Prediction of start-up yield stress of gelled crude oil by support vector regression

Key words:  support vector machine  nonlinear regression  start-up yield stress  gelled crude oil  pipe flow experiment

Abstract:

The start-up yield stress of gelled crude oil is affected by many factors such as start-up temperature, start-up flow rate, temperature drop and shutdown duration. Support vector regression (SVR) is a common and efficient technique for nonlinear multivariate analysis. Based on the data collected on the experimental loop for oil rheological property research，the startup properties of gelled crude oil were studied under different start-up temperatures，start-up flow rates，temperature drops during shutdown period and shutdown durations. In addition, the SVR method was applied to the experimental data, and a new formula with improved precision was obtained to predict the start-up yield stress of gelled crude oil. The contrast result between the prediction value and the measured data validated the reliability of the formula.