Knowledge Graphs: Bridging Oil & Gas Data Silos to Improve Decision Making

This article was co-authored by Ramakrishna Manne and Sachin Padhye and supported by Eric Regel.

Data Silos

The oil and gas sector is now generating data at a greater volume than ever before. A typical offshore platform generates 1-2 TB of data every day. Data is expanding not only in quantity but also in variety given the increasing number of in-field equipment sensors. It is commonly stored across disparate databases, servers, content management systems, wikis, data dictionaries, glossaries, and even emails. The complexity of this data ecosystem presents challenges to finding and consuming appropriate and accurate data using traditional methods. This in turn makes it hard to generate insights essential to transforming business and operations. 

Relational Databases versus Graph Databases

Traditional methods of using relational databases have worked well so far for Oil & Gas companies. However, they are falling short of driving key insights given the rapid expansion of volume and variety of internal and external data. A technique known as “knowledge graphs” is a new way to gather better insights faster and more economically compared to traditional methods. 

Knowledge graphs use the concept of “Ontology” which is a semantic generalized data model that defines types of things (classes) along with their properties (attributes) and links properties to other classes. When plotted, it represents the building blocks or model of a graph database with the classes and properties acting as nodes and the relationships acting as links. 

Business Value

Knowledge Graphs provide the ability to gain faster insights to improve productivity, minimize risks, and increase uptime. This in turn has the potential to increase NPV and reduce OPEX. Geosoft (acquired by Seequent in 2019), a geophysical and geological software and services company, reported in their survey “Exploration Information Management report” about 20% – 40% potential improvement in productivity of those who respond to requests for data. Survey results showed that 74% of responders spend more than 10% of their time managing data with 34% spending between 10%-20%, 40% spending more than 20%, and 20% spending more than 30% of their time on managing data.

A) Superior Performance

Knowledge Graphs outperform relational databases in developing insights (not known previously) faster and more economically especially as the complexity of queries increases. They can find links across various domains, functions, and regions. These links help drive insights effectively by reducing the need to reproduce or rebuild data. These links allow knowledge graphs to propagate and further increase their ability to discover insights.

B) Increased Flexibility

Knowledge graphs are not limited by time and size as relational databases are. They can be integrated throughout the timeline of a project as opposed to relational databases which can only integrate specific and predefined data sources. They enable the reuse of data immediately and effectively. This eliminates the cost of movement, transformation, and/or recreation of more data to make it useful. With knowledge graphs, there is no need to create “Shadow IT” systems.

Three Pillars of Success

While Knowledge graphs offer several benefits, they take diligent planning and development of the right capabilities to achieve success. The right platform with relevant tolls is necessary to develop them. Three pillars that are vital to delivering successfully are:

1. Enabling automated quality growth,
2. High accuracy,
3. Proper governance and change management. 

Applications

Knowledge graphs can take the “control of work” or “permit-to-work” processes to the next level. This is because of the early identification of hidden risks coupled with insights discovery that can potentially prevent or at least mitigate these risks based on data from past incident reports, lessons learned, completed permits, after-action reviews, MOCs, and regulatory requirements. Currently, without the use of knowledge graphs, relevant data is usually stored in a dozen different systems, which might or might not be linked through traditional static methods. Ascertaining hidden risks associated with a specific type of equipment, tool, manufacturer, location, elevation, operating condition, time, personnel, etc. is very challenging given the current data silos unless another independent system is created from scratch. Knowledge graphs will cover all these systems and databases defined above. Moreover, all linked data will be available directly within the permit-to-work system itself, as the operator completes the permit form online. Consequently, all new insights can be fed early during the scheduling/planning phase of the work order. The new hazard/ risk assessment augmented with fresh insights can in turn reduce millions of dollars in incident costs.