基于快速标量传递算法的瑞雷波频散曲线反演研究

为了提高近地表瑞雷波频散曲线反演的效率和精度, 引入快速标量传递算法来计算瑞雷波频散曲线正演理论值。通过对比加入线性约束条件前后遗传算法(GA)与模拟退火法(SA)在反演瑞雷波频散曲线中的表现, 提出将计算速度快的蒙特卡洛法(MC)作为辅助手段来快速识别地层类型, 然后在GA和SA中加入线性约束条件来提高收敛速度, 并将GA得到的反演结果作为SA的初始状态, 同时适当地缩小搜索范围, 通过联合反演来克服GA的早熟问题。用上述方法计算和验证三层地层模型、含噪声数据以及工区实际模型, 结果表明该方法高效、准确、稳定性强, 有很强的全局寻优能力, 并具有一定的抗噪能力。

Inversion Research of Rayleigh Wave Dispersion Curve Based on Fast Scalar Transfer Algorithm

In order to improve the efficiency and accuracy of the inversion of Rayleigh wave dispersion curves near the surface, fast scalar transfer algorithm which has the characteristics of high computational efficiency is introduced to calculate the forward theoretical value of Rayleigh wave dispersion curve. The performances of genetic algorithm (GA), simulated annealing algorithm (SA) in the inversion of Rayleigh wave dispersion curves before and after adding linear constraints are compared. On this basis, linear constraints are added to GA and SA to improve the speed of convergence, and Monte Carlo method (MC) with fast computing speed is used to identify the types of formation as a supplementary means. Then the inversion results obtained by GA are taken as the initial state of SA as well as narrowing search scope appropriately, and this kind of joint inversion is carried out to overcome the premature problem of GA. Using the above method to calculate the three-layer model, noisecontaining data and actual model of the work area. The results show that the method above is efficient, accurate and stable, and it has strong ability of global optimization and anti-noise ability to a certain extent.