I-Corps: A Deep Learning Toolbox for Subsurface Imaging in Exploration for Oil and Gas

这个i-Corps项目的更广泛的影响/商业潜力是满足美国日益增长的能源消费需求，并保持健康的石油和天然气产量。机器视觉和自动数据拾取是跟上地震勘探数据倍增的可行途径。该产品是第一个以减少地震数据处理中的人工负担为目标的产品，该产品采用了一种新颖而高效的转移学习策略的深度神经网络。如果成功，这种新颖的深度迁移学习方法将为许多其他工业应用采用先进的深度神经网络技术奠定基础。除了石油和天然气行业，该产品还可以用于其他重要领域，如地热能勘探。这个i-Corps项目是为了探索一种产品的市场潜力，该产品可以自动识别嵌入地震数据中的独特模式。该方案的智能优点在于提出了一种新的深度迁移学习方法，该方法通过利用区分无监督特征学习来学习对合成数据和真实数据之间的变化更不变的高层表示，从而减轻了深度神经网络的训练负担，并解决了训练数据短缺的问题。我们的建议的创新之处在于，它旨在建立一个可靠的系统来执行地震数据处理所需的各种模式识别任务，并适应通过地震调查获得的不同数据集。该产品有可能节省高达20%至50%的总地震处理时间。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

The broader impact/commercial potential of this I-Corps project is to meet the increasing demand of energy consumption and maintain a healthy oil and gas output in the US. Machine vision and automatic data picking are viable approaches to keep up with the data multiplication of seismic surveys. The proposed product is the first one that is aiming at minimizing labor burden in seismic data processing by using a deep neural network with a novel and efficient transfer learning strategy. If successful, the novel deep transfer learning approach will build the foundation for adopting advanced deep neural network technologies by many other industrial applications. In addition to the Oil and Gas industry, the product can also be used in other important areas such as geothermal energy exploration.This I-Corps project is to explore the market potential for a product to automatically identify unique patterns embedded in the seismic data. The intellectual merit of this proposal lies in the theme of a novel deep transfer learning approach, which will ease the training burden of the deep neural network and solve the training data shortage problem by utilizing discriminative unsupervised feature learning to learn high-level representations that are more invariant to variations between the synthetic and real data. The innovativeness of our proposal is that it is aimed to build a reliable system to perform various pattern recognition tasks required by the seismic data processing and be adaptive to different datasets acquired through seismic surveys. This product has the potential to save up to 20% to 50% of the total seismic processing time.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.