Optimization of differential pressure signal acquisition for recognition of gas–liquid two-phase flow patterns in pipeline-riser system

Eighteen differential pressure signals were investigated for the recognition of gas–liquid two-phase flow patterns in a long pipeline-riser system. The recognition was performed by a BP neural network based on the multi-scale wavelet analysis of either single or combine signals. In order to evaluate the performance of different signals for recognition, three parameters were proposed, namely the recognition rate, the measuring length (distance between the pressure taps) and the measuring position. The effects of the measuring length, the measuring position, and the geometric shape of the measuring section on the recognition rate were analyzed. Recognition rates of the signals on the horizontal pipeline were weakly correlated with the measuring length and the measuring position. While for the signals on the inclined sections, the recognition rates were influenced by the measuring position. Both the optimal single signal and optimal combined signals were obtained for the fast recognition of flow patterns.