井间层析成像的平滑SIRT算法

SMOOTHING SIRT ARITHMETIC FOR CROSS-WELL TRAVEL-TIME TOMOGRAPHY

井间层析成像是一个非线性反演问题 ,为了保证反演过程的稳定性 ,并提高迭代的收敛速度 ,对常规的SIRT算法进行了改进 ,即引入了平滑算子对梯度场进行动态平滑 ,并且通过线性搜索确定速度更新的步长。将这种改进的算法称为平滑SIRT算法。模型试算的结果证明 ,该方法提高了迭代的收敛速度 ,而且反演结果不受初始模型的影响 。

In order to assure the stability of inversion and enhance the convergence speed of iteration, the traditional SIRT arithmetic is improved. A smoothing operator is introduced, and the gradient field is smoothed dynamicly. The step of velocity updating is determined by the linear searching method. The improved method is called after smoothing SIRT arithmetic. The inversion result of synthetic cross\|well data shows that the convergence speed has been enhanced by this method. The inversion result is robust with the different initial model, which makes the stability of the inversion improved.