基于随机反传和筛选模型的微震逆时定位成像

对于常规的逆时定位成像方法,成像结果中强震源的成像值通常远大于并且会掩盖弱震源;同时,成像结果中假象的压制与消除也一直是该技术中颇受关注且比较难解决的问题.对此,本文结合了混合成像条件与高通滤波,从图像对比度的角度加强定位成像效果.提出了反传检波点随机选择的方法,通过重复进行随机选择与随机分组,从而得到不同震源的、包括一些冗余在内的更多信息,通过对信息的融合以提高定位可靠性.提出了筛选模型的概念,将成像过程中各点的波场反传序列引入震源判断标准,构建函数以大致量化震源存在的可能性,结合阈值,构造出由0和1组成"筛选模型",对成像结果进行通过性选择,以此消除假象并提高震源识别的正确性.通过简单模型和复杂模型,验证了本文提出方法的有效性以及对各类干扰因素的适应性与抵抗性.

Time-reverse imaging of source location based on random backward propagation and sifting model

Microseismic events in exploration seismology, mainly caused by hydraulic fracturing, drilling, or fault displacement, have great significance. Space and time coordinates of the events are usually the overriding concern. When the data is of low signal to noise ratio and the first arrival is unable to be accurately picked up, time-reverse imaging could make the microseismic record focus at the source location by backward propagating. However, in conventional time-reverse imaging method, the imaging value of a strong microseismic source is usually large, and will cover up the weak source. Besides, suppression or elimination of artifacts in the imaging result is also a problem hard to solve.
In this paper, we combine hybrid imaging condition with high-pass filter, enhance the imaging effect from the perspective of image contrast. We propose the random selection strategy of receivers in back propagating. The randomly selected receivers in backward propagating are divided into groups randomly (in the case of enough receivers), then the process is repeated several times. Each time corresponds to once conventional time-reverse imaging. It facilitates acquiring more information including much redundance of sources. Reliability increases by fusing the information. Then we propose a concept called sifting model. It introduces the backward propagating sequence to the source judgment standard. Based on this, we construct two discrete functions to value its degrees of dispersion towards zero value and the peak value's time coordinate. Hence the rough possibilities of the existence of a source at a point can be evaluated. Combined with a threshold, a sifting model consisting of 0 and 1 is formed. It is used to eliminate artifacts by passing selection. To demonstrate the effectiveness, experiments using synthetic acoustic data are performed, and we add various interference factors to verify its stability.