In drilling operations, the rheological properties of the drilling fluid (‘mud’) are measured at rigsite at ambient temperature using sensors in real-time. These measurements, however, are required to be reported at the API standard temperature (120/150 oF). As the rheological properties of the drilling fluids vary significantly with temperature, it is essential to calibrate sensor measurements at the ambient temperature to the API standard. Previous attempts in the literature to build data-driven frameworks for predicting drilling fluids behavior demonstrate limited success due to restrained data access, neglect of the physics, and/or use of improper algorithms, such as neural networks which are shown to perform poorly despite their popularity. In this work, we develop a digital twin to calibrate the rigsite rheology measurements for obtaining the API standard properties by exploring the use of a set of promising machine learning algorithms. A dataset composed of various drilling fluids composition with rheological measurements at both rigsite and API conditions is collected. Our results demonstrate that the ensemble algorithm outperform other commonly used methods, such as regularized regression, polynomial regression, and neural networks. The optimized integrative model is deployed on a platform at rig for use at real-time drilling operations.

This content is only available via PDF.
You can access this article if you purchase or spend a download.