Ultrasonic level sensors for flowmetering of non-Newtonian fluids in open Venturi channels: Using data fusion based on Artificial Neural Network and Support Vector Machines

In drilling operations related to oil & gas or geothermal applications, the improved monitoring and control of the flowrate of drilling fluid is important. This will help in reducing cost as well as improving system and Health, Safety and Environmental (HSE) performances. A relatively accurate and low-cost flow monitoring system functioning as a supervisory unit for the drilling fluid in the return path would be useful for this purpose. Inclusion of appropriate sensors and modifying the geometry of an already existing open channel in the transport of drilling fluids are possible approaches for estimating the flowrate of the drilling fluid. Forming a Venturi flume in the already existing open channel structure of the transporting conduit for the drilling fluid offers some interesting possibilities. Using a set of three ultrasonic level meters for determining the levels at various points in the open channel and fusing the data from other sensors in the test loop, the flow rate in a Venturi channel is successfully estimated. Two different empirical models using Artificial Neural Network (ANN) and Support Vector Machine (SVM) are used as alternatives to the mass balance approach. For the flowrate of drilling fluid in the range of (250-550) kg/min, the performances of ANN and SVM models are much better than that of the mechanistic model. The sampling rate for SVM is about 90 times more than that of the mechanistic model. Whereas, the sampling rate of ANN is about 100 times more than that of the SVM model. The Mean Absolute Percentage Error (MAPE) for both empirical models is less than 2%.