论文部分内容阅读
薄层厚度预测一直是公认的难题之一,其难度就在于如何准确地提取薄层的地震属性。传统的方法是利用砂厚和视振幅间的Widess原理,即在调谐厚度之内,视振幅与砂厚呈单调上升关系,而波峰及波谷间的时差变化不大[1];或是利用频率信息[2],其基本原理是随着砂层厚度增大,地震波主频变低。但是理论和实际资料都表明,不同的砂厚和地层组合对地震波的动力学信息影响很大,各种参数与砂厚都是非线性关系,使用单一的信息不可能准确预测薄层厚度。因此我们采用了时域、频域多种参数,利用解决非线性问题的有力武器──神经网络预测薄层厚度。另外,我们还对资料进行了高保真、高信噪比、高分辨率处理,取得了令人满意的效果。
Prediction of sheet thickness has always been one of the accepted problems, the difficulty lies in how to accurately extract the seismic attributes of the sheet. The traditional method is to use the Widess principle between sand thickness and apparent amplitude, that is, within the tuning thickness, the relationship between amplitude and thickness of sand is monotonically increasing while the time difference between the wave crests and troughs is not much changed [1] The basic principle of the information [2] is that as the sand thickness increases, the seismic wave frequency becomes lower. However, both theoretical and practical data indicate that different sand thickness and formation combinations have a great influence on the seismic information of seismic waves, and all parameters are nonlinear with sand thickness. It is impossible to accurately predict the thickness of thin layers using a single piece of information. Therefore, we use a variety of parameters in time and frequency domain, using powerful weapon to solve nonlinear problems ─ ─ neural network prediction of thin layer thickness. In addition, we also carried out high-fidelity data, high signal-noise ratio, high-resolution processing, and achieved satisfactory results.