Estimating Rheological Properties of Non-Newtonian Drilling Fluids Using Ultrasonic-Through-Transmission Combined with Machine Learning Methods

During drilling an oil/gas well, situations where the pressure integrity of the well is lost are great risks to assets, humans, and the environment, and represent high costs to the drilling operation. The non-Newtonian drilling fluid that is circulated while drilling serves several purposes, and its properties are essential to fulfil them. The density (p)enables the drilling mud to maintain pressure integrity during drilling operations. The plastic viscosity (μ p)enables the fluid to transport cuttings, while the gel strength (S)keeps the cuttings suspended when circulation is stopped. The industry standard is to perform manual drilling fluid checks every six hours to measure these three important rheological properties. To increase safety, efficiency and enable more automated drilling operations, sensor technology to perform these measurements in real-time is in high demand. Non-invasive ultrasonic measurement techniques in combination with machine learning represents one promising and easily implementable solution to meet these demands. In the presented work acoustic properties of different drilling fluids are studied. Three different pairs of ultrasonic transducers were utilized to evaluate the propagational properties of 22 different samples of water-based drilling fluids. The parameters measured in the ultrasonic experimental setup is received signal amplitude, time of flight and the lateral distance between receiver and transmitter. The trained machine learning models predicted density with a mean absolute percentage error (MAPE)in the range of 0.84% to 0.95%, plastic viscosity from 4.4% to 7%. The models for gel strength were not nearly as accurate, with MAPE ranging from 15% to 19%. This shows that the measurement principle has potential to develop a sensor system capable of meeting the demands regarding safety, efficiency and automation.