The shape of information used to execute drilling operations is rugged. A quick survey of all of the information shows data expressed in test reports, images, time series data, Cartesian data, text, spreadsheets, maps, code, and engineering drawings. Traditional database representations of data as tables with rows and columns does not match the new reality of all of the information forms that must be brought to bear on critical drilling decisions. For us to possess an expansive view of all of the information we must move beyond the limitations of relational data into the realm of graphs.
Representation of data in graph form has become popular in recent years, particularly in those domains where the data is rugged. The medical field was the first to recognize the value of graph forms to create cognitive applications like machine-based automated of disease diagnosis. We contend that the rugged nature of drilling operations also lends itself well to graph representation. The objective here is to create something known as a "corpus"—a repository of data in a wide variety of forms that can be traversed with a query language that is purpose-built for graphs. The ultimate aim is to create a Knowledge System that is suitable for natural language queries by humans.
In our talk we will describe how such a corpus is built and how to apply it specifically to drilling operations, with a focus on what works well alongside the challenges and (temporary) failures.
Knowledge Graphs have a unique character and are appropriate for certain classes of applications. We believe that Knowledge Graphs do deserve a place in the oil and gas industry given these attributes that map well to the characteristics of the industry's data forms.