Analysis of gas–liquid behavior in eccentric horizontal annuli with image processing and artificial intelligence techniques

In drilling operations estimation of gas–liquid behavior such as flow patterns and liquid holdup is beneficial in terms of cost, time and efficient usage of resources for the well to be opened. There is a lack of research for hydraulic behavior of two phase fluids in annular geometries. One of the aims of this study is to observe the flow patterns experimentally in two phase eccentric annulus. The second aim is to detect the liquid holdup of these flows using digital image processing techniques instead of emprical correlations or mechanistic models. The last aim is to estimate the flow pattern and liquid holdup for two phase (air and water) flow in horizontal eccentric annulus. This is conducted by using artificial intelligence techniques rather than conventional mechanistic models. In this study, nearest neighbor algorithm, backpropagation neural networks, and decision trees are used as the artificial intelligence techniques. Flow is generalized by representing the flow patternsas superficial Reynolds numbers for both liquid and gas phase. The results showed that the back propagation neural network model provided the best results as an estimation model for flow pattern identification whereas regression decision tree had the best performance for liquid holdup determination. In air and water flow, 7 observed flow patterns are classified correctly with an accuracy of 90.38% and liquid holdup is estimated with an average absolute percent error of 17.06%.