Shear velocity log is an important input during the reservoir characterization. It is widely used during seismic reservoir characterization such as lithology, porosity and fluid estimation, four-dimensional seismic studies, geomechanical and wellbore stability studies. Shear velocity log is not necessarily acquired in all wells due to the costs. Therefore, different methods have been developed to estimate it using other logs. Methods such as empirical relations, multi-regression analysis, artificial neural networks, and rock physic modelling have been widely used for this purpose. In this study, the shear velocity log is estimated using empirical relations, multi-regression analysis, and artificial neural networks in one of the gas reservoirs in south of Iran. In addition, utilizing rock physics models, an equation is developed for this purpose. For empirical relations, Castagna and Brocher's relations were employed. Empirical relations estimated the shear velocity log with significant error. Multi-regression analysis was carried out employing three logs: Compressional velocity, density and porosity. In the artificial neural network, the training is performed by the Levenberg-Marquardt algorithm with four input layers, ten hidden layers and one target layer. The outcome of multi-regression and ANN estimations are compared well with the measured shear velocity log. Finally, an equation is developed using rock physics concepts to estimate the shear velocity log. The linear form of Gassmann's relation is used on the first step. The small and negligible terms are eliminated to simplify the equation. This equation requires compressional velocity and density logs. It also needs bulk modulus of rock forming mineral and pore filling fluids as well as porosity that are commonly known in oil and gas reservoirs. Utilizing this equation, shear velocity log is estimated in our case study. The average difference between the measured and estimated shear velocity log using our method is 10 m/s and 0.33%, respectively. Since, this equation considers the impact of mineral and fluids, we predict that it could be applicable in most of geological environments.

剪切速度测井是储层表征过程中的重要输入。它被广泛用于地震储层表征，如岩性，孔隙度和流体估算，四维地震研究，岩土力学和井眼稳定性研究。由于成本原因，不一定在所有井中都获得剪切速度测井曲线。因此，已经开发出不同的方法来使用其他日志进行估计。经验关系，多元回归分析，人工神经网络和岩石物理建模等方法已被广泛用于此目的。在这项研究中，使用经验关系，多元回归分析和人工神经网络估算了伊朗南部一个气藏中的剪切速度测井。此外，利用岩石物理学模型，为此目的开发了一个方程。对于经验关系，使用了Castagna和Brocher的关系。经验关系估计剪切速度测井有明显误差。使用三个测井曲线进行多元回归分析：压缩速度，密度和孔隙率。在人工神经网络中，训练是通过Levenberg-Marquardt算法执行的，该算法具有四个输入层，十个隐藏层和一个目标层。将多元回归和ANN估计的结果与测得的剪切速度测井结果进行了很好的比较。最后，使用岩石物理学概念开发了一个方程，以估算剪切速度测井。第一步使用加斯曼关系的线性形式。消除了小的和可忽略的项以简化方程式。该方程式需要压缩速度和密度测井曲线。它还需要岩石形成的矿物和孔隙填充液的体积模量以及油气储层中通常已知的孔隙度。利用这个方程，在我们的案例研究中估计了剪切速度测井。使用我们的方法测得的和估计的剪切速度测井之间的平均差分别为10 m / s和0.33％。由于该方程式考虑了矿物和流体的影响，因此我们预测该方程式可能适用于大多数地质环境。