Wet gas metering using a Venturi-meter and Support Vector Machines

Abstract

A new approach to the measurement of wet gas flows is introduced in this paper. Support Vector Machine (SVM) was employed in wet gas metering. Typical features were extracted from the signals obtained by a throat-extended Venturi meter. The features and the corresponding flow rates (targets) were used to train the SVM model. The trained model was then used to predict the flow rates of wet gas. Experimental results suggest that this method provides a solution that is much better than the empirical formulas. The average prediction error of this method is smaller than that of the empirical formulas by about 50%. This method is also proved to be better than the technique using a venturi-meter and neural network.