Allowing computers to do more of the legwork in exploration can increase subsurface understanding.

Automation has dramatically redefined the economics and productivity levels across many industries, including manufacturing, automotive, airlines, pharmaceuticals and banking. In most cases, the "automation wave" in these industries was preceded by extensive mechanization targeted at increasing individual productivity. The technical and economic feasibility of automation depends upon the computing device's ability to dramatically expand human capabilities for productivity, accuracy and repeatability. How might the growing automation trend in other industries impact exploration and production (E&P) petrotechnical processes, productivity and manpower shortages?
To date, we have seen a few isolated implementations of automation in our geophysical workflows, notably in the areas of seismic event detection. Looking ahead, the challenge is to understand which petrotechnical tasks are good candidates for substantial automation, which might effectively use targeted automation to support individual productivity, and in which cases automation is of limited or no value.
Waves of automation
The Industrial Revolution brought extensive mechanization across the textile, iron and steel, transportation, mining, and other core industries of the 18th and 19th centuries. In the 20th century, Henry Ford's moving assembly line in the automotive industry and a wide array of business machines across banking and other industries continued the trend to work process mechanization. The advent of computer technology over the past half-century has powered a broad wave of automation that is streamlining or replacing many mechanized work processes. Automation has dramatically redefined the economics and productivity levels across many industries, for example: machine vision inspection for manufacturing, pharmaceuticals and automotive; robotic assembly lines for automotive and heavy industries; automated ticketing for airlines and sports/entertainment; product tracking for retail and manufacturing; and automated teller machines for banking.
Automation value
What compelling value does automation deliver across so many industries to justify billions of dollars of investment? Automation can provide substantial business benefits through cost-effective provision of superior speed, accuracy and reliability. In other words, automation is accepted when it is much faster, much more accurate and much more reliable than the manual or mechanized status quo. The application of process automation is generally most successful for simple and repetitive tasks that are impossible, or at the least, challenging for the average worker.
Machine vision sensing is an example
of a key emerging automation component for manufacturing, automotive and pharmaceuticals industries. These vision-sensing systems typically integrate optical sensing hardware with image processing and pattern recognition software to identify manufacturing defects and substandard components. Specific applications include detecting product over/under filling, labeling errors, product contaminants
and structure defects. For a modest economic investment, these automated systems deliver value through increased productivity, reduced waste and improved product quality.
History of E&P automation
While mainframe computer systems dramatically changed the face of seismic processing from the mid-'60s onward, seismic interpretation remained a manual paper-and-pencil activity for many years. However, the popularization of 3-D seismic quickly overwhelmed the traditional manual interpretation approach. Initial attempts to mechanize these workflows in the early 1980s resulted in technology innovations like GSI's SeisCrop Table.
By transcribing 3-D seismic data to film strips, seismic interpreters were able to quickly scroll through seismic horizon slices and perform dynamic mapping on overlain Mylar sheets. This early approach to seismic interpretation mechanization was soon overshadowed by the introduction of the computer interpretation workstation, ushering in the era of Computer Aided Exploration (CAEX) (Figure 1). Led in particular by a small, Houston-based technology company, Landmark Graphics Corp., the CAEX revolution completely changed the productivity equation for 3-D seismic interpretation. However, these systems did more than just mechanize the visualization and interpretation of 3-D seismic volumes - they introduced automated seismic event tracking. The extremely tedious task of interpreting every seismic line in a volume was replaced by automated extrapolation of an interpreted sparse grid or individual "seed points." Over the years these initial rudimentary seismic event trackers have evolved to more sophisticated waveform and volume trackers. Other attempts at automating seismic workflows have been much less successful, often resulting in mystical "black-box" techniques. While this paper is highlighting automation successes with seismic data, we are also beginning to see automation emerge in a number of drill path and field development planning approaches.
Automated fault interpretation
After several years of mediocre results, recent advancements in automated fault interpretation are beginning to generate outstanding results. An intuitive and adaptive approach that seems to be particularly promising has been developed by the BP Center for Visualization. This technique is very analogous to machine vision sensing, combining image processing and pattern recognition technology. Image processing is used to highlight discontinuities in the seismic data and identify potential fault surfaces. Fault segment patterns are then automatically detected on time-slice surfaces and are presented to the interpreter in an azimuthal context. By automating the time-consuming and repetitive task of picking individual fault segments, intellectual power is focused upon applying geologic knowledge to the creation of fault plane surfaces. The productivity gains associated with these new workflows provide opportunities for more detailed and accurate fault analyses and a greater focus upon the overall reservoir structure and contents. The interpreter is saved from menial work, and valuable time is refocused to expert activities that better capitalize on the skills and imaginations of geoscientists.
Automated velocity analysis
Exciting automation innovations are also emerging in prestack seismic processing workflows, specifically in the area of velocity analysis. Implementations ranging from "function tuning" to multiple prestack imaging iterations are becoming commercially available. The AutoImager approach from Data Modeling Inc. is an interesting automation technique which also shares similarities with machine vision sensing. In the same way that vision sensing supports 100% quality inspection, this novel approach supports automated, high-density velocity analysis at every surface location and time sample if desired. Rapid prestack imaging provides both data quality control and a tool for estimating velocities through multiple 3-D prestack time imaging iterations. Automated velocity analysis transforms a very tedious semblance or stack-power picking process into an approach where intellectual power is focused upon discerning image quality and geologic plausibility. The value of these new automation techniques lies not just in reduced project cycle time but also in a number of benefits derived from the creation of a high-density velocity field, including better horizon and fault imaging, more accurate prestack/ poststack inversion, and more accurate pressure prediction, as well as support of emerging workflows for 4-D and multicomponent seismic processing and interpretation.
Trust but verify
As we adapt our E&P processes to a new wave of automation tools, we can borrow a reminder from a bygone era - trust but verify. As we grow to trust our automation tools to produce objective and consistent results, it is essential to focus our attention upon verifying these results within a subsurface, geologic context. Learning from our failures associated with simply drilling "amplitude vs. offset anomalies" or "dominant frequency attributes," we need to use automation tools as a means to identify geologically sound prospective reservoirs and drilling targets. In particular, "black box" approaches that employ artificial neural networks and other computer learning techniques will need to provide analysis tools to support easy and robust user validation.
Social issues
In a number of E&P forums, particularly in the United States and Canada, we hear ominous presentations about the "demographic crisis" or "big crew change." The replacement of mundane work processes with automation would seem to be a desirable and perhaps an essential development to maintain required levels of productivity into the future. However, similar transformations in other industries have brought tremendous disruption to the individual workers and periods of widespread angst. As these industries restructured, they generally found that a recasting of roles resulted in better opportunities for applying human intellectual power. In the future E&P industry, we need our best minds, globally, to focus upon the increasingly challenging task of finding, developing and extracting petroleum from increasingly smaller and more complex reservoirs. As an industry, we should embrace technology that will remove the repetitive and mundane from our work processes and free up our brain cycles for creativity, intuition and decision-focused tasks.
Future E&P automation
With the proven success of automated horizon tracking and more recent automation innovations in volume and fault interpretation and velocity analysis, what lies ahead? We should expect a maturation of computer learning techniques such as artificial neural networks, delivering on some of the original but unfulfilled promise of artificial intelligence. In particular, we should see the emergence of more accurate pattern recognition techniques, leveraging work in biometrics (face and fingerprint recognition and the emerging multimodal biometrics technology) as well as other industries. As we learn to extend these pattern recognition techniques into 3-D, 4-D and beyond, a greater sophistication will emerge in our ability to track seismic characteristics associated with not just lithology and fluid but also stratigraphy and petroleum system components. Beyond seismic we will begin to see better automated optimization of individual well plans and full field development scenarios, including targeting, path layout and completions design. Over time, the office automation we have discussed here will merge more and more with the E&P operations automation being fueled by digital oilfield initiatives. Automation technology is beginning to play an increasing role in the E&P industry, and the results will depend upon our ability to effect change and manage the associated social and technology implications. We must choose wisely. The best solutions will empower E&P professionals to greater success while making our lives more challenging and rewarding.