A new research grant awarded under the United States Department of Energy’s Small Business Innovation Research (SBIR) program aims to develop technologies that can harness large data sets to help drillers make time-critical decisions. E-Spectrum Technologies, a supplier of technology-driven telemetry solutions for upstream energy markets, in partnership with the Harold Vance Department of Petroleum Engineering at Texas A&M University, has been awarded a Phase I grant to begin development and commercialization of a machine learning-based drilling optimization system.
Such machine learning algorithms are being used in turbomachinery design, too, where solving a physics equation may well be too time-consuming and expensive.
The objective of the grant is to develop and commercialize a real-time computer advisory system to help drillers make more effective decisions and optimize the Rate of Penetration (ROP) achieved during drilling operations. The advisory system will use transformational digital technologies such as distributed processing and machine learning techniques to quickly identify ongoing or incipient vibration and loading patterns that can damage drill bits and slow the drilling process. Features of the drilling advisor include the ability to: operate in geothermal wells at temperatures up to 250°C; perform downhole bit dysfunction identification using machine learning; and transmit near-bit data using high-speed short-hop EM telemetry.
The system will use machine learning algorithms hosted on a near-bit embedded computer to identify incipient bit dysfunctions and pass them to an MWD telemetry system via a high-speed EM short-hop link. This bit dysfunction information will be transmitted to the surface using E-Spectrum’s Drill Dog MWD telemetry platform which is being thermally upgraded under a separate research grant to operate at temperatures up to 250°C while retaining compatibility with E-Spectrum’s proprietary suite of signal denoising algorithms.
At the surface, a PC-based dynamic advisory application using machine learning and data mining algorithms will integrate this incipient failure information with numerous other data streams to provide drillers with critical real-time advice on how to set drilling parameters in order to optimize ROP and avoid damaging drill string components.
While the Phase I research is targeted at ROP optimization, the advisory system will be designed to be modular and scalable to allow future incorporation of expanded data inputs and data driven algorithms to address other drilling problems that are currently difficult or impossible to solve using physics-based models.
