Automatic Seismic to Well Tie

Author: Valentin Tschannen - Fraunhofer ITWM, Germany.

This package aims to provide basic utilities to automate the seismic to well tie process. 该软件包旨在提供基本工具，以自动化地震到井的标定过程。

Part of the package is dedicated to the creation and training of a neural network to perform wavelet extraction. It also contains utilities to create synthetic training data. 部分包致力于创建和训练神经网络以执行子波提取。它还包含用于创建合成训练数据的实用程序。

The other modules are dedicated to the remaining operations (log processing, depth to time conversion), and use random search and bayesian optimization for the automatic tuning of hyper-parameters. 其他模块则专注于其余操作（测井处理、深度时间转换），并使用随机搜索和贝叶斯优化来自动调整超参数。

Package content

• README.md, Makefile, setup.py, environment.yml, main.sh
• scripts : shell scripts for the installation 用于安装的 shell 脚本
• tests : unit/intergration tests 单元/集成测试
• documentation : documentation built during installation 安装过程中构建的文档
• notebooks : jupyter notebooks demonstrating the use of the library 演示库的用法的 jupyter notebook
• experiments : config files to train a neural network and folder where results are saved 用于训练神经网络的配置文件和保存结果的文件夹
• wtie : python package

Installation 安装

Linux

(TODO: docker) Python dependencies are listed in the file enivronment_linux.yml. Assuming that you installed the Anaconda platform and that your system's os is unix-like with the tool make, you can install the wtie package by running the following command in a shell console in your base environment (alternatively, take a look at the Makefile to see the steps that you need to reproduce to install on your system): Python 依赖项列在文件 enivronment_linux.yml 中。假设您已经安装了 Anaconda 平台，并且您的系统是类 unix 操作系统并带有 make 工具，您可以在您的 base 环境下的 shell 控制台 中运行以下命令来安装 wtie 包（或者，您可以查看 Makefile 文件以了解在您的系统上需要重现的安装步骤）：

make install

This will create a conda environment named wtie and install the package (this may take several minutes, especially if you are going through a network adapter). 这将创建一个名为 wtie 的 conda 环境并安装该包（这可能需要几分钟，特别是如果您通过网络适配器进行安装）。 If there were no problems, you are ready to work with the package! 如果没有问题，您就可以使用该包了！ Don't forget to activate the environment before working with the package: 在使用该包之前，不要忘记激活环境：

conda activate wtie

Windows

(TODO: better Windows support) Python dependencies are listed in the file enivronment_windows.yml. For this guide I assume that you installed the Anaconda platform. Go to the main well-tie directory and go through the follwoing steps: Python 依赖项列在文件 enivronment_windows.yml 中。本指南假设您已经安装了 Anaconda 平台。转到主 well-tie 目录并执行以下步骤： First, open an Anaconda prompt and create a new conda environement: 首先，打开一个 Anaconda prompt 并创建一个新的 conda 环境：

conda env create -f environment_windows.yml

This will create a conda environment named wtie. 这将创建一个名为 wtie 的 conda 环境。 Second, open a Windows command prompt and activate the environment: 其次，打开一个 Windows 命令提示符并激活环境：

conda activate wtie

Then, in the same prompt, install the python wtie package: 然后，在同一个提示符中，安装 python wtie 包：

python setup.py develop

Unfortunatly, Windows does not seem to be able to streamline the installation of the noise package that we use to generate correlated synthetic reflectivity series. To install this package you should follow instructions given here. 不幸的是，Windows 似乎无法简化我们用来生成相关合成反射系数序列的 noise 包的安装。要安装此包，您应遵循此处给出的说明。

Finally, from the same prompt, move to the tests folder and run the test suite: 最后，在同一个提示符中，移动到 tests 文件夹并运行测试套件：

pytest -v --basetemp="./tmp"

Don't forget to activate the environment everytime you want to work with the package: 每次想要使用该包时，都不要忘记激活环境：

conda activate wtie

Tutorial 教程

A series of notebooks demonstrates the use of the library to perform an automatic well tie. The weights of a pretrained network as well as data coming from the open Volve and Poseidon datasets are provided in the data folder. NOTE that both datasets come with their own license. 一系列 notebooks 演示了如何使用该库执行自动井标定。预训练网络的权重以及来自开放的 Volve 和 Poseidon 数据集的数据在 data 文件夹中提供。请注意，这两个数据集都有其自己的许可证。

Explanations about how to train your own neural network are given here. 关于如何训练您自己的神经网络的说明在这里给出。

Publications 出版物

A publication titled "Partial automation of the seismic to well tie with deep learning and Bayesian optimization" is available in open access in the journal Computers and Geosciences. It can be accessed via https://doi.org/10.1016/j.cageo.2022.105120 . You are free to download and modify the source code for your work. If you find this work useful, don't forget to cite our publication. 一篇题为“Partial automation of the seismic to well tie with deep learning and Bayesian optimization”的出版物可在《Computers and Geosciences》期刊上开放获取。可通过 https://doi.org/10.1016/j.cageo.2022.105120 访问。您可以自由下载和修改源代码用于您的工作。如果您觉得这项工作有用，请不要忘记引用我们的出版物。