Halliburton’s ‘hybrid’ digitaltwin

Combined physical and data driven offering goes beyond ‘fashionable’ neural networks.

No, it’s not Frankenstein. Halliburton’s hybrid digital twin (HDT), as revealed in a recent white paper represents an attempt to marry the two worlds of forward, physical modeling and data-driven analytics*. The authors recognize that in oil and gas, ‘modeling has existed for decades and the availability of high-performance computing and software tools has allowed for their widespread acceptance.’ But HPC can’t simulate everything. There is a need to couple physical models with data-driven analytics into a ‘hybrid digital twin.’

Halliburton proposes a digital twin for predicting future behavior and performance of the physical asset, and even a digital twin for ‘system of systems thinking,’ to cater for ‘interoperability and emergent behaviors.’ Physical models are routinely based on engineering assumptions and validated on limited data sets. Such weaknesses can be offset with data-driven models. The adaptive nature of the HDT is claimed to provide ‘significant benefits’ in well construction and production planning, ‘where variations between wells and fields are the norm.’

The authors argue against the use of ‘fashionable’ neural networks alone. Neural nets ‘only comprehend’ the data and ignore the underlying physics. The HDT promises a system that understands both and ‘will have a widespread impact.’

The HDT leverages concepts developed by Matthew Franchek of the University of Houston. See for instance his work on BOP condition monitoring SPE 189987-PA.

* See also our editorial on this subject.

Click here to comment on this article

Click here to view this article in context on a desktop

© Oil IT Journal - all rights reserved.