自适应波形多道匹配追踪断层识别技术

Fault identification based on multichannel adaptive waveforms matching pursuit

相干体属性技术是识别断层的常见方法,但在应用中也会引起假象且损失有用信息。地震数据可表示成多组局部平面波(连续反射信息)与不连续信息及随机噪声之和,从地震数据中分离出连续反射信息则有利于刻画不连续地质信息。自适应波形多道匹配追踪断层识别技术以常规相干体及倾角属性技术为基础,首先通过相干扫描方式确定同相轴组数、各组同相轴两个方向的视倾角、到达时及延迟时等初始信息;然后建立目标函数,反演出各组连续同相轴波形,因不需预先设定原子库,故反演出的波形物理意义更明确,可视为多道匹配追踪技术。通过实际资料试验,所得连续反射数据体和压制了连续反射体的残差数据体都有助于识别断层,且残差数据体剔除了多组同相轴干涉引起的假象,更好地描述了同相轴扭曲的不连续的线状地质特征,比常规相干体能识别更多细节(如高角度断面波);在连续数据体上提取的相干体比在原始数据体上的更清晰、断层组合关系更合理且噪声更弱。

In oil and gas exploration,coherence cube attributes are most used to identify faults.But this will introduce sometimes a variety of illusions and lost some useful information.In this paper,we propose a new approach of fault identification to avoid the problems.Seismic data can be represented as a sum of multi-group local plane waves(continuous reflection),discontinuous reflections,and random noises.An extraction of continuous reflection information from the sum would enhance discontinuous geological information.First,the coherence scanning is used to obtain initial information,like number of event groups,apparent dips of two directions of each group,arrival time and delay time,etc.Then objective functions are founded,and waveforms of continuous event group are obtained by inversion.Therefore waveforms obtained by the inversion have clearer physical meanings since atom dictionaries would not be set in advance.This approach can be regarded as multichannel matching pursuit.Real data tests demonstrate lots of benefits of the proposed approach.On one hand,since the extraction of continuous reflection information eliminate illusions caused by the interference of event groups,residual data volumes show better discontinuous linear geological features of twist events and more details(large-angle waves from fault surface)than conventional coherence cube.On the other hand,coherence cubesextracted from continuous reflection data volumes are clearer and with less noise than conventional data volumes,more reasonable fault combinations can also be identified.