LSSVM model of oil holdup of oil-water two phase flow using thermal method based on HHT

Abstract

A new model based on least square support vector machines (LSSVM) and Hilbert-Huang transform (HHT) has been proposed for the first time, which is capable of forecasting oil holdup of oil-water two phase flow. Owing to the temperature signal of oil-water two phase flow is greatly disturbed by noises, which results in a limited measurement range of oil holdup. In order to solve the problem, a new signal processing method based on the improved HHT is used. With its ideal performance on local adaptability and time-frequency analysis, noises are removed. Experimental studies were carried out to compare the HHT with the wavelet transform. The instantaneous fluctuated amplitude and the standard deviation are obtained form the last residue component of HHT, together with total flux were employed as inputs of LSSVM model. In order to improve the predictive accuracy of the LSSVM model, a Genetic Arithmetic (GA) has been adopted to determine the optimal parameters of LSSVM automatically. The experiment results show that the average measurement error of LSSVM model was 0.542% in the range of 8% to 90% oil holdup.