In this paper we make the case that data science captures value in well construction when data analysis methods, such as machine learning, are underpinned by first principles derived from physics and engineering and supported by deep domain expertise.

Despite receiving wide attention in recent years, many organizations currently struggle to derive value from their data science efforts. In our experience, disappointment arises for a multitude of reasons, which we discuss in detail. Key issues that often hinder value capture include poor data management, challenges in working with WITSML data, lack of well construction domain expertise by data science teams, inadequate use of physics and engineering and failure to adopt data science solutions into existing or new well construction workflows.

Although by no means comprehensive, we provide a summary of important data that pertains to the well construction process. We further discuss high-level areas where data science can add value to well construction through analysis of such data. Data science initiatives typically fit within at least one of the following categories: Historical Studies, Well Planning, Real-Time Well Construction Execution and Post-Drill Learning Capture. Historical studies are often good places for data science teams to initially focus their efforts. However, as insights are drawn and potential for value is shown, organizations should consider extending capabilities developed to carry-out historical studies to support well planning and real-time well construction execution workflows.

A large portion of this paper is dedicated to discussing ways that organizations can work to improve their abilities to derive value from data science efforts. Most of the discussion focuses on steps that data science teams can take today. However, our commentary on data management and governance is more forward looking. Important topics which we cover include:

  • Data management and governance.

  • Serving data to data scientists.

  • Working with WITSML data.

  • Basic skills and technologies needed by data science teams.

  • Importance of building common capabilities for working with data.

  • Need for physics and engineering to inform data analysis.

  • Importance of identifying data quality issues.

  • Importance of activity-based data filtering when working with WITSML data.

  • Dysfunction detection using WITSML data.

  • Application of statistics and machine learning.

We conclude by examining several historical data science case studies for well construction. Each example highlights the need to connect data and some physical or engineering process (i.e., "engineering with data") to deliver value through data science.

You can access this article if you purchase or spend a download.