提出者：

本文出现在 E&P 时事通讯中。请在此处订阅勘探与生产通讯 。

在数据源方面，石油和天然气钻屑样本历来被忽视。事实上，许多地质样本只是在地质核心储存库中聚集了灰尘。

然而，随着数字化的不断发展，该行业的人士将会注意到岩屑数字化概念背后的动力发生了转变。地球科学界越来越多的人意识到将钻屑数据集成到地下机器学习 (ML) 工作流程中以支持准确和富有洞察力的决策的潜力。 

为了满足充分利用以前未开发的数据源的需求，Earth Science Analytics 最近宣布与地质实验室和服务公司 Rockwash Geodata 合作。通过此次合作，地球科学分析公司将其技术专长与其软件相结合，将大量休眠石油和天然气数据转化为高价值的数字资产。

挑战

虽然视觉切割数据的价值越来越被视为行业信息的重要来源，但迄今为止，利用可用数据仍具有挑战性。岩屑数据始终存在，但需要亲自访问。更糟糕的是，背景化和分析数据所需的工作量和时间导致许多人忽视了插条作为可行的信息来源。 

Rockwash Geodata 的努力为改变这一现状做出了很大的努力。该团队已将成千上万的重要地质资源带入数字环境。感谢 Rockwash Geodata 的努力，现已在全球许多油井的整个地层部分收集了广泛的数字数据点。这些数字可用数据还为计算机驱动的深度学习技术生成了优秀的候选数据集。 

解决方案

此次合作将 Earth Science Analytics 的基于网络的云原生地球科学软件与 Rockwash Geodata 的地质专业知识相结合。Rockwash Geodata 的专家现在可以获取岩屑照片的专有数据库，并使用 Earth Science Analytics 的 ML 工作流程，根据大块岩性对照片进行分类。这将创建一个完全标记的插枝样本照片数据集。

通过合作，这些公司可以实现图像数据的民主化。这确保他们可以更好地对图像进行情境化和可视化，确定趋势以进行预测并支持更好的业务决策。 

通过将这些新创建的数字输入与一组有质量保证的传统测井曲线相结合，地球科学分析可以生成一组高质量的岩石特性预测，这些预测可用于构建更大的储层规模解释。这将通过为岩石特性预测提供另一层理解和信心来支持石油物理和地球科学工作流程。

由于现在已经生成了来自数千个切屑样本的数字图像，机器学习工作流程可以帮助最大限度地发挥其价值。简而言之，此次合作正在帮助运营商最大限度地发挥岩屑数据的价值。

过程

为了从图像中创造价值，我们必须采用以数据为中心或以标签为中心的方法。 

为了提供准确的模型，始终需要行业专家通过标签引入知识。在这种情况下，地球科学分析专家专注于创建一组本体，这些本体的重点是从切割样本中提取重要信息。 

Earth Science Analytics 的流程首先与 Rockwash Geodata 合作，将数字岩屑图像置于上下文中，了解每张图像中存储的信息。通过正确地将所有数据集和类型置于上下文中，这些公司的目标是跨规模传播数据，从小到大。其次，他们对每张图像进行标记，添加人类专业知识并利用地层信息等辅助数据。 

此阶段完成后，地球科学分析将使用计算机视觉方法来训练机器学习模型，帮助支持丰富、快速和大规模的图像解释。 

最后，进行整合，将岩屑与其他测井数据和地震数据一起进行审查。这提供了宝贵的质量控制步骤，并允许地球科学分析传播信息，支持和加速客户的业务决策。

支持高效运营 

地球科学分析公司的新合作伙伴关系意味着这些以前经常未开发的资产可以提供真正的数据价值，帮助地球科学界的人们更广泛地了解地下。 

这种方法确保用户不仅可以确认他们已经知道的事情，还可以了解他们不知道的事情。用户可以看到良好地层的示例，还可以看到可能存在特定的无法解释的钻井问题的示例，从而帮助为许多未解答的问题提供解决方案。 

最终，这将使勘探与生产行业的人员变得更加高效，因为可用的额外信息将降低钻干井的风险。不仅如此，虽然岩屑规模解释工作流程对于石油和天然气勘探中的良好业务决策至关重要，但它们也可以更广泛地应用于地下 CO ₂封存和采矿业。 

这项技术将实现并加速多个领域的数字化转型，地球科学分析公司期待双方的合作，从人工智能数字岩屑数据中创造真正的价值。 

作者简介：  Eirik Larsen 是 Earth Science Analytics 的联合创始人兼首席解决方案官。 

Presented by:

This article appears in the E&P newsletter. Subscribe to the E&P newsletter here.

Oil and gas cuttings samples have traditionally been overlooked when it comes to a data source. Indeed, many geological samples have simply gathered dust in geological core repositories.

However, in line with ongoing digitalization, those in the sector will have noticed a shift in momentum behind the concept of cuttings digitalization. More and more of those in the geoscience community are realizing the potential for integrating cuttings data into subsurface machine learning (ML) workflows to support accurate and insightful decision making. 

To support the appetite to get the most out of what was a previously an untapped data-source, Earth Science Analytics recently announced a collaboration with Rockwash Geodata, a geological laboratory and services company. The partnership will see Earth Science Analytics combine its technical expertise with its software, transforming vast quantities of dormant oil and gas data into high-value digital assets.

The challenge

While the value of visual cuttings data is increasingly being viewed as an important source of industry information, utilizing the available data has so far proved challenging. The cuttings data were always there, but access required physical visitation. To make matters worse, the workload and time required to contextualize and analyze the data led many to overlook the cuttings as a viable source of information. 

Rockwash Geodata’s efforts have gone a long way to change this. The team has brought thousands upon thousands of important geological resources into the digital environment. An extensive collection of digital datapoints throughout the entire stratigraphic section of many wells worldwide now exists thanks to Rockwash Geodata’s efforts. This digitally available data have also generated an excellent candidate dataset for computer-driven deep learning techniques. 

The solution

The collaboration sees the amalgamation of Earth Science Analytics' web-based cloud-native geoscience software and Rockwash Geodata’s geological expertise. Rockwash Geodata’s experts can now take a proprietary database of cuttings photographs and, using ML workflows from Earth Science Analytics, categorize the photos in terms of bulk lithology. This creates a fully labeled dataset of cuttings sample photographs.

By working together, the companies can democratize the image data. This ensures they can better contextualize and visualize the images, establishing trends to make predictions and support better business decision making. 

By combining these newly created digital inputs with a quality assured set of traditional log suite curves, Earth Science Analytics can generate a set of high-quality rock property predictions, which can be used to build larger, reservoir-scale interpretations. This will support petrophysical and geoscience workflows by providing another layer of understanding and confidence to the rock property predictions.

And as these digital images from thousands of cuttings samples have now been produced, the ML workflows can help to maximize their value. Put simply, the collaboration is helping operators to maximize the value from cuttings data.

The process

To create value from an image, we must embrace a data-centric or label-centric approach. 

The introduction of knowledge by industry experts via labeling has always been required in the quest to deliver accurate models. In this case, Earth Science Analytics experts focus on creating a collection of ontologies that are focused on extracting the important information from the cutting samples. 

Earth Science Analytics' process begins by working with Rockwash Geodata to contextualize the digital cuttings images, understanding the information that is stored in each one. By properly contextualizing all datasets and types, the companies aim to propagate data across scales, from the very small to the very large.  Secondly, they label each image, adding human expertise and leveraging auxiliary data such as stratigraphic information. 

Once this stage is complete, Earth Science Analytics uses computer vision methods to train ML models, helping to support rich, fast and large-scale interpretation of the images. 

Finally, integration takes place, as cuttings are reviewed alongside other well-log data and seismic data. This provides valuable QC steps and allows Earth Science Analytics to propagate information, supporting and accelerating the business decisions of its clients.

Supporting efficient operations 

Earth Science Analytics' new partnership means these previously often untapped assets can provide real data value, helping those in the geoscience community get a broader picture of the subsurface. 

This approach ensures users can not only confirm the things they already know, but also cast a light on what they don’t. Users can see examples of good formations, but also where there may have been a particular unexplained drilling problem, helping provide the solution to many unanswered questions. 

Ultimately, this will allow those in the E&P sector to become more efficient, as the extra information available will reduce the risk of drilling dry wells. Not only that, but while cuttings-scale interpretation workflows are essential for good business decision-making in oil and gas exploration, they can also be used wider for sub-surface CO₂ storage and mining industries. 

This technology will enable and accelerate digital transformation across multiple sectors, and Earth Science Analytics looks forward to its collaboration to create real value from AI-ready digital cuttings data. 

About the author: Eirik Larsen is co-founder and chief solutions officer of Earth Science Analytics.