Soft Measurement of Water Content in Oil-Water Two-Phase Flow Based on RS-SVM Classifier and GA-NN Predictor

Measurement of water content in oil-water mixing flow was restricted by special problems such as narrow measuring range and low accuracy. A simulated multi-sensor measurement system in the laboratory was established, and the influence of multi-factor such as temperature, and salinity content on the measurement was investigated by numerical simulation combined with experimental test. A soft measurement model based on rough set-support vector machine (RS-SVM) classifier and genetic algorithm-neural network (GA-NN) predictors was reported in this paper. Investigation results indicate that RS-SVM classifier effectively realized the pattern identification for water holdup states via fuzzy reasoning and self-learning, and GA-NN predictors are capable of subsection forecasting water content in the different water holdup patterns, as well as adjusting the model parameters adaptively in terms of online measuring range. Compared with the actual laboratory analyzed results, the soft model proposed can be effectively used for estimating the water content in oil-water mixture in all-round measuring range.