基于CPU/GPU协同加速叠前逆时偏移方法研究

Research of pre-stack reverse time migration approach based on CPU / GPU synergetic accelerating

为了提高地震资料数据处理的计算效率与成像精度,对于逆时偏移算法,采用一阶速度-应力波动方程,使用交错网络高阶有限差分算法进行求解。利用随机边界条件和PML边界条件进行地震波场正向延拓,既减少波场存储量,又能保证波场数据准确性,同时,使用CPU/GPU协同并行计算技术来提高运算效率。实验结果表明,CPU/GPU协同并行加速计算效果明显,叠前逆时偏移算法可对大地层倾角和偏移孔径清晰成像,能够对复杂构造地质体精确成像。

In order to improve the computational efficiency and imaging accuracy of seismic data processing,this paper applied one order stress-velocity wave equations and staggered grid high order finite difference algorithm in the inverse time migration algorithm. Using Random boundary condition and PML boundary condition to proceed the forward continuation of seismic wave field can not only reduce the storage of wave field,but also ensure the accuracy of wave field data. At the same time,we use CPU / GPU synergetic parallel accelerating to improve computation efficiency. The experimental results show that the pre-stack reverse time migration algorithm can form clear image for big stratigraphic dip and migration aperture and accurate image for complex tectonic geologic body.