in wind is the major high-frequency component. Seismic random noise is a temporal and spatial random process. As a superposed wave-field, it is composed of vibrograms and wave profiles when the distance or time is a constant. Therefore, the characteristics of the simulated noise record are compared with the real noise in the time domain (vibrogram) and space domain (wave profile), respectively, which include frequency spectrum (wave number spectrum in the space domain), power spectral density, phase locus (only in time domain), mean, variance, kurtosis, skewness, frequency distribution, and cumulative distribution function in the time domain. The comparison results both of vibrogram and wave profiles show the similarity between the simulated noise and the real one. The comparison results demonstrate the feasibility of the proposed method. When a theoretical model of seismic random noise is built, the simulated noise in different data collection regions can be obtained by adjusting the parameters, and noise propagation characters can be analyzed in theory. The simulated noise in the corresponding regions can be used as the background noise instead of white Gaussian noise. A more suitable filtering method and its parameters can be selected and adjusted by analyzing the main component of noise and its mathematical expression.
- 出版日期2015-12
- 单位吉林大学
